r AD-AO84 326 OPERATING SYSTEMS INC WOODLAND HILLS C -A F/ 9/4 M
UL SATELLITE AND MISSILE DATA GENERATION FOR AIS.(U)
DEC 79 6 M SILVA, C A MONTGOMERY F30602-78-C-0274
fCLASSIFIED OSI-R79037 RADC -TR79314 IN.
IIIIIIIIIIIIIIIIIIIIIEEEEIhEEEEEEIhIhEEEEEmuIIIIIIIIIuuuIIIIIIuuuluuu,I EEE...'..lI
SATELLITE AND MISSILE DATASGENERATION FOR AIS
Operating Systems, Inc.
S Dr. Georgette M4. T. SilvaDr. Christine A. Montgomery
APPROVED FOR PUBLIC RELEASE; DISTRIBUTION UNUMITER
MAY 19 9 I80
AROME AIR DEVELOPMENT CENTERAir Force Systems Command
C),.f Griffiss Air Force Base, New York 13441
805 6 .0
This report has been reviewed by the UDC Public Affairs.Off lea (iA)andi4 1.releasable to the National Technical Information Service (NTIS).At NTIS it will be releasable to the general public, including forsig-nations.
RADC-TR-79-314 has been revieVkf and Ta approved for publication.
APPROVED:ZBIGNIEII L. PARKOVICZProject Engineer
APPROVED:
HOWARD DAVISTechnical DirectorIntelligence &Reconnaissance Division
FOR HE CMMANER: ing Chief, Plans Off ice
If your address has changed or if you wish to be removed from the RADCmailing list, or if the addressee is no longer employed by your organiza-tion, please notify RADC (IRDT), Griffiss AFB NY 13441. This will assistus in maintaining a current mailing list.
Do not return this copy. Retain or destroy.
UNCLASSIFIEDSECURIT 4S6SIFICATION OF THIS PAGE (When Does Enterld)
[ Rq) REPORT DOCUMENTATION PAGE BEFRE D Ls TIGFREP GOVT ACCESSION NO. ECIPIENT'S CATALOG NUMBER
6 SJATELLITE AND MISSILE DATA GENETIO FOR A . / Fialechnical et..... ...... i ............ ...:- .......- r I Sep 78--31 Aug7,/
* Georgette M.T.ISilva F3466-7-C-uJI4Christine A. Montgomery
9. PERFORMING ORGANIZATION NAME AND ADDRESS 1. PROGRAM ELEMENT. PROJECT. TASK
Operating Systems, Inc. AREA & WORK UNI
21031 Ventura Boulevard
Woodland Hills CA 91364 F
,. CONTROLLING OFFICE NAME AND ADDRESS - 9
Rome Air Development Center (IRDT) Decm'=MGriffiss AFB NY 13441 "_9UMSER OF PA
14. MONITORING AGENCY NAME & ADDRESS(If dillfer* 5. SECURITY CLASS. (of this report)
Same " UNCLASSIFIEDSame S.. DECLASSIFI CATION DOWNGRADING
N/ASCHE D
ULE
16. DISTRIBUTION STATEMENT (of this Report)
Approved for public release; distribution unlimited.
17. DISTRIBUTION STATEMENT (of the abstracI entered in Block 20, if different from Report)
Same
IS SUPPLEMENTARY NOTES
RADC Project Engineer: Zbigniew L. Pankowicz (IRDT)
19 KEY WORDS (Continue on reverse aide if necessary and identify by block number)Intelligence Data Processing Computational LinguisticsAdvanced Indicator System Content AnalysisEvents Reports Discourse ProcessingRecord Generation Message Text ProcessingArtificial Intelligence Missile and Satellite Context
20 ABSTRACT tConflnue on reveres Ode If necessary end Identify by block nmrnber)This final technical report describes a 12 month R&D effort under ContractF30602-78-C-0274, "Satellite and Missile Data Generation for AIS". Extendedabstract included in the Report. All examples of messages and message frag-ments (sentences) in the report are based upon facts and events described inthe daily press, specifically the Los Angeles Times (July 1979), Aviation Week(February 1978 - July 1979), American and Soviet Military Trends since theCuban Missile Crisis by John M. Collins, published by the Center for Strategicand International Studies, Georgetown University, Washington DC, 1978 (Cont'd)
DD " AN7 , 1473 UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE (ahen Dat Ert.r
9j9Y-
-__- - ---..--
UNCLASSIFIEDSECURITY CLASSIFICATION OF THIS PAGE(17,w Date Entoed)
Item 20 (Cont'd)
and Jane's All the World's Aircraft, 1976 -1977. The messages used as
examples resemble the actual messages only in their grammatical structure.Most objqcts, attributes, and time/location parameters used refer to theevents connected with the launch, mission, deorbit, breakup and impact of
0 K. ow
UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE(91en DWG. Efn#*e"
ABSTRACT
This final report presints the results of work performed under RADC contract No.F30602-78-C-0274. he effort described specifically addressed the problem ofderiving Indicator and descriptor data from the narrative text portions of a class ofIntelligence messages dealing with events related to missile and satellite launchingsused for Input to the Advanced Indicator System (AIS) data base.
The introductory section briefly discusses the intelligence problem which OSI'sevent processing technology intends to solve, summarizes the technology developedunder this contract, and presents the conclusions drawn on the basis of the resultsobtained.
Section 2 offers a summary of OSI's methodological approach to the analysis anddescription of event reports. This methodology, initially developed on the basis ofmessages dealing with air activities, was, under this contract, extended to coverreports of events involving missile and satellite launchings and related events./ Sub-section 2.2 presents the characteristics of the maximal unit of analysis: the EVENTREPORT, while subsection 2.3 discusses the characteristics of messages from thepoint of view of their conceptual organization and that of their linguistic organization,and gives details of the analytical procedures adopted for their analysis. Subsection2.4 discusses two representational constructs of fundamental importance in eventprocessing: the Template, and the Event Record. Subsection 2.6 outlines someIssues involved In the problem of reference, while subsection 2.6 provides guidelinesfor the establishment of a research corpus.
,'Section 3 describes the Missile and Satellite domains, and presents the results oftheir analysis In terms of a domain definition. The discussion includes a characteriza-tion of the event report In terms of its component messages; a list of the messagetypes encountered in the domains under consideration; a list of the event typesIdentified together with their descriptor system, and a definition of the sublanguageIn terms of Its vocabulary and syntax.
Section 4 focuses on the Implementation of OSl's message text analysis system,MATRES II. It briefly reviews the principles underlying OSI's event processing techo-nology and offers an overview of MATRES II. he computer programs which embodyOSI's approach to the automated analysis of ssage text are written in FORTH, Pro-log, and SNOBOL4, and run on a PDP 11/46 tnder the RSX 11D operating system.Finally, the analytical processe4 utilized by I4TRES II are illustrated by means ofexamples.
Mi
CONTENTS
3.2.3.2 The Grammar ......................................................... 93.2.3.2.1 The Declarative Sentence .......................... 10
3.2.3.2.2 The Simple Sentence ................................ 10I.3.2.3.2.3 The Noun Phrase ..................................... 11
3.2.3.2.4 Nominallzatlons......................................... 123.2.3.2.5 Relative Clauses ..................................... 133.2.3.2.6 Noun Phrase Apposition........ ........ o........143.2.3.2.7 The Verb Group......-....................... o....... 183.2o3.2.8 Adverblals ....... o .... o............... ....... 163.2.3.2.9 The Concepts of Time and Space.......... ..... 173.2.3.2.10 Location References.... ........... ............ 193.2.3.2.11 Complementation ........................o ............o...193.2.3.2.12 Passive Sentences.........o ... o......... o......... 213.2.3.2.13 Ambiguity.,...- ............... o_.................. o... 213.2.3.2.14 The paraphrase Problem ......ooo....... o.......... 213.2.3.2.15 Functional Synonyms...... .o..................... 23
3.2.3.3 The Lexicon.....o..........o................................ .......... 25
4.0 IMPLEMENTATION ...... ...... ............. ............. 14.1 Principles of Discourse Processing.. .- 0.......... ............... ............ 14.2 The MATRES 11IText Processing System.......... ............................ 2
4.2.1 General Remarks. ... 0............ ...... ..............000 ........ 24.2.2 Functional Description ......... .0...... 0............. .............. 74.2.3 The Event Representation Language (ERL).-... o....-..... o........... 94.2.4 The ERL Control Mechanism....... ......... . . . 00..0........... ......... 114.2.5 Advantages of Prolog Representation .............. .. .............. 1
4.3 Illustration of the Understanding Process ..o..... ...... 0.............. ...... 124.3.1 Inputting a Sentence. ... o....-... .... o.... o........ o....... ..... ...... 124.3.2 Parsing a Sentence..... ....... ........ ......-. o............ o.o.... .. 134.3.3 Interpreting the Parse Tree......o.... ....... 000.......... 0............ ..... 1 74o3.4 Identifying the Reported Source of an Event Report.............. 204.3.5 Identifying the Reported Status of an Event.......... o............. .... 21
5.0 REFERENCES,........................ ........................ ..... 5.1
Vi
CONTENTS
1.0 INTRODUCTION AND SUMMARY ............................................................................... 1
1.1 Introduction .................................................................................................... I
1.1.1 Problem Statement .............................................................................. 1
1.1.2 Toward a Solution ................................................................................. 3
1.2 Summary ......................................................................................................... 6
1.2.1 Scope of Current Effort ........................................................................ 6
1.2.2 Current Capabilities of OSI's Message Text ProcessingSystem. 7
1.3 Conclusions .................................................................................................... 9
2.0 METHODOLOGICAL APPROACH ............................................. 12.1 Preliminary Notions ...................................................................................... 12.2 Characteristics of EVENT REPORTS .............................................................. 42.3 Characteristics of Messages ...................................................................... 5
2.3.1 The Conceptual Organization of Message Text .................................. 02.3.2 The Description of Message Content ................................................. 72.3.3 The Linguistic Organization of Message Text ..................................... 82.3.4 Description at the Event Level ........................................................... 9
2.3.4.1 Vocabulary Analysis ............................................................ 9
2.3.4.2 Lexical Analysis ................................................................... 102.3.4.3 The Study of Text Level Relations ........................................ 14
2.4 Representational Issues .............................................................................. 172.4.1 The Template ..................................................................................... 172.4.2 The Event Record ........................ .............. 21
2.6 The Problem of Reference .......................................................................... 232.5.1 Definite Referring Expressions ........................................................... 24
2.6.1.1 Pronominal Reference ................................. 242.6.1.2 Definite Noun Phrases ....................... I ................................... 25
2.6 Guidelines for the Establishment of a Research Corpus ............................... 262.6.1 Data Collection .................................................................................. 262.6.2 Data Sanitization ................................................................................ 272.6.3 The KWIC Concordance ....................................... 28
3.0 MISSILE AND SATELLITE DOMAIN DEFINITION ................ 8 ....................................... 13.1 The Missile and Satellite Worlds ................................................................... 13.2 Domain Definition ......................................................................................... 2
3.2.1 Research Corpus ................................................................................ 23.2.1.1 Source Data Sanitization ....................................................... 23.2.1.2 Examples of Transcribed Text .............................................. 3
3.2.2 The Structure of EVENT REPORTS ....................................................... 43.2.2.1 Message Types .................................................................... 53.2.2.2 Event Types and other Concepts .......................................... 63.2.2.3 Descriptor System ................................................................ 7
3.2.3 Language Definition ............................................................................ 93.2.3.1 Relationship between General English and the Reporting
Language. 9
APPENDICES
Appendix A - Sanitized Message Text ....................... A-i
Appendix B - Lexicon ...................................... B-1
Appendix C - Templates and Auxiliary Procedures ........... C-1
Appendix D - FSA Listing .................................. D-1
Appendix E - Examples of System Input/Output .............. E-1
vii
EVALUATION
The objective of this effort consists in augmenting and extendingthe capabilities of the existing message text processing methodologyto the subject domain of satellite and missile sightings. The workdescribed deals with the analysis of textual reports of events andthe synthesis of relevant information elements in a format suitablefor automated input to the AIS database.
The effort is directed at providing coypputerized aids to the I&Wanalyst in distilling the contents of incoming text messages intocompact, formatted, computer processable content representations insupport of his mission to predict the future on the basis of informationdescribing past and present events. The analyst's difficulties resultfrom the fact that the volume of message traffic is normally veryhigh and increases sharply in a crisis situation. This preventsefficient handling and full exploitation of the enormous amount ofvariables contained in the message traffic under both normal andcritical operating conditions. Since a computer experiences nodifficulty in processing large numbers of variables, the notion ofautomating this task provides a logical solution in the context ofinformation explosion.
The significance of the subject effort consists in computer modelingof the analyst's cognitive activities in reading and understandingmessage text, transforming its contents into information items ofinterest, and building a conceptual model of the information conveyedin the message. In order to accomplish this task, the computer mustbe equipped with representations of both linguistic and extra-linguistic knowledge inherent in cognitive faculties of the analyst.The approach to computer modeling of understanding relies heavilyon the recent and current theoretical advances in computationallinguistics, language theory, artificial intelligence and cognitivepsychology.
ZBIGNIEW L. PANKOWICZProject Engineer
viii
Li%
I!
1.0 INTRODUCTION AND SUMMARY
1.1 Introduction
This final report presents the results of work performed under RADC contract No.
F30602-78-C-0274. T ie effort described specifically addressed the problem of deriv-
Ing indicator and descriptor data from the narrative text portions of a class of intelli-
gence messages dealing with events related to missile and satellite launchings used for
Input to the Advanced Indicator System (AIS) data base. The following sections briefly
discuss the Intelligence problem which OSI's event processing technology Intends to
solve (1.1.1.), and summarize the development of the technology thus far (1.1.2.)
1.7.1 Problem Statement. The task of an intelligence analyst is to predict the future on
the basis of Information describing what has happened in the past and what events are
currently taking place.
At the global level, the questions the analyst asks himself are: "What is happening?"
"What does It mean In terms of my knowledge about similar events In the past?", "What
Is going to happen next"? He is concerned with certain states of affairs, and events
signifying changes In these states of affairs.
When working with a single message, the analyst seeks answers to at last the following
questions:
1. What Is Its Information content?
2. How reliable Is the source?
3. How "credible" Is the data?
His evaluations of Incoming Information are bas3d on his cognitive models of certain kinds
of situations, the personalities, entities, and processes Involved, and the potentialities
1-1
and constraints associated with changes In an existing state of affairs.
Given the volume of Information he must sift, and the complexity of the cognitive models
involved, the difficulties of the analyst's task are obvious. Aids to support his analytical
processes clearly must involve means for distilling the content of incoming information
Into a form which is compact, usable, and compatible with his view of the world.
Information on the world situation comes to the &W analyst mainly in the form of intelli-
gence messages, which are electrically received in an I&W center 24 hours a day. The
messages come from many different originators, and are largely in the form of narrative
text. The volume of message traffic is extremely high, and in a crisis situation,
Increases dramatically. Even under normal operating conditions it is very difficult for an
analyst to Isolate items of information from message text and to assimilate and correlate
these Items into a pattern of events of indications significance. In a crisis situation, the
analyst is completely saturated with data, and the performance of his task demands
superhuman capabilities for handling the enormous number of variables which are con-
tained in the message traffic.
A computer, on the other hand, can process large numbers of variables. Thus, the notion
of offloading some of the variable processing functions onto the machine seems to pro-
vide a logical solution to the information problem.
One of the interesting developments in this direction is the Advanced Indications System
t" (AIS), which currently has the capability to statistically analyze Intelligence data and to
display such data in a logical and useful form.
Briefly, the AIS provides the following:
a. A structure for continuous objective and systematic monitoring of selected indicator
time series.
1-2
b. Computer based logic for detecting the significant patterns in current data, compar-
Ing them to past activities, and quantifying the departure from normally observed
activity.
c. The capability to interact at a single point with a comprehensive data base to
assimilate, investigate, present and resolve the unusual situations detected.
At the current stage of development, the messages to be analyzed by the AIS are manu-
ally sorted into functional threat focused Indications categories. Data elements called
'indicators' and 'descriptors', which are n-ary structures of various types*, are then
derived from relevant segments of certain messages, and subsequently entered into the
system's data base for the daily update. Due to the experimental and developmental
status of the AIS, update of the descriptor/indicator data base is currently performed
manually by contractor personnel. When the system achieves operational status, update
of the AIS data base will be performed by an I&W analyst. Despite the considerable
benefits he will derive from the AIS, maintaining the AIS data base constitutes still
another task for the overburdened analyst.
1.1.2 Toward a Solution. For the past several years RADC has been sponsoring an
exploratory and developmental program related to the design and development of a gen-
eral methodology for the efficient and effective exploitation of the content of electri-
cally transmitted intelligence messages. The long term goal of this work is to develop a
system which would assist the analyst in creating and maintaining formatted data bases
dorived from natural language text, and thus offload some of the processing functions
from the analyst to the computer. Such a system should provide the analyst with infor-
mation which is needed for the attainment of his particular goal, i.e., Information which is
These are described in detai: in a classified appendix to the OSI Findl TechnicalReport RADC-TR-77-194, June 1977.
1-3
relevant to his task, Is of high epistemic standing, and therefore useful to solving his
problems.
Figure 1-1 shows the components of a total system for message exploitation and
highlights the focus of the RADC program.
As mentioned above, the work described ih this report is concerned with the analysis of
textual reports of events and the synthesis of relevant information elements in a format
suitable for automated input to the AIS data base.
Specifically, the program addresses the problem of automating the analysis of the narra-
tive text portions of intelligence messages describing events, with the aim of transform-
Ing them into succinct, formatted, computer processable content representations.
The automated generation of information elements from narrative message text requires
that the computer in some sense "understand" natural language text. Within the con-
text of the work described here, we say that a computer system understands an input
text Insofar as it can construct an adequate representation of the information content
of that text. Specifically, we require that the output of the computer understanding
process, when applied to some message text, furnish the analyst with at least those
Information elements that he would himself have extracted from that particular text.
OSI's approach to the problem of computer "understanding" leans heavily on theoretical
advances In several disciplines, including theoretical linguistics, computational linguistics,
artificial Intelligence, text linguistics, and cognitive psychology. A survey of the field as
related to the work reported here can be found In Silva and Montgomery (1978) and Silva
et al. (1979).
1-4
.V1
e4n
-
-~ en
00x
I- 48
l..A
k
L~j L--..!: L, =
M C= EN
LOCnC.D L.
X L.) C I.-Cl= C
cc 1= C) CL
-: LLJ >
aL 0
--
1L-
-C-
V) . - - - - 3
The aim is to model the cognitive activities of the human analyst as he reads and under-
stands message text, distilling its contents into information items of interest to him, and
building a conceptual model of the information conveyed by the message.
In order to model this human cognitive activity, the computer must be equipped with
representations of both linguistic and extra-linguistic knowledge, and a means of mani-
pulating such representations for the analysis of text and synthesis of information ele-
ments. The elements must then be presented in a clear and useful format suitable for
the task at hand.
1.2 Summary
1.2.1 Scope of Current Effort. The scope of the effort described here included extend-
Ing and augmenting the capabilities of the message text processing methodology origi-
nally developed under previous contracts to the missile and satellite subject domain.
Briefly, the work involved the establishment of a research corpus; the development of a
transcription scheme for the sanitization of messages; extensions to the linguistic
methodology; extensions and additions to the ATN grammar constructed under a previous
contract to accept a wider range of linguistic structures; the refinement of the notion of
"template" -- the fundamental information structure developed for the organization and
representation of knowledge about events; the development of templates for the missile
and satellite domains; and finally, additions and extensions to the existing algorithms for
the Interpretation of narrative text and its subsequent transformation Into formal content
representations.
A major effort was devoted to the development of additional program modules to accom-
modate new syntactic construction types in the missile and satellite domains, and to the
provislon.of adequate system control.
1-6
1.2.2 Current Capabilities of 031's Message Text Processing System. The OSI message
text analysis system has the capability to digest narrative text and systematically
transform it into concise, machine' processable content representations called 'event
records', In which a message can be viewed from several perspectives: time, location,
organization involved, activity type, etc.
Specifically, the current capabilities of the system are:
a. It determines the key event'described in a message on the basis of an analysis of
Its first sentence and presents it to the analyst in a form that answers the basic
"what is happening?" question (1.1.1.).
b. It provides information useful for determining the reliability of the source by recog-
nizing and displaying the reported source of an event. For example, if an event is
reported by a foreign news agency, the name of that news agency Is displayed in
the event record under the heading Infosource.
c. It provides information helpful to evaluate the credibility of the source data, by
highlighting the probabilistic information associated with a report of an event. Words
such as 'possible', 'probable', 'successfully' constitute judgments of the originator of
a message as to the reliability of the data reported. Such words are preserved dur-
Ing processing and remain associated with the term they modify in the text. For an
example see Table 1-1 below, where the word 'successfit4ly' indicates certitude of
the successful completion of the deorbit event described.
Table 1-1 shows a hypothetical (partially transcribed) input sentence describing a major
event and the corresponding event record produced by MATRES II.
1-7
Table 1-1 Example Input and Output by MATRES II--------------------------------------------------------------------------------------------
I THE NEWSAGENCY PRESSNAME ANNOUNCED THAT THE TWOUNIDENTIFIED SATELLITES WHICH WERE LAUNCHEDFROM PLACENAMEI ON 09 MARCH 1973 AT ZULUTIME WERESUCCESSFULLY DEORBITED INTO THE PLACENAME2 ON 09 APR 1978BY THE POLITNAME ON REVOLUTION 3NMBR
Infosource= THE NEWSAGENCY PRESSNAMEEvent: DEORBITAction= SUCCESSFULLY DEORBITEDAgent= BY THE POLITNAMEObject: SATELLITE... Equipment= UNIDENTIFIED SATELLITES.. .Number= TWO...Relative= LAUNCHED FROM PLACENAME1 ON 09 MARCH 1978 AT
ZULUTIME
Location= INTO THE PLACENAME2Revolution= ON REVOLUTION 3NMBRDate= ON 09 APR 1978
--------------------------------------------------------------------------------------------
The report is divided into four major sections.
Section 2 offers a summary of OSI's methodological approach to the analysis and
description of event reports. This methodology, initially developed on the basis of mes-
sages dealing with air activities, was, under this contract, extended to cover reports of
events Involving missile and satellite launchings and related events. Subsection 2.2
presents the characteristics of the maximal unit of analysis: the EVENT REPORT, while
subsection 2.3 discusses the characteristics of messages from the point of view of their
conceptual organization and that of their linguistic organization, and gives examples of
the analytical procedures adopted for their analysis. Subsection 2.4 discusses two
representational constructs of fundamental Importance in event processing: the Tem-
plate, and the Event Record. Subsection 2.5 outlines some Issues Involved in the prob-
lem of reference, while subsection 2.6 provides guidelines for the establishment of a
research corpus.
1-8
Section 3 describes the Missile and Satellite domains, and presents the results of their
analysis In terms of a domain definition. The discussion Includes a characterization of
the event report in terms of its component messages; a list of the message types
encountered in the domains under consideration; a list of the event types identified
together with their descriptor system, and a definition of the sublanguage in terms of its
vocabulary and syntax.
Section 4 focuses on the implementation of OSI's message text analysis system,
MATRES i. It briefly reviews the principles underlying OSl's event processing technol-
ogy and offers an overview of MATRES II. The computer programs which embody OSI's
approach to the automated analysis of message text are written in FORTH, Prolog, and
SNOBOL4, and run on a PDP 11/45 under the RSX 11D operating system. Finally, the
analytical processes utilized by MATRES II are illustrated by means of examples.
Section 5 contains a list of references to books, journal articles, conference papers,
doctoral dissertations, and other publications of relevance to the work described here.
Much of the information contained In these publications has Inspired and guided the
RADC-sponsored developmental work on automated data base generation since Its
Inception several years ago.
Appendices A-E contain a listing of sanitized message text (A), a listing of the combined
lexicons for the air activities, missile, and satellite domains (B), a listing of the ERL tem-
plates and their ancilliary procedurgs as encoded In Prolog (C), a listing of the FSA char-
acter processing algorithm (D), and a set of examples of system Input/output (E).
1.8 Conclusions
OSI's message text analysis methodology, while Initially developed on the basis of a res-
tricted subject domain -- reports of air activities -- has successfully been applied to
1-9 t
I
P 1.1
two new subject domains -- those of missile and satellite reports. It has been demon-
strated to be general in Its applicability to different subject areas and therefore exten-
sible in a non-trivial manner to reports of events involving the physical movement of
objects such as aircraft, ships, missiles, and satellites.
It Is shown that the automated analysis of event data needs to take into account all
aspects of event reporting and requires a truly interdisciplinary approach. Several lev-
els of analysis are identified, each involving a different aspect of event reporting, and
each based upon different considerations.
The three sublanguage domains studied thus far consist of descriptions of events involv-
Ing aircraft activities and launchings of missiles and satellites, and related events. In all
three cases, the source data are contained in the text portions of military messages
typical of these subject domains, consisting of a report title summarizing a given event,
followed by one or more declarative sentences describing that event (and optionally,
other related events).
In all three cases, the semantics and the syntax of event descriptions are constrained
by two factors. One, by the particular subject domain, and two, by the fact that the
events described are limited to what is observable and what should be reported accord-
Ing to a reporting procedure. This results in a substantial number of participial construc-
tions of various types, complex nominalizations and agentless passives, as well as a
range of types of quantification, conjunction, complementation, ellipsis, and anaphora.
The sublanguages, although less extensive in their inventory of syntactic constructions
than event reports in journalistic narrative, nevertheless contains certain constructions
which present challenging semantic problems. Such problems include thc treatment of
"respectively" constructions, as well as certain types of definite anaphora which not
1-10
V
only transcend sentence boundaries and, In some cases, even message boundaries, but
often are of the kind that have no explicit referent In the previous discourse.
Of the three languages studied thus far, the discourse structure of the satellite reports
Is considerably more complex than that of missile reports or reports in the air activities
domain. While in air activities reports the description of a given event is often com-
pleted within a single sentence (e.g., a particular aircraft penetrated enemy airspace at
a specific location and a specific time), in missile and satellite reports the complete
specification of the properties of an event and of the object(s) involved more frequently
requires several sentences, and not uncommonly, several messages. Thus, a report on
some launch operation can consist of an Initial, rather skeletal statement, followed by
one or more messages received over a period of time which update the previous report,
adding to and sometimes changing previous specifications.
Although event reports in the air activities domain may also involve several messages,
the update problem is much simpler, since the attributes of a flight event are fewer and
less complex than those of satellite events. In any case, the boundaries of a discourse
relevant to a single event can range from a single sentence to several messages. The
problem of assembling the total mental "picture" relating to any given event can only be
approached on the discourse level.
The major impact of adding a new subject domain to the system's repertoire was felt in
two related areas: vocabulary analysis and template construction. Although there exists
a core of overlapping vocabulary for the three subject domains, domain-specific usage
of event-related terms necessitates the construction of separate templates for each
domain, even In those cases where they share the central concept.
To see this, consider the FLIGHT concept, which is shared by the three domains. In the
1-11
air activities domain, it happens to be the key concept, and therefore it has a complex
Internal structure. In the missile and satellite domains, it is referred to only tangentially
to express the duration of a mission: "THE MISSILE IMPACTED AFTER A FIVE MINUTE
FLIGHT".
The syntax of the three sublanguages, on 1he other hand, exhibited a large number of
similarities, so that the addition of the new domains only required development of a small
number of new constructions, specifically, relative clause constructions and appositive
postmodofication, which was relatively rare in the air activities reports and had there-
fore not been sufficiently developed.
MATRES II is still at an early stage of development. No attempt has been made to
represent a complete set of semantic structures for the interpretation of the whole
range of linguistic expressions occurring in the messages studied.
On the contrary, the aim was to find a small set of structures that would describe as
many event types as possible, but still be simple and compact enough to be the basis of
a manageable and understandable computer system.
This goal has to a great extent been achieved. By limiting the scope of the system's
knowledge, It has been possible to model a complex domain of practical significance, and
to Implement algorithms that "understand" this domain In a limited sense.
From the theoretical point of view, it is Important to stress that the development
described In this report is only a first step toward a formal characterization of the rela-
tionship between knowledge-based language understanding and the generation of Indi-
cator and descriptor data.
However, from the pragmatic point of view, the work carried out under the current con-
tract has demonstrated that OS's Initial design concept was sound, and can be
1-12
developed into an automated support system for I&W functions. The timeframe of anIoperational development is obviously a function of the degree of automated versus
Interactive processes - - the more Interaction, the less Is required of the knowledge
base.
The concepts underlying the MATRES II design and implementation appear sufficiently
useful that the system has already aroused considerable interest both within and out-
side the intelligence community. With some additional developmental effort, it should be
possible to field an experimental MATRES system in the near future, allowing interested
users the possibility of hands-on evaluation of the I&W data base generation concept.
1-13
.11
2.0 METHODOLOGICAL APPROACH
2.1 Preliminary Notions
The method of approach which OSI has adopted since the inception of the RADC explora-
tory and developmental program for Automated Data Base Generation has been to look
ahead to the potential capabilities of a future system for both interactive and fully
automated exploitation of the narrative text of intelligence messages, and to develop a
methodology that will remain valid for applications of considerably greater scope than
the one currently under development.
This section offers a summary of OSI's methodological approach to the automated gen-
eration of indicator and descriptor data from the narrative text portions of intelligence
messages reporting on events related to movements of physical objects such as air-
craft, ships, missiles and satellites.
This methodology, initially developed on the basis of messages dealing with air activities,
was, under this contract, extended to cover reports of events involving missile and
satellite launchings and related events.
OSI's methodological approach is centered around the notion of "event", which is
adopted as the logical unit of analysis, and thus becomes the basis for describing intelli-
gence Information. Although the concept of an event Is fundamental to many research
endeavors, no standardized terminology for describing or classifying events seems to
exist. In many cases, the definition of an event Is arbitrary and tailored to a particular
field or purpose. Thus, In physics, the term 'event' usually refers to a point in the
space/time continuum, while in mathematical statistics It has the broadest meaning, that
of any proposition, whether true or not. The philosopher Russell (1956), regards the
event concept as a primitive (i.e., as undefined) and then uses it to define a series of
2-1
time points. In another usage, 'event' refers to a fact. In a related approach, the event
concept encompasses the parameters of 'action/time/location,' and is used to refer to
a type of activity. For a detailed treatment of the event concept as initially developed
by OSI, the reader is referred to Kuhns (1974), or Silva et al., (1979).
For the purpose of the work described in this report, we have expanded the definition of
an event as previously stated, to encompass all the states, processes, and activities
associated with an object or a set of objects from the inception to the termination of
Its/their mission. The term used for the aggregate of these states, actions, activities,
and processes is "global event", symbolized as EVENT.
To illustrate this usage of the term EVENT, consider the story of Skylab, as reported in
the news media (Los Angeles Times, Newsweek, and Aviation Week between 1973 and
1979).
Briefly, Skylab was the orbital workshop which was launched from the Kennedy Space
Center on the 14th of May, 1973. It incurred serious damage at lift-off time and was
later repaired by the astronauts previously scheduled to rendezvous with it. The
astronauts, launched in a modified Apollo service module, attached themselves to Skylab
by "umbilical" cords (connections to life support systems), and salvaged the ailing
spacecraft. Subsequently, Skylab was used for experiments, and was later abandoned.
Recently, it caused much concern, because it was obvious Skylab was going to deorbit
and crash to earth. It did indeed reenter the earth's atmosphere, upon which it began
to break up Into pieces, some of which burned up on reentry. Others impacted in the
Indian Ocean, and yet others landed in Western Australia.
From a global perspective, the Skylab story constitutes an EVENT and all the various
news items which Informed us of the state of Skylab over the last six years constitute
2-2
an EVENT REPORT.
A global event has a complex Internal structure composed of smaller interrelated units:
the launch, the deorbit, the Impact, etc. These smaller units are referred to as 'events',
or 'atomic events'. An 'atomic event' is roughly either a property that an object has at a
point in time or over a time interval, or a relation that holds among a set of objects or
locations at a point in time or over a time Interval.
Descriptions of atomic events take two forms: Intensional descriptions, and extensional
descriptions.
An Intensional description Is an abstract description of a class of Individuals in terms of
a set of invariant properties common to all members of the class. Thus, the intensional
description of the class of launch events would state that all such events are associ-
ated with an object that can be launched. In the satellite domain, the object is a satel-
lite. It usually has some specified mission, and is associated with an orbit, which itself
Is described in terms of several parameters such as apogee, inclination, perigee, and
period. Other entities associated with a satellite launch are the launch system used for
Injecting the satellite into orbit, the launch site, and the time and date of the launch.
An extensional description Involves one individual, I.e., a unique member of a class of
Individuals in the world being modeled. A simple example is the description of a specific
launch event involving a particular spacecraft (e.g., Skylab), launched from a particular
launch site (the Kennedy Space Center), by a particular launch system (a Saturn-5 type
launch vehicle), Into a particular orbit (the Skylab orbit had an Inclination of 50 degrees
to the equator), at a particular time and date (1300 hours on 14 May 1973).
The representational construct for intensional descriptions of events and their associ-
ated concepts is the "template" (see subsection 2.4.1), while that for the extensional
2-3
description of events and their associated concepts is the "event record" (see subsec-
tion 2.4.2). The relation between a template and its corresponding event record is
roughly the same as that which holds between an intensional description of a concept
and its extension. Thus, the set of event records describing events of the same class,
I.e., event records related to a particular template, constitute the extension of the con-
cept described by the template.
The collection of reports describing an EVENT constitute an EVENT REPORT. In the mili-
tary environment, an EVENT REPORT comprises a variable number of discrete intelligence
messages, which are often received hours, days, weeks, or even years apart.
The messages fall into definable classes which have important methodological implica-
tions for event processing. Each message type has an internal conceptual organization
reflected In the linguistic organization of the message text. The next few subsections
describe the major characteristics of EVENT REPORTS, and the internal conceptual and
linguistic organization of messege texL.
2.2 Characteristics of EVENT REPORTS
One of the fundamental properties of an EVENT REPORT is that it is coherent with
respect to a global theme. Thus, If the central theme of an EVENT REPORT is the mission
of a particular spacecraft, the atomic events described by individual messages, including
the comments offered and the inferences stated, are all In some intuitive way related to
that theme.
Precise rules for establishing coherence at this level, however, are very difficult to for-
4 mulate. Although it Is clear that determining coherence involves domain-specific
knowledge of objects, their properties and their behavior In the real world, as well as
knowledge of conventions governing reporting procedures, the inferential procedures for
2-4
K establishing coherence cannot be explicitly formulated at this time.
An attempt was made to identify some of the factors which might enter into the estab-
lishment of links between messages in the missile and satellite domain. Two notions
emerged as crucial In this area: the time reference of a reported event, and the
discourse referent, i.e., the entity discussed. (For further discussion, see subsections
2.3.4.3 and 2.5).
It Is clear that the procedures involved In event processing at the level of the EVENT
RECORD must be based upon a cognitive theory of discourse comprehension utilizing real
world knowledge of how the objects, facts, processes and events are organized in a
particular subject domain.
2.3 Characteristics of Messages
As mentioned above, the individual messages in message sequences constituting EVENT
REPORTS usually fall into several identifiable classes. From the point of view of
automated computer analysis, a distinction must be made between those messages that
contain new event descriptions (i.e., descriptions of events reported for the first time),
and those that either confirm previously reported events, request changes in the param-
eters of some previously reported event, add information to previously underspecified
parameters, or provide summaries of global events. From an operational point oi view, a
first report involves creating a new data element, while elaborations, requests for
change and updates involve changes and/or additions to an already existing structure.
Messages have a complex internal structure comprising header information, followed by
either formatted, semi-formatted, and/or unformatted (narrative) text portions, before
ending with some special symbols signalling the conclusion of the message.
2-6
Since this work is concerned mainly with the narrative text portions of messages, the
latter are described in terms of three components: a 'pre-text' component, the 'text'
component, and a 'post-text' component. The next two subsections focus on the
characteristics of the 'text' component of messages.
2.3.1 The Conceptual Organization of Message Text. This subsection focuses on the
conceptual organization of message text. In general. the 'text' component of a message
contains information as to the time and location of a given atomic event, and may contain
additional data giving the context of the event sequence or chain of related events,
properties of objects involved, the source of the information, and some interpretation of
the event.
The majority of event-related messages have a characteristic structure which may be
represented by the following formula, where the parentheses enclose optional elements,
and curly brackets enclose alternatives:
() E L (T) (())S represents references to the source of the information, and is an optional element.
The following single-sentence message contains a source reference represented by the
string THE SOVIET NEWS AGENCY TASS:
THE SOVIET NEWS AGENCY TASS ANNOUNCEDTHAT IMPACT OF COSMOS-954 TOOK PLACENEAR YELLOWKNIFE, CANADA.
E symbolizes the key event being reported in the message, while L represents the loca-
tion and T the time of the given event. In the above example, the key event is the
Impact of Cosmos-954, and the location Is Yellowknife, Canada. The time of the impact
Is not mentioned. When time Information Is omitted, It can be derived from the header of
2-6
Is not mentioned. When time Information is omitted, it can be derived from the header of
the message, or from a combination of other factors, Including the tense of the main
verb. The latter procedure often involves complex inferential processes which are dif-
ficult to implement.
E' symbolizes further information on the key event, e.g., properties of the objects
involved, and other occurrences in the chain of events reported.
I represents further Information on the event. The latter often takes the form of evalua-
tive comments on surrounding circumstances or consists of sentences describing
relevant historical background. Interpretative comments are often absent from mes-
sages, but in those cases, more detailed information on the key event itself is usually
given.
Messages containing more than one paragraph are generally structured along the same
lines, where each additional paragraph reports a related event, with associated time and
location data, details, and Interpretive comments.
In event-related messages, a description of the key event and its parameters occurs in
the first sentence of the text. Thus, It is always the first sentence of the message
text which Introduces the TOPIC of the message.
2.3.2 The Description of Message Content. In order to describe the information content
of message text,'we utilize the "Text Grammar" approach as developed in the writings
of Petofl (1971), Petofl and Rieser (1973), van Dijk (1972, 1979), van Dijk and Petofi
(1977), and Dressier (1978a, 1978b).
As the name Implies, the unit of analysis of a text grammar is a text, in our case, the
message text. The aim of a text grammar Is to provide an abstract linguistic description
of a text -- Including a description of the structure of its Individual sentences --
2-7
utilizing as many levels of analysis as necessary to make explicit the content structure
of the text.
Our approach to this problem is to describe the meaning content of the message in terms
of a "Message Grammar" in which the "primitives" are event classes, and the relations
are text-level relations. The latter may be optional or obligatory and determine the con-
nectivity or non-connectivity between events.
The definition of the Message Grammar Is based upon a systematic study of the
language used in a particular subject domain. The results of the study for the missile
and satellite sublanguage are pesented in Section 3 of this report.
The formulation of a message grammar for event data requires a multi-level approach,
beginning with the description of the text at the level of words and sentences and con-
cluding with a description of the information content of the message in terms of higher-
level abstract units. Each level of description involves its own units and principles of
analysis.
2.3.3 The Linguistic Organization of Message Text. At the linguistic level a message
text Is characterized in terms of a sequence of propositions. It is important to note,
however, that a message text is not just an arbitrary, unrelated set of propositions; it is
a coherent, structured conceptual unit, whose individual propositions are linked by means
of text-level relations including coreference, temporal relations, causal relations, entail-
ments and presuppositions.
The formulation of conditions for textual coherence Is an essential part of the linguistic
characterization of a message text and plays an important role In message text
analysis. Some of these relations are explicitly expressed In the surface structure of
the text; others are Inferred during the Interpretation process on the basis of contex-
2-8
tual and real world knowledge. In addition, it is Important to note that for a text to be
coherent, Its propositions need not necessarily be connected In an explicit manner at
the text level, since much of the Information conveyed by a text is Implied. It is suffi-
clent that a text be coherent at a higher level of abstraction with respect to the TOPIC
of the message, i.e., the entity discussed.
The next higher level of analysis is the Event Level. The linguistic level is related to the
Event level by a set of mapping rules which transform the individual propositions of the
text into event representations. The mapping rules are based upon a study of the
correspondences between form and content in a given domain. Examples of such
correspondences taken from the domains under consideration are given in the next sec-
tion.
2.3.4 Description at the Event Level. This level involves the description of atomic
events in terms of their properties, Including time, location, action, objects, and related
facts. For each subject domain, the description of message content at this level
requires that the set of characteristic event classes be Identified together with the set
of operations and relations on events that hold within the domain. This, in turn, requires
a thorough analysis of the vocabulary of the sublanguage, and lexical analysis, and a
study of text-level relations. The following sections sketch OSl's approach to these
requirements.
2.3.4.1 Vocabulary Analysis. A KWIC concordance of the message sentences is first
prepared to serve as a basic analytical tool for development of the dictionary and gram-
mar (see subsection 2.6.3). The vocabulary yielded by the Initial KWIC Index, Is parti-
tioned Into those Items which are entered Into the system's dictionary and those that
are members of open sets and must be treated heuristically.
A dictionary, or lexicon, Is a file of Information about the words of the language.
2-9
IV
The first category is made up of matural language words (nouns, pronouns, adjectives,
verbs, adverbs, conjunctions, subjunctions, etc.). The second category comprises
number strings, date time groups, regiment numbers, division numbers, and other
alphanumeric strings such as designations of equipment types and launch systems.
Furthermore, this category also includes geographic designations, names of countries,
areas, zones, cities, rivers, etc., vhich require the use of special dictionaries for their
recognition.
Heuristics for automatically recognizing and labelling lexical items which are members of
open sets (e.g., types of equipment, place names, geographical coordinates) involves
writing specific recognition routines involving anything from pattern-matching to syntac-
tic prediction.
In addition to a common core of English language words utilized across subject domains,
each Individual subject domain requires Its special dictionary reflecting domain-specific
nomenclature and usage.
Since the lexicon has to mesh both with the parser and the Interpretive routines, the
task of building the lexicon is a continuous effort extending over the entire period of
development.
2.3.4.2 Lexical Analysis. The results of the vocabulary analysis form the basis of the
lexical analysis, which in turn underlies the definition of new templates.
The goal of lexical analysis is the description of the concepts expressed by words and
the Identification of their conceptual relations to other words*. All key concepts, their
Each word entry is accompanied by information regarding its syntactic category, inwhat syntactic environments it can occur, and what some of its semantic propertiesare.Note that the term "word" Is used loosely to refer to single words or combinations ofwords generally considered to form a unit of meaning.
2-10
attributes, their Internal argument structure, and their external relations to other con-
cepts must be determined for each subject domain. The following paragraphs provide
some background on OSl's approach to lexical analysis.
Information on words is characteristically acquired by experience. It is generally
accepted that such information Includes spelling, pronunciation, inflected and derived
forms, major syntactic category, knowledge about how words combine with other words
to form grammatical phrases and sentences, knowledge about semantic relations
between words (e.g., synonymy, antonymy, hyponymy), knowledge about semantic fields,
rules for appropriate use in a given situation, and other encyclopedic facts. Knowledge
about words underlies a person's faculty to produce and comprehend language and to
communicate with other people in a manner appropriate to a given situation. The mean-
Ings of many words depend upon functional and perceptual attributes, as well as the
place that word occupies within a system of concepts.
One of the goals of lexical analysis is to define the boundaries of the lexical universe
for the domain under study. Within a given subject domain, representations of concepts
expressed by words need only Include what is relevant to the particular domain.
To Illustrate the kinds of issues Jnvolved in formulating hypotheses about concepts
expressed by words, their Internal relations and their links to other concepts, we offer
the following two examples.
2.3.4.2.1 Example 1: Hyponymy. flyponymy is generally considered to be one of the
most Important principles underlyng the organization of nominal concepts. It has been
the subject of many linguistic Investigations -- especially in connection with kin terms,
color terms, and plant and animal taxonomles.
According to Lyons (1968), hyponymy Is the relation which holds between a more
2-11
specific, or subordinate, lexical unit and a more general, or superordinate, lexical unit, as
exemplified by the pair of words 'Skylab' and 'spacecraft', where 'Skylab' Is a hyponym
of 'spacecraft'. Under this definition, Pioneer 11, Voyager, and Viking are co-hyponyms
of the superordinate term 'spacecraft'. Hyponymy has sometimes been defined in terms
of the logical relation of class-inclusion (see, for example, Carnap, 1956). According to
Lyons (1977), however, there are problems attaching to the definition of hyponymy in
terms of the logic of classes. He proposes that hyponymy be defined in terms of unila-
teral implication, as follows:
Word Wi is a hyponym of word Wj if, for any x, the sentence
"x is a Wi" entails the sentence "x Is a WJ".
The relation of hyponymy imposes a hierarchical structure upon a vocabulary and upon
particular fields within a vocabulary. Knowledge of hyponymy relations Is essential to
the resolution of discourse reference, and plays an Important role in information retrieval.
In Library Science, it underlies the assignment of "See also" references.
Hyponymy Is an external relation between concepts. Example 2 below illustrates a
class of internal relations which are often referred to as "case" relations (Fillmore,
1968).
2.3.4.2.2 Example 2. Case Relations. Natural language cases are widely recognized as
an Important organizing principle in. the analysis and description of natural language
data. Many of the advanced projects for natural Janguage understanding In the U.S. as
well as abroad embody some sort of case system.
A "case" is a binary relation which holds between the predicate (usually, but not neces-
sarily, realized as a verb) and one of its arguments. A case analysis determines the
semantic roles of the components of an expression with respect to a central concept.
2-12
.1
Consider, for example, the sentence:
A SATURN-5 TYPE LAUNCHER WAS USED TO PLACE THE
SPACECRAFT INTO ORBIT.
The main predicative concept "USE" denotes an action, and as such determines the
roles of the other components of the sentence with respect to it. Thus, "A Saturn-5
type launcher" fulfills the role of "instrument", and "to place the spacecarft into orbit"
the role of "purpose" in relation to "USE". These roles are purely semantic and
describe the Internal relations of the concept USE to other constructs in its immediate.
syntactic environment. Continuing our analysis, we note that the "purpose" argument
can be further decomposed into the predicative concept "place", the noun phrase "the
spacecraft", and the prepositional phrase "into orbit". The noun phrase "the space-
craft" fulfills the role of "object", while the prepositional phrase "into orbit" fulfills the
role of "locational goal" in relation to "place".
Correlations between the text of our example and its logical argument structure are
expressed by what linguists refer to as the "selectional restrictions" or the "selectional
preferences" of a particular verb. Some of the selectional preferences for the verb
"use" are given below:
(a) The Agent, if expressed in surface structure, corresponds to
the "logical" subject.
(b) The Instrument is expressed by the object noun phrase denoting
any object which can be "used".
1 (c) The Purpose, if expressed in surface structure, is a "to"
complement, or a "for" complement.
Internal relations may be obligatory or optional. For example, with the predicate USE, the
2-13
purpose for which something is used need not always be explicitly expressed in surface
structure. The verbal gerund "Using Message Spooling Processors in a Non-Interactive
tNetwork" does not spell out the purpose. It does, however, give an indication of the
environment ("in a non-interactive network"). Both Purpose and Environment are optional
arguments of USE. Mention of the Instrument argument, is, however, mandatory.
In conclusion, this brief analysis of the two examples given above identifies only a small
number of the semantic relations which can hold among words. An in-depth lexical study
must come to grips with issues relating to polysemy and lexical disambiguation, including
the problem of core senses and the way a core sense can be extended to provide other
senses; with morphological relations, including inflectional and derivational word forma-
tion, as well as with the meaning of compounds, with the notion of "semantic field", and
with Issues of representation. in addition, the role of presupposition must be clarified.
Presuppositions and entailments are of particular importance for event data analysis,
because they make predictions about possible sequences of events. If any of these
predictions are violated, they must be brought to the attention of the analyst.
2.3.4.3 The Study of Text Level Relations. By "text-level relations" we mean those rela-
tions that connect atomic propositions in a running text. They include such relations as
synonymy, hyponymy, part/whole relations, causal relations, temporal relations, pronomi-
na! reference, noun phrase reference, and temporal reference.
Factors that enter into text level relations are implications in the semantic content of
constituent propositions, lexical equivalence, and syntactic devices such as time and
place relators, logical connectors, and the use of proforms.
For example, consider the following message fragment, which consists of two clauses,
each describing an atomic event:
2-14
-1
THE SATELLITE DEORBITED AFTER COMPLETING A 28 DAY MISSION.
The event described by the main clause (the satellite deorbited), stands in the time
relation of succession to the event described by the subordinate clause: the deorbit of
the satellite is understood to have taken place after a mission which lasted for 28 days.
This time relation is explicitly stated in the text, and can be derived from the semantics
of the two clauses, including the meaning of the subordinate conjunction 'after'. The
semantic interpretation of the time relation between the two propositions is defined as
follows:
BEFORE(P,P')
where P represents the 28 day mission, and P' the deorbit of the satellite.
In other cases, links between propositions of a text can be established by means of the
hyponymy relation holding between words contained in tie text. To illustrate this, con-
sider the following pair of consecutive sentences:
(a) SPUTNIK 1 WAS OBSERVED IN ORBITOVER THE INDIAN OCEAN ON FEBRUARY 11.
(b) THE SATELLITE LATER REENTERED THE EARTH'S ATMOSPHEREAND BROKE UP UPON REENTRY.
Here the noun phrase "the satellite" is a more general term for "Sputnik 1", the relation
between the two terms being one of hyponymy (subsection 2.3.4.2). The link between
the two sentences Is one of noun phrase reference.
Equivalence between two noun phrases Is sometimes explicitly stated in a text as in (c)
below:
(c) THE FIRST EARTH SATELLITE, ALSO KNOWN AS SPUTNIK 1...
2-15
In the previous example, the information which permitts the resolution of the noun phrase
reference is present in the immediate context. At times lexical connection between
sentences may not be given overtly, but may depend upon factual knowledge that the
originator of the text assumes on the part of the reader. Consider the following pair of
sentences where the definite noun phrase refers to a concept not mentioned in the pre-
vious context, but is conceptually linked to it.
(d) A FLIGHT OF TWO AIRCRAFT WAS ACTIVE OVERPLACENAME DURING THE LATE ZULUTIME HOURS.
(e) THE PILOTS WERE IN COMMUNICATION WITH ANUNIDENTIFIED CONTROL AND REPORTING CENTER.
The problem here is to establish the referents of the definite noun phrase "the pilots",
which has not been mentioned in the previous context. In cases such as these, the
disambiguation may be handled by what Chafe (1972) describes as "foregrounding".
Chafe argues that whenever a new concept is introduced into a communication, that
concept introduces a number of related concepts into the local context or foreground.
For example, once the concept of "aircraft" has been mentioned in a text, we can use
definite reference for its engine, fuselage, wings, and even its pilot. Thus, in the pair of
sentences above, sentence (d) introduces the concept of "aircraft" which then
automatically brings into focus closely associated concepts including the no' on of
"pilot", which is referred to by a definite noun phrase in (e).
In an automated system, the fact that aircraft have pilots is stored in the permanent
knowledge base of the system and is thus readily accessible to the routines for ana-
phora resolution.
2-16
LWill
2.4 Representational Issues
This section discusses two representational constructs of fundamental importance to
event processing: the Template and the Event Record.
2.4.1 The Template. Taking the event as the primary unit of analysis, OSI has developed
the concept of a "template" as an organizing principle for the uniform representation of
information on events and event-related entities, as viewed from the perspective of the
user/analyst In the context of a particular task domain.
Events and event-related entities are described as n-ary relations, where the n-ary
relationships is named by a predicate symbol and the arguments of the relation
correspond to the "roles" of case theory (see subsection 2.3.4.2).
An important part of the information encoded in templates consists of what linguists call
the "selectional restrictions" or "selectional preferences" in a particular domain. In
essence, the latter express the correspondences between syntax and semantics, and
play a fundamental role in semantic interpretation. They form the basis for constructing
the procedures which map syntactically analyzed input sentences into event records
and thus have the function of reducing the many ways a concept can be expressed in
natural language to a systematic representation of the information content of the input
sentence.
Although most correspondences between syntax and semantics hold across subject
domains, each domain seems to present some idiosyncratic usage. It is important, there-
fore, that the correspondences between syntax and semantics be studied in depth for
each subject domain before they can be translated into algorithms and incorporated into
templates.
2-17
In the domains under investigation there are templates for classes of objects (aircraft,
missiles), classes of events (flights, launchings), classes of relations (temporal, causal),
rand other concepts such as the date time group. For example, a template describing the
class of flight events in the context of the air activities domain includes parameters
such as time of flight, the aircraft involved in the flight, the purpose of the flight (its
mission) , the point of departure, the destination, the current location, the farthest point
reached, direction, altitude, and path. The aircraft involved in a fiight are in turn
described in terms of aircraft-related parameters such as equipment type, nationality,
organizational subordination, and so on. These are all parameters which can enter into a
flight event and are therefore part of the cognitive model of the analyst, i.e., of his view
of what a flight involves.
Table 2-a provides an informal description of the LAUNCH template as developed for the
Missile and Satellite domains. The template embodies a set of descriptors selected from
a small set of descriptor types. Each descriptor has a procedure attached to it which
incorporates the information necessary to relate abstract descriptions of concepts to
syntactic structures. They are essentially mapping rules which effect the transforma-
tion of parsed sentences into event records,
In a sense a template is likc a class declaration, a structure naming the attributes
(descriptor slots) which are optionally or obligatorily associated with instances of that
class, and specifying the values those attributes can have.
Templates ilntcgate procedural knowledge with a richly structured declarative rGpresen-
tation. In this sense, templates have a lot in common with both Wilks'(1977) recursive
semantic formulas and Bobrow and Winograd's (1977) KRL Specialization Units.
2-18
Table 2-a. Informal Description of the LAUNCH ConceptIn the Missile and Satellite Domain.
+------------------------------------- -------------
IDescriptive Elements 11 Procedural Elements I
------------------------------- ---- -----------------
I ~IOBL/I I ProceduresIII /Ill forI
IDescriptor I Filler Specification 1/OPTI lfilling slotsI+------------------------------ ----4-+-------------+-
lAgent I If expressed, then 1 11 If conditions hold, II logical subject of IOPT 11 fill Agent slot with II sentence. Head noun 1 11I subject noun phrase II with feature NATION 1 1I 1
+------------------------------ ------------------ 4.
I If no Agent, then 11I According to whichI Object in logical 11I conditions hold,I subject position; 11I construct ObjectI
lObject I otherwise in object iOBL 11 template from either IIposition. Allowable I il subject nounphrase or IIfeatures: MISSILE 11I object nounphraseII and SATELLITE I I
-------------------------------- ---- ----------------
I Either subject I IlTest headnoun of subject II nounphrase with I I for feature BOOSTER II headnoun with 1 11I feature BOOSTER, or 1 11
ILaunchsys I PP with prep BY and JOPT IlSearch VMODS list II I headnoun with feature I I for specifiedIII BOOSTER I I Iprepositional phrase I
--------------------------------- ---- --------------- 4-
ILaunchsite I PP with headnoun I I ISearch 'JMODS list III (+LOC) I Ilfor specifiedI
I II II constituentI+------------------------------ ----,-+-------------+-
III Il5earch VMODS list IlInclinationlPP with headnoun I Ilfor specifiedI
II (+INCL) I I Iconstituent
+------------------------------- ----------------- +-
2-19
Table 2-a (contd)+------------------------------------------------+-
S I I II II IPP with Preps TO or I IlSearch VMODS list I
IDestinationhINTO and headnoun I iIfor specifiedI iwith feature LOC I 1 Iconstituent
II I II------------------------------------ ----------------
I i 1. ADV (TYME REF) or I IlSearch VMODS list IITime I 2. PP with TYME prep IOPT 11for specified II I and headnoun with I l1constituent II I feature TYME I II 1+------------------------------------ ------------ +-
I I I IlSearch VMODS list II I PP with DATE-node IOPT l Ifor specified IiDate I I I Iconstituent II I I II I
+----- + --------- --- +..+ ------------ +
Each subject domain yields its own template inventory corresponding to the events and
objects and their internal and external relations -- optional or obligatory -- which have
Informational significance within that domain. In the Missile and Satellite domain, some of
the key concepts for which templates have been constructed are: "launch", "deorbit",
"reentry", "breakup", "Impact", "missile", "satellite", and "DTG". (For a complete listing
of the template inventory and the auxiliary procedures as encoded in the language Pro-
log, see Appendix C).
In summary, the template is an information structure which provides the means for coding
the analyst's cognitive models in terms of logical data structures which are susceptible
to automatic processing. In other words, templates provide a framework for the
representation of higher-level conceptual information approximating that which a human
reader has of a given subject matter.
Logically, templates can be viewed as relational network models of memory In which
primitives are relations, words, and word senses. Procedurally, they can best be
described as the fundamental knowledge structures which mediate the correlations
between syntactic structures and their corresponding information content.
2-20
, ,... ..,.. __.._ . .. . - , ,. - , - ,, . ,,1: . -
Each template describes a class of entities in terms of those properties which are nor-
mally associated with that class in a particular task domain. A template thus reflects
the information user's conceptualization of the domain, i.e., his view of what that class of
entities involves.
2.4.2 The Event Record. While templates are abstract data structures for the represen-
tation of event classes, Event Records are concrete data structures for the representa-
tion of individual events. An event record is the description of a single individual, i.e., a
unique member of a class of individuals in the world being modeled. A simple example is
the description of a specific spacecraft (e.g., Skylab) which was launched from
specific launch site, at a given time and date. Table 2-b illustrates the event record
corresponding to the LAUNCH sentence below:
THE SKYLAB ORBITAL WORKSHOP, A CONVERTED S-4BTHIRD STAGE FROM A SATURN-5 LAUNCH VEHICLE,WAS LAUNCHED FROM THE KENNEDY SPACE CENTERAT 1300 HOURS ON 14 MAY 1973.
Table 2-b. Event Record for LAUNCH Sentence+-----------------------------------------------------------------------------------t
Event: LAUNCHAction: LAUNCHEDObject: SATELLITE... Equipment= SKYLAB ORBITAL WORKSHOP... Number=... Relative= A CONVERTED S-4B THIRD STAGE FROM A
SATURN-5 LAUNCH VEHICLE
LaunchslIte= FROM THE KENNEDY SPACE CENTERTime = AT 1300 HOURSDate = ON 14 MAY 1973
+-----------------------------------------------------------------------------------
Notice that the above sentence does not provide values for all the descriptors associ-
ated with the LAUNCH template. Templates represent an aggregate of 611 the possible
parameters and attributes which can be associated with an event or object in the con-
2-21
J L
text of a given domain. Many of these parameters or attributes are optional and there-
fore need not be present in any one actual description In a text,
Note also that the appositive "a converted S-4B third stage from a Saturn-5 launch
vehicle" is analyzed as a relative clause and stored in the "Relative" descriptor slot.
For a discussion of tle current treatment of relative clauses the reader is referred to
the subsections on relative clauses and appositive postmodification in Section 3 of this
report.
Event Records have several important properties which render them particularly useful
as a support tool for the I&W analyst:
e They reflect the analysts view of the world, and are thus compatible with their
cognitive models of objects, events and states of affairs in their area of exper-
tise.
* They are discrete representations of events, objects, and their properties and
are usable for the construction of a data base.
9 They are so designed as to allow flexible retrieval of information not only by
event type, but also by other associated parameters, such as object(s) involved
in the event, and time and location indicators.
e The information stored in these data structures is in a format which lends itself
readily to further processing. This processing may be related to storage and
retrieval functions, may be statistical in nature, or may be part of the inference
making mechanisms to be developed for a future system (e.g., a system for event
prediction).
The next section deals primarily with discourse reference. The section begins by briefly
defining the problem of reference, and then goes on to discuss several forms of
2-22
reference common In our sublanguage.
2.6 The Problem of Reference
A more complete computer 'understanding' of narrative text requires the machine to
have the ability to deal with anaphoric language in a perspicuous and systematic way.
This ability is especially important for the analysis of larger texts characteristic of intel-
ligence messages.
Two of the major surveys discussing the problems of anaphora are Nash-Webber (1977)
and Hirst (1979).
Computer-based attempts to handle anaphoric expressions are described in Baranovsky
(1970), Burton (1976), Charniak (1972, 1973), Deutsch (1975), Hobbs (1976),
Rumelhart (1975), Wilks (1975), Winograd (1972), Woods (1972), and others.
Information contained in a message falls Into two distinct categories. At each point in a
message text some of the information is "new": i.e., it is being introduced into the text
by the sentence being analyzed at that point. Other information is "given", or "old", i.e.,
It has previously been Introduced In the text and we assume, stored in memory, much as
the human processes sequential text. Interpreting a text requires identifying given con-
cepts in memory and attaching the new Information to them. In the following discussion,
the term "refenrjng expression" Is used to denote those parts of a sentence that com-
municate given information.
The problem of reference, then, Is the problem of Identifying the concepts referred to In
a text. Such concepts need not explicitly be expressed as segments In a text; they
very often aremntltles which are assumed to be in the reader's mind.*
* For a discussion of the general p)roblem of reference see RADC-TR-77-194, Vol. I,Part I, Section 2.1.
2-23
h6.1_
Anaphoric expressions comprise pronouns, pro-verbs, some definite noun phrases and
-1 ellipses. The ensuing remarks are restricted to definite referring expressions.
2.6.1 Definite Referring Expressions. For the purposes of discussion it is useful to dis-
tinguish two kinds of definite referring expressions: pronouns and non-pronominal defin-
ite noun phrases. The reason for this distinction is that the processes needed to iden-
tify the concepts referred to by pronouns differ from those needed for the resolution of
non-pronominal definite noun phrases.
The test data used under this contract contains both pronominal and non-pronominal
definite noun phrases (e.g., it, their, they, the spacecraft, the three missiles launched
from PLACENAME).
2.5.1.1 Pronominal Reference. Pronouns carry little information in themselves. Consider
the following sentence:
IT IMPACTED IN THE AUSTRALIAN OUTBACK.
The pronoun "it" in the above sentence only tells us that the subject of "impacted"
must be singular. To help identify the referent of "it" we must determine its functional
role in the sentence -- Its "case" in the sense of Fillmore (see subsection 2.3.4.2 --
and make use of the syntactic and semantic restrictions on the slot occupied by the
pronoun. Then the previous context can be searched for a concept that satisfies those
restrictions. In the given example the restrictions are first, that the referent of "it"
must be singular, and second, that it must be a physical object capable of impacting. In
the given subject domain, such an object is either a missile or a satellite.
The following paragraphs suggest a method for handling pronoun references of the kind
Illustrated above.
2-24
The above sentence would first be syntactically analyzed into a propositional structure,
and subsequently undergo case assignment; the pronoun 'it' is assigned the role of
Object. As mentioned above, 'It' carries the lexical information that it is definite and
singular. The fact that it was assigned the Object role indicates that its referent must
belong to either the missile or the satellite class. This information is included in the res-
trictions specified in the intensional description of the IMPACT event concept. Once the
program determines the restrictions on the referent for "it", it proceeds to examine all
descriptions recently stored in memory to find one which meets the semantic feature
tests for the pronoun's case slot. If such a description is found, the program will ensure
that it agrees in number with Vit" before it is accepted as a possible referent. If the
test is positive, the program creates a link from the Object slot of the event record
representing the sentence under discussion to the description just identified.
When pronouns refer back over longer portions of a text the resolution process is more
complicated, since the larger the previous context, the more numerous are the potential
referents of the pronoun and the more complex the inferential processing that has to be
performed.
The next section examines definite noun phrases which are used anaphorically, i.e., the
noun at the head of the construction refers to a specific object or concept.
2.5.1.2 Definite Noun Phrases. The major difference between pronouns and definite
noun phrases Is that the latter carry more information. Thus, the phrase 'the four SS-1 1
missiles' contains information specifying both the general class (missiles) and the type
(SS-1 1) of the objects referred to.
The resolution of such noun phrases is basically a problem of finding a matrhing descrip-
tion in memory. The methods required to decide whether a given entity fits 9 vven
2-26
description are often very complex and Include all kinds of Inferential processes.
,. In the simplest cases are those which Involve matching a definite description with an
object that has been described in the same way previously in the text. Other cases
may Involve complex routines using the semantics of descriptors as well as their syntac-
tic form, the Immediate linguistic context, the world knowledge stored in the system,
external data files (special dictionaries, glossaries, etc.), and sometimes even the
extensional event data base in order to decide whether two descriptions match.
Sometimes a definite noun phrase may refer to an entire chain of events. In such cases
it is necessary to appeal to structures made up of units defined at a higher conceptual
level than that customary in traditional linguistic analyses.
For a computer program to do this, It must be able to refer not to the individual sen-
tences of the message, but to the chain of events represented by these sentences
expressed in higher level conceptual categories appropriate for event description, thus
making use of the "Message Grammar".
One of the areas into which this research might profitably be extended is exploring the
possibility of resolving anaphora not only within the limited context of a single message,
but also within the larger context of an EVENT REPORT.
2.6 Guidelines for the Establishment of a Research Corpus
Because of the nature of the data for which this system is developed, a special
approach Is taken to the establishment of a research corpus.
2.6.1 Data Collection. For any given subject domain, the first task relates to the collec-
tion, "sanitizatlon", and organization of a representative sample of intelligence messages
to serve as the source data for the analysis and characterization of the reporting
language.
2-26
Data should be gathered In cohesive units which form a conceptual whole in the given
domain. Thus, in the missile and satellite domain, messages were collected in sets con-
stituting EVENT REPORTS. Only if so organized can a systematic study of the domain be
undertaken.
A sufficient amount of data needs to be collected for each subject domain. The notion
"sufficient data" can only be quantified by a step by step collection of data in each
subject domain. It is a well known fact that each new batch of data collected for any
given domain yields less and less new Information. At a certain point the graph of addi-
tional new items vs. total information levels off. it is at that point that further data col-
lection becomes unproductive.
2.6.2 Data Sanitization. The source data gathered for purposes of development is usu-
ally classified. It can therefore not be used in its original form for in-house development
of computer programs i, the analysis of message text. Such developmental work can
be accomplished far more easily with unclassified test data than with actual messages,
provided the test data has the same grammatical structure as the original material.
To achieve this, a procedure to generate unclassified message text from classified data
was developed. The procedure consists of two steps.
e Step 1. Narrative message text is transcribed according to a set of transcription
rules, according to which all proper names of objects, their attributes, and
time/location parameters are replaced by placeholders. For example, names of
political entitles (e.g., USA, UK) are replaced by the string "politname". Specific
j geographic placenames (e.g., Cape Kennedy, Florida) are replaced by the string
"placename". fames of persons (e.g., Pete Conrad, Valery Bykovsky) are
replaced by the string "personname, while times and dates are replaced by
2-27
appropriate placeholders. A complete list of the sanitization rules developed for
the Missile and Satellite domains is given in subsection 3.2.1. The resulting text
is used for the study of the grammatical structure of the reporting language and
for system development.
e Step 2. For purposes of demonstrating the system, the transcribed text is input
to a special routine which places all events in the future and randomly replaces
the placeholders by either totally ficticious names or by names referring to
objects in outer space.
This method preserves the syntactic structure of the original language for linguistic
analysis. At the same time, it provides some appearance of verisimilitude.
Thus, the transcribed sentence (a) below, might be changed into sentence (b):
a. MISCLASS LAUNCHED FROM PLACENAME1 TO PLACENAME2, DAYNO NMTH 4NMBR.
b. UM-67 ICBM LAUNCHED FROM BETELGEUSE TO RIGEL 23 JULY 1987.*
2.6.3 The KWIC Concordance. Next, the sanitized data is put into a machine readable
form and processed by a standard KWIC index program, which yields a key-word-in-
context concordance, orthographic type counts, .and some limited statistics. The utility
of KWIC concordances and associated frequency lists both as an aid to linguistic
analysis and for determining priorities in the parser need hardly be stressed. An excerpt
from a KWIC concordance produced for the Missile and Satellite domains is shown in
Table 2-1.
A listing of each word form in its context permits a mo!-e precise charactorization of the
sublanguage under investigation than would be possible otherwise. By its very form, the
Note: M-67 Is the name of a galactic cluster in the Cancer constellation. UM-67 isa fictitious name. Betelgeuse and Rigel are stars in the Orion constellation.
2-28
KWIC index facilitates the determination of the word classes and relations characteristic
of a particular sublanguage, and thus forms the basis for defining both the scope and the
vocabulary specific to a reporting language and the scope of the sublanguage grammar
to be used by the parser.
The three procedural steps described above lay the foundations for the conceptual
analysis of the source data.
2
2-29
7~~N 0i vq.IW NdIC W NpNC NrhrN.4f t N.Nr.NnlN N Nf4N C
w 7 w C -0 N7 -- -0 hra 40 WLI. rr
-70 0 L 62 - 4. a C - ~ x2 Lzoorcra...o-~ ! -a o rmht 7 OnI 4.-W sI. 4
C7~4C3 er.L CLI Ow 3 0 0 a - IZLIL.
3 7'..Zw 7C h ht . . w.- O c ,WE IL
-.a~ IIL r a.- .77h-Z r LIZ ha.a04-NLI~ha1iZX. .4._. I W IO ,. r-. s -0. 7 O
2 ~ ~ ~ - ,7 C 74 00 !7Zc7L 4. E.w~hhZL44 ~ ~ x 0LWL4.Q -. 7 ~r -x ~ z w t .- 4.4iJLm4C 4 C.h OO
-w3s ,- OO 74 70' . 7 LI =aa Z~r ~ Iw00haaUN~ ~ ~ ~ ~ ~ ~~~~~W CI...2 3 4 )hh ~ ... .j7L
CZ 20ha77M cq~a G2 _47-C- -rtuo a s cmTo ~ ~ ~ ~ 'a aM x0C w7- Xa~h x -C.r..ta Xz zrc 7.
E_ CI.J U .:.aC 2-2 -0 rrcL habs.j 40'J --. Y h w
CL 0 . .i . 0.-'2 -C C4 0 ( 3 . Z .wa~ a -3 7 . 4 U 4 C h (o ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 3 ha(L 4 0a SWL 0 0= ' 7 ... Oa07 ~ ~ = coca.( wa 777 .4 .4 0 0 rr . r r
U.. - Z t07 r a2 LI0 haahormw.4..7 7L 0 . 4..C &7
!Uw 7 L w2
ha . . m g7 O . O O C O 0 0 0 0 0wW-) .. 407t..4
4-. m~m2 ...hha ww . 140(7 0 00 00 0 00 0
C ~ ~ ~ ~ = zx z z z z 44 07 0..7aLLUIII 'a 74.U da. Uaaahhhhh"
a.~~~~ -C 40 - I07 4 44 4 4 -0- --- IC .2 OhhaS.77L Q..4.4323a 0Sz...C4C zcCC4C44LU74
LI).' 4 7 00 000LILLIL)ILI ULLIL!!!77
- 7.- . Cw C : j l a g, 7 7 7 0 -w h a . t0 0 7 7 0 0 w 0 0 . .3 0
rr C .-a7 C3r4.r XS - -..a, t - . ~ s
000.., 0 0 c 4 C.h Nw _-0 74 M.- 47 7 0 0000 0-ILIIU I 00~Lh L L7 4 OhSC 7 .cch
000 0 N m. . - 074 074 -LI 0O..wCOO 00 w 5
~~~C ta,4. 5" .: 0 --I4 70 . 7 7 4 . h
- ~ W jw _ .47 0 7 04-4_W4 7w ha I 0al 4 .N 7 ( 0 , ha mhh 47 4; 1Uh h 0 4 .Z 0a
- zh.. - - - - 4 7 a 0 0 ,~ - W haE
ha t4 - 0 C 2h 4 z maw4W4
ha x _x OfC Z 77
4.7~I 7m 0z -w wS7 0 . 7 7 4 .u~wZM - 304 1
-xw47c~c 4-7 C-.jsa7JW J4
imu-a 412-30
3.0 MISSILE AND SATELLITE DOMAIN DEFINITION
This section begins with a brief introduction of the missile and satellite worlds, and con-
tinues with a detailed description of the two domains in the form of a domain definition.
8.1 The Missile and Satellite Worlds
The missile and satellite worlds have a large number of properties in common.
Both worlds contain political entities (e.g., U.S., USSR, Uganda), which carry out inten-
tional actions (e.g., they launch satellites, or missiles).
Both worlds are characterized by certain human actions whose instigators normally
remain unidentified. Thus, objects are identified; missions are assessed or confirmed;
events are expected, announced or confirmed; actions are attempted.
In both worlds time and location indications are of crucial importance.
There are, however, some significant differences, due to the differences in function
between missiles and satellites. Satellites but not missiles are launched into a given
orbit; satellites but not missiles perform orbital maneuvers; satellites but not missiles
have well-defined missions; some satellites perform manned flights, in which case they
may rendezvous and dock with other spacecraft; finally, satellites but not missiles are
deorbited, abandoned, or reactivated.
Another Important difference concerns the time span associated with the series of
actions and events connected with a launch. Thus, while the duration of the chain of
events associated with missile launchings Is measured in minutes, that of events associ-
ated with satellite launchings Is usually measured In days, months, or even years, as in
the case of Skylab.
The characteristic objects In the satellite world are spacecraft. As mentioned above,
3-1
the range of actions a spacecraft normally performs is wider than that of missiles.
Spacecraft can be launched, placed into orbit, deorbited, and recovered; they can per-
form orbital maneuvers, rendezvous and dock with other spacecraft, they can deorbit
themselves, reenter the earth's atmosphere, break up, burn up, crash to earth and
Impact. They can be manned or unmanned. They can have different kinds of missions.
Their physical properties are varied.
Missiles are simpler to describe and the number of atomic events they can participate
In, although similar, are much more restricted.
3.2 Domain Definition
3.2.1 Research Corpus. The test corpus used as a basis for the developmental work
described in this report consists of 280 messages organized into EVENT REPORTS.
3.2.1.1 Source Data Sanitization. Approximately 20% of the messages collected were
transcribed according to the sanitization rules discussed in subsection 2.5.2. For a list-
ing of the sanitized message texts, see Appendix A.
The list of placeholders used for the transcription of the missile and satellite test corpus
Is given below.
Placeholder List
Note that all propernames used in this list are taken from the sources listed in the Fore-
word to this report.
" SPANAME (spacecraft name): COSMOS-724, COSMOS-651, VOY.GER, SOYUZ-21, MOLNIYA-2A PIONEER-11, etc.
" SPAIYPE (spacecraft type): SOYUZ, COSMOS, PIONEER
" SPACLASS (spacecraft class): ESV
* SPASTANAME (space station name): SALYUT-4
" MISNAME (missile name): SS-N-10, SS-14, SS-N-6,
3-2
.1
& MISCLASS (missile class): INTERMEDIATE BALLISTIC MISSILE, ICBM, SLBM,DRONE, SAM
@ LSTYPE (launch system type): SS-4, SATURN-1B
* PRESSNAME: TASS
* PLACENAME: TYURATAM, PLESETSK, THE AUSTRALIAN OUTBACK
* SEANAME: INDIAN OCEAN, BLACK SEA
e TRAC (test range acronym):
a POLITNAME: SOVIET UNION, USSR, USA
e POLiTADJ: SOVIET
e NATNAME: (Used for nationals of countries) SOVIETS, AUSTRALIANS
e PERSONNAME: CHARLES (PETE) CONRAD, YURIJ ROMANENKO, GEORGIJ GRECHKO
w COORDINATES: 68-28N 46-29E, 49-33N 160-24E, 70N 36E, (Note the varietyof formats used for the expression of coordinates).
* ZULUTIME: 0940Z, 281600Z, (Note the different formats).
a 1NMBR: 1, 2, etc.
* 2NMBR: 10, 11, etc. (and similarly for 3NMBR, etc. Note that 1 NMBR or 2NMBRmight designate a day of the month, and that 2NMBR or 4NMBR might refer to ayear.)
a NMBRNAME: SEVEN, SIX, etc.
• 1NMBRTH: 1ST, 2ND, etc. (and similarly for 2NMBRTH, etc.)
e NMBRNAMETH: FIFTH, SIXTH, etc.
* RNMBR- used for Roman Numerals like 1, 11, V, VII.
0 NMTH: used for name of month, e.g., JANUARY, NOVEMBER
The list of "placeholders" remains open-ended. New placeholders are incorpo-ated as
required by the material uoder study.
3.2.1.2 Examples of Transcribed Text. The following examples show sanitized vcrsions
of real message text.
Example 1
4 MISNAME LAUNCHED FROM PLACENAME 2NMBR MNTH 4NMBR.AT ZULUTIME1, A MISNAME WAS LAUNCHED FROM PLACENAME1.IT IMPACTED NEAR PLACENAME2 AT ZULUTIME2.
3-3
Example 2
PRESSNAME ANNOUNCES LAUNCH OF SPANAME.AT ZULUTIME, PRESSNAME-THE POLITADJ NEWS AGENCY-ANNOUNCED THE LAUNCH OF
rSPANAME, THE UNIDENTIFIED MILITARY SUPPORT SPACLASS WHICHWAS LAUNCHED FROM THE PLACENAME MISSILE AND SPACE CENTER AT ZULUTIME TODAY.ORBITAL PARAMETERS AS CONTAINED IN THE ANNOUNCEMENT ARE:
APOGEE NMBR KILOMETERSPERIGEE NMBR KILOMETERSINCLINATION NMBR DEGREESPERIOD NMBR MINUTES
The transcription procedure is very time consuming. The exercise of developing a saniti-
zation procedure, however, has proved useful. It has provided a means of partitioning
the message vocabulary into those items which are common English words and those that
are specific to the domain (nomenclature, specific ways of referring to times and dates,
geographic locations, etc.) and therefore require special recognition procedures in an
operational environment.
The set of messages sanitized according to the above procedure under this contract are
contained in Appendix A.
3.2.2 The Structure of EVENT REPORTS. As one wolid expect, the structure of the
EVENT REPORTS in the two domains reflects the state of bifairs in the real world. EVENT
REPORTS describing sate!lite-related activities consist of several message3, each
describing one or more atomic events.
For example, in the satellite domain, a first message always describes a launch, and is
Invariably followed by a message confirming the launch. A sub3equent mc~sage reports
the announcement of the launch by a foreign press agency and usually provideS details
of orbital parameters. The foreign press announcement is optionally followed by one or
j imore messages describing related events such as orbital maneuvers, staMi!ization
maneuvers, a rendezvous, a docking, a deorbit, an impact and/or a recovery.
3-4
7 -:7L
The last message in an EVENT REPORT is usually a summary statement desciibing tha
global event.
EVENT REPORTS describing missile launching and related events are much more compact.
Often the launch, the reentry and the impact of a particular missile are described in a
single message.
This structural difference has implications for the process models which will apply to the
two domains; the model for the satellite domains will be much more complex than that
for the missile domain.
3.2.2.1 Message Types. In order to characterize EVENT REPORTS at the global level,
their internal composition in terms of message types was studied.
Individual messages were classified according to the type of information they contain.
As pointed out in subsection 2.3, it is important to distinguish between those messages
that refer to an event for the first time, and those that confirm events, elaborate on
their properties, request changes, report on related events, offer comments, or provide
global summaries of EVENTS.
Several distinct message types were identified:
a. Messages describing a new event.
b. Messages confirming an event.
c. Messages elaborating on the parameters of a previously reported event.
d. A report of a report, e.g., a message reporting on a foreign news agency pressrelease concerning an event.
e. Data Summary Messages.
f. Closure Messages
g. Follow up Messages.
h. Requests for changes to previously reported facts.
The message types studied In detail under this contract Include those listed under
3-5
a-d. Types e-h are characterized by an abundance of evaluative statements,
hypothetical statements, and statements expressing opinions or certain inferences.
The current study was limited to statements describing events aiid therefore
excludes message types e-h.
3.2.2.2 Event Typc5 and other Concepts. The fullowing major Event types were ideniti-
fied:
a. LAUNCH
b. DEORBIT
c. DOCKING
d. REENTRY (AND DECAY)
e. BREAKUP
f. IMPACT
A number of other concepts with a complex internal structure were identified:
a. Objects: Missiles and Satellites
b. The date time group (DTG)
c. Orbital Parameters: Apogee, Perigee, Period and Inclination.
The relations which hold between events in the missile and satellite domains, are mainly
temporal relations of succession in time and relations of presupposition and entailment.
A spacecraft which is in orbit now must have necessarily been launched some time pre-
viously. A spacecraft which is reentering the earth's atmosphere will break up, burn up,
or crash to earth.
A complete listing of the templates developed for the event types and other concepts
listed above, including their ancilliary procedures as encoded in Prolog is given In
Appendix C.
3-6
The DEORBIT template, as encoded in Prolog, is explicated In subsection 4.3.3.
3.2.2.3 Descriptor System. The following descriptors were Identified for the Missile and
Satellite Domains:
3-7
Table 3-1. Missile and Satellite Descriptor System---------------------------------------------------------------------------------------------------- IIA. Higher-Level descriptorsI Infosource The source of a report of an event
e.g., a foreign news agency
I Status Status of event (expected to happen;I failed to happen; confirmed)'B. Event-related descriptorsI Agent Animate instigator of an action.
I Object The entity that moves or changes orwhose position or existence is beingdescribed.
I Location The location of the object at somepoint in time.
I Destination Projected or actual destinationof the object at the end of themission.
I Mission Mission of satellite
I Revolution Usually revolution on which1C. Orbit-related descriptorsI ApogeeI PerigeeI PeriodI Inclination
IC. Launch-related descriptorsI Launchsite Site from which a missile or
satellite was launched
Launchsystem System used for firing missileor satellite.
ID. Date/Time Group DescriptorsI Time Time of observed eventI Date Date of observed eventI Duration Duration of an eventlE. Object-related descriptors
Equipment Class Spacecraft or missile class
I Setspecification Number of ob ects
Further description Usually a relative clauseI or appositive clause
---------------------------------------------------------------------------
3-8
3.2.3 Language Definition. The reporting language used in the two domains has special
grammatical and lexical properties which justify calling it a sublanguage.
3.2.3.1 Relationship between General English and the Reporting Language. The report-
ing language consists of declarative sentences. There are no questions or commands.
Furthermore, the reporting language is charqcterized by grammatical constructions which
deviate from those of "normal" English, (e.g., dropped articles, dropped prepositions).
Although it might be possible to first recover the deleted material and then subject the
expanded text to a general parsing grammar, it turned out more convenient to write a
specialized grammar stating the allowable combinations of word classes directly. Our
number of rules still is considerable smaller than those that would be required for general
English.
As an example of idiosyncratic vocabulary usage, consider the verb "deploy". In ordi-
nary discursive prose, this verb is normally used with 'an animate Agent: somebody
deploys something. In our corpus, inanimate objects like aircraft more often than not
deploy themselves. Constructions with the Agent expressed in surface structure occur
less frequently.
3.2.3.2 The Grammar. In the MATRES II System, the linguistic structure is defined by
means of an augmented transition network grammar in terms of familiar linguistic
categories such as sentence, nounphrase, verbgroup, prepositional phrase and adverb.
In order to expedite processing, a number of language specific categories, not usually
found In traditional grammars, were added. Thus, the familiar definition of prepositional
phrase in (a) was augmented to encompass dates (b):
(a) pp - prep + nounphrase
Wb) pp prep + date
3-9
where 'date' is a non-terminal of the grammar with its own internal structure.
In this section we give an informal description of the major grammatical phenomena which
are covered by the grammar, and of the analyses which are given them by the ATN
parser.
3.2.3.2.1 The Declarative Sentence. The only class of sentences handled by the current
version of the grammar are declarative sentences. As mentioned above there are no
other sentence types in the corpus.
A declarative sentence may be a simple sentence as in (1), a complex sentence with
embedded nominal clauses as in (2) and (3), or sentences with adverbial subordinate
clauses as in (4) and (5). The current version of the grammar does not handle coordina-
tion.
1. SKYLAB DEORBITED OVER CANADA.
2. TASS ANNOUNCED THAT SKYLAB DEORBITED OVER CANADA.
3. SKYLAB FAILED TO IMPACT IN CANADA.
4. THE SATELLITE DEORBITED AFTER A 13 DAY MISSION.
6. THE SATELLITE DEORBITED FOLLOWING A 13 DAY MISSION.
The MATRES II grammar analyzes a declarative sentence as a list having as its first ele-
ment a simple sentence, which may be followed optionally by a sentence conjunction,
and either another simple sentence or a noun phrase.
3.2.3.2.2 The Simple Sentence. A simple sentence may have six components, of which
only the main predicate Is mandatory. The components are: voice (active or passive),
subject (a noun phrase), a verb group, optionally followed by a direct object, a comple-
ment, and one or more post-verb modifiers.
3-10
The grammar analyzes a simple sentence as a six-branched node structure. The first
branch points to the voice node, the second to the subject node, the third branch to the
verb group, the fourth to the object, the fifth to a complement, and the sixth to a list of
adverbial modifiers.
3.2.3.2.3 The Noun Phrase. In the simplest case, a noun phrase may consist of a pro-
noun or a proper noun optionally followed by an appositive construction. On the other
extreme, a noun phrase may consist of a determiner followed by a list of pre-head
modifiers, a head noun, and a list of post-head modifiers.
A determiner may consist simply of an article (e. g., 'THE'), a quantifier (eg. 'ALL'), or a
number phrase (eg. 'AS MANY AS SIX'), or it may be a complex structure involving two or
three of these constituents, as shown in the examples below:
ALL THE MISSILES
ALL SIX MISSILES
THE SIX MISSILES
ALL OF THE SIX MISSILES
Pre-head modifiers may include adjectives, nouns, past participles, and present pariici-
pies. In both domains analyzed, head nouns are typically preceded by several modifiers
referring to various attributes. Example (8) is taken from the aircraft domain, while
example (9), is from the satellite domain.
(8) RETURNING UGANDAN UBBC SR-71 AIRCRAFT
(9) A FIRST GENERATION HIGH RESOLUTION PHOTOGRAPHIC SATELLITE
Possible post-head modifiers are relative clauses, reduced relative clauses, sppos tives,
3-11
-Aft
and prepositional phrases. An example of each is given In (10) through (13), respec-
tively.
(10) SKYLAB, WHICH WAS LAUNCHED FROM THE KENNEDY
SPACE CENTER ON 14 MAY 1973,....
(11) THE SATELLITE, LAUNCHED FROM THE KENNEDY
SPACE CENTER ON 14 MAY 1973 ....
(12) THE SKYLAB ORBITAL WORKSHOP, A CONVERTED S-4B
THIRD STAGE FROM A SATURN-5 LAUNCH VEHICLE, DEORBITED .......
(13) THE AIRCRAFT FROM ENTEBBE
A noun phrase Is analyzed as a four-branched node. The first branch points to a deter-
miner (possibly null, as in (10)), the second to a list of pre-head modifiers, the third to
the head noun, and the fourth to a list of post-head modifiers.
The current version of the grammar only allows simple noun phrases (i.e., those without
post-head modifiers) to occur as direct objects or prepositional objects.
3.2.3.2.4 Nominalizations. The following constructions, referred to as nominalizations,
are very frequent in our research corpus:
a. THE DEORBIT OF SKYLAB
b. THE IMPACT OF SKYLAB
c. THE REENTRY OF SKYLAB
These nominalizations are parsed as noun phrases and are later converted into proposi-
tional structures in the ERL "machine" before semantic interpretation. Nouns like DEOR-
BIT, IMPACT and REENTRY, which denote events, become the main predicates of their
3-12
respective propositions, and the objects of the preposition 'of' become the logical sub-
jects. Such nouns are marked by the feature EVENTIVE in the lexicon.
3.2.3.2.5 Relative Clauses. Among the post nominal modifiers, the relative clause is one
of the most frequent constructions. A relative clause may start with a relative pronoun
(e.g., which, that), in which case we refer ;o it as a "wh" relative (example a), or it may
start with a past participle (b) or a present participle (c) optionally preceded by an
adverb, in which case we refer to It as a reduced relative.
(a) THE UNIDENTIFIED MISSILE, WHICH WAS SOFTLANDEDNEAR THE SEANAME TODAY, WAS FIRED FROM PLACENAME.
(b) SATNAME, THE FIRST GENERATION HIGH RESOLUTIONPHOTOGRAPHIC SATELLITE LAUNCHED FROM PLACENAMEON 2NMBR MONTHMAME, WAS DEORBITED DURINGITS 3NMBRTH REVOLUTION.
(c) THE SPACECRAFT, CARRYING TWO ASTRONAUTS,WAS SUCCESSFULLY INJECTED INTO AN ORBITINCLINED 2NMBR TO THE EQUATOR.
Relative clauses usually describe an event related to the main event of the sentence.
As such they provide links to previous sentences or even messages.
In the current version of the system, relative clauses are parsed as sentences and
stored In the postmodifier list of the head noun. The Interpretive routines recognize
relative clauses and store them as a unit in the "Relative" slot of event templates. A
more sophisticated system, however, would break down relative clauses and build event
records for them, which would be suitably connected to the main template. This was not
done under this contract simply for lack of time.
The current version of the grammar also allows for relative clauses with a sentential
antecedent as In the sentence below:
3-13
THE AIRCRAFT WAS APPARENTLY POSITIONED AT
0125N3470E, WHICH WOULD PLACE IT APPROXIMATELY
6 KMS INSIDE UGANDAN AIRSPACE.
Such relatives are stored in the post modifier list of the main verb. The following exrm-
pie illustrates how relative clauses are stored in templates:
(a) THE UNIDENTIFIED MISSILE , WHICH WAS SOFTLANDED NEARTHE SEANAME TODAY, WAS FIRED FROM PLACENAME.
Event: LAUNCHAction= FIREDObject: MISSILE...Equipment= UNIDENTIFIED MISSILE...Number=...Relative= SOFTLANDED NEAR THE SEANAME TODAYLaunchsite= FROM PLACENAME
3.2.3.2.6 Noun Phrase Apposition. Noun phrase apposition is very common in the sub-
languages under study.
For two or more noun phrases to be appositives, i.e., in apposition, they must iiormally be
identical in reference or else the reference of one must be included in the reference of
the other. For example, in (a), THE SOVIET NEWS AGENCY and TASS refer to the same
organization.
(a) THE SOVIET NEWS AGENCY, TASS, ANNOUNCED THAT ........
The semantic relationship between the two noun phrases is one of "appellation", a sub-
class of the more general "equivalence" relationship (c.f., Quirk, 1972). With "appella-
tion", both noun phrases are definite and the second is typically a proper noun.
The converse of "appellation" is "designation" also a subclass of the "equivalence"
relationship. With Jesignation -- as in the case of appellation -- both appositives are
commonly definite ,un phrases, but here the second appositive is less specific than the
3-14
first, as illustrated in (b)
(b) TASS, THE SOVIET NEWS AGENCY, ANNOUNCED THAT ..........
A third subclass of the equivalence relationship is that of "identification". With identifi-
cation, the first appositive is typically an Indefinite noun phrase and the second apposi-
tive is more specific as in (c) below:
(c) AN UNIDENTIFIED MISSILE, PERHAPS A DRONE .......
Here, there is no longer unique equivalence as there was with (a) and (b); the second
appositive identifies -- often only tentatively -- wvhat is given in the first Zppositive.
A second type of major semantic relationship in strict non-r-estrictive noun-phrase appo-
sition is that of "attribution". "Attribution" involves predication rather than equivalence.
The second appositive is commonly an indefinite noun phrase (although it can also be
definite), and can be replaced by a relative clause:
(d) SPANAME, A LOW RESOLUTION PHOTOGRAPHIC SATELLITE,
LAUNCHED FROM TYURATAM AT OOOZ ON 30 DEC 1955,..............
Figure 3-2 summarizes the semantic relationships in the types of appositives id.rtified
so far in the three sJbject domains analyzed thus far.
*---------------------------------------------------------------------+l i APPELLATION
I EQUIVALENCE IDENTIFICATION
DESIGNAT ION
I ATTRIBUTION
--------------------------------------------------------------------- 4
Figure 7-2. Semantic Relationships in non-restrictive
noun-phrase apposition.
Because all the subclasses of noun phrase apposition found in the aircraft, missile, and
satellite materials to date are such that the second appositive can be replaced by a
relative clause, it was decided to analyze them as relative clauses in the AIN grammar,
3-15
and to shift the burden of distinguishing appositives from relative clauses to the inter-
pretive component (the ERL "machine") -- if such distinctions were useful in an an
operational environment.
The following example, taken from the the Missile domain, shows how a sentence con-
taining an appositive is analyzed by the current version of the MATRES II system.
* > AN UNIDENTIFIED MISSILE, PROBABLY A MISNAME,*WAS FIRED FROM PLACENAME AT APPROXIMATELY ZULUTIME TODAY.Event: LAUNCHAction= FIREDObject: MISSILE...Equipment= UNIDENTIFIED MISSILE...Number=...Relative= PROBABLY A MISNAMELaurfchsite = FROM PLACENAMETime= AT APPROXIMATELY ZULUTIME TODAY
3.2.3.2.7 The Verb Group. The verb group may consist of an auxiliary tcliuwed by a
verb, as in (14), or an auxiliary followed by a copilia followed hy an adjective, as ini (15).
(14) HAVE BEEN CONDUCTING
(15) HAVE BEEN ACTIVE
In (14) the auxiliary is 'HAVE BEEN', while in (15) the auxiliary is 'HAVE', and 'BEEN' is the
copula.
Some verbs (eg. 'CONDUCT', 'PENETRATE') must be followed by a direct object consti-
tuenit, which is another noun phrase. Other verb3 (eg. 'ARRIVE') never have a direc
object, while for others (eg. 'OPERATE') the object is optional. Suc.h verbs are mark-d
TRANS, INTRANS or left unmarked in the lexicon.
3.2.3.2.8 Adverbials. Adverbial modifiers include prepositional phrases and adverbs, and
may occur before the subject, as in (16), after the verb (and the object, if there is one)
as in (17), or embedded within the verb group, as is the case with 'CURRENTLY' in (18).
3-16
(1 6) AT APPROXIMATELY 9 AM ON 25 MAY 1973,
AN APOLLO SPACECRAFT CARRYING THREE ASTRONAUTS
WAS LAUNCHED FROM THE KENNEDY SPACE CENTER.
(1 7) THE SATELLITE WAS LAUNCHED AT APPROXIMATELY
1330 HOURS ON 14 MAY 1973.
(18) THE SATELLITE IS CURRENTLY ORBITING OVER CANADA.
3.2.3.2.9 The Concepts of Time and Space. Two of the key concepts in intelligence
reporting normally expressed as adverbials are the concepts of "time" and "space".
Knowledge of temporal patterns of given classes of events and time order relations
between event classes, coupled with location data, can often assist the analyst in
creating an overview of the current situation and in predicting possible outcomes.
Knowledge of all events that occurred at about the sam.e time or within a given time-
span, in a given geographic area, can greatly enhance the analyst's capability to make
accurate assessments of world situations.
Time icferences are particularly important, since in many cases they provide the only
explicit link between the various messages constituting an EVENT REPORT (i.e., all mes-
sages referring to the same global event).
For example, consider an event referring to the launch of a new satellite. A first mes-
sage may report the observed launching of an "unidentified" satellite. Although there is
no name available for the satellite, such a message usually gives precise information as
to the time, date and site of the launching. A subsequent message may state that the
satellite launched at such and such a time from such and such a launchsite is now con-
firmed in orbit, but still lack information on the class of the satellite launched. The mes-
3-17
sage identifying the satellite by name is sometimes not received until Vha foreign news
agency announces the launch, or when the satellite is otherwise identified. Notice that
up to that point the only available link between the received messages is the !aunch
time and date and possibly the site.
The linguistic realization of time and location adverbials in our particular mes.age do,.in
often differs rather sharply from that in "normal" English.
Time phrases in the sublanguages under consideration tend to occur either at the begin-
ning of sentences or at the end. In addition, they have special characteristics which
render them easy to isolate from the rest of the sentence. First, they usually begin with
certain prepositions (e.g., at, between, during, in, on, since, from ... to, until). Second,
they necessarily contain one or more of a set of words designating the months of the
year (including abbreviations); numbers, either spelled, in numeric form, or a form peculiar
to intelligence reporting (zulutime); and third, expressions like "today", "this date", this
year". The prepositions and the time expressions in the third group form closed sets
and can be easily identified on the basis of a fairly restricted sample for any given class
of messages.
The following are some examples of time and date expressions taken from the intelli-
gence messages reporting on aircraft activities and missile and satellite launchings:
* AT 0940Z; AT 0940; AT 09400oz; AT APPROXIMATELY 0940Z; AT ABOUT 0940Z
e AT THE SAME TIME
e BETWEEN 0115 AND 0332Z; BETWEEN 0115-0332Z; BETWEEN 0115:0332Z
* BY 201600Z
e CURRENTLY
* DURING THE EARLY 0200Z HOUR
e DURING THE MORNING HOURS
* EARLIER; EARLIER TODAY
3-18
i
* FROM 1400-1625Z
* SINCE 0119; SINCE ABOUT 01 19Z; SIINCE APPROXIMATELY 0119
Corresponding date expressions are:
* DURING THE YEAR
* IN 1977; IN OCTOBER 1977; OCTOBER 1977
a ON 12 DECEMBER; ON 12 DECEMBER 1977; ON 12/13 DECEMBER
* ON THE 3RD; ON 3RD APR; ON THE 3RD APR
* 12 DECEMBER; 12 DEC 74; 12 DECEMBER 1974; DECEMBER 1977
* THIS DATE; ON THIS DATE; ON THAT DATE
* THIS YEAR* TODAY; EARLIER TODAY
Date phrases are analyzed as three branched nodes. The first branch points to the day,
the second to the month, and the third to the year (the third is often null).
3.2.3.2.10 Location References. Natural language expressions describing locations are
far more varied than expression referring to time. While many such expressions may
only be identifiable by means of linguistic analysis, there are nevertheless some classes
that could be handled by a finite state character processing algorithm. Obvious candi-
dates are references to location by means of coordinates (e.g., 5026N7138E, 50-26N
071 -38E, and variations thereof).
3.2.3.2 11 Complementation. The current version of the grammar handles two types of
complements: "that"-complements and "to"-complements, the only types found in the
reporting languages under investigation.
Examples of 'that'-complements from the missile and satellite corpus are:
a. ...SUGGESTS THAT A MALFUNCTION MAY HAVE OCCURRED AS EARLYAS 2NMBR MNTH.
b. ...SUGGESTS THAT NMBRNAME WERE CARRIED.
c. ...ANNOUNCED THAT SPANAME WAS LAUNCHED FROM THE POLITNAME.
d. ...SUGGESTS THAT THE MISSLE WAS A MISNAME.
e. ...INDICATED THAT T-E SECOND STAGE ENGINE FAILED TO IGNITEAS PROGRAMMED FOLLOWING FIRST STAGE SHUTDOWN.
3-19
f ... INDICATIONS ARE THAT THE SECOND STAGE OF THE LAUNCHVEHICLE FAILED TO IGNITE.
g ...INDICATES THAT THE SPACECRAFT HAS BEEN ABANDONED.
h. ...INDICATED THAT THE SPACECRAFT, IF SUCCESSFUL, WOULD HAVEBEEN INSERTED iNTO AN ORBIT SIMILAR TO PREVIOUS NAVSATS.
I. ...INDICATES THAT THIS WAS A NOMINAL MISNAME MOD 1NMBR,...
As the above examples show, 'that'-complements are complete declarative sentences
preceded by the word 'that'. In our corpus, they are found in object position of verbs,
as well as after certain nouns. The parser analyzes 'that'-complements as sentences
and stores them in the 'Compl' register.
'To'-complements consist of the 'to' marker followed by the first verb of the predicate,
which must be untensed. Examples of 'to'-complements from the missile and satellite
corpus are:
a. ...FAILED '0 ACHIEVE EARTH ORBIT AND REENTERED.
b. ...WAS TO BE INCLINED 2NMBR DEGREES TO THE EQUATOR.
c. ...WAS THE NMBRNAMETH MISNAME TO BE LAUNCHED FROM PLACENAMETHIS YEAR.
d. ...APPEAR TO BE PRESENTLY CONDUCTING A SYSTEMATIC CHECKOUTOF THE ON-BOARD SYSTEMS.
e ...WAS ALLOWED TO CONTINUE ITS ORIGINAL WESTWARD DRIFT IN
SILENCE.
f. ...ATTEMPTED TO GEOPOSITION THE SATELLITE ON 1NMBR MNTH.
g. ...FAILED TO IGNITE.
h. ...FAILED TO IGNITE AS PROGRAMMED FOLLOWING FIRST STAGESHUTDOWN.
I. ...IS EXPECTED TO RENDEZVOUS AND DOCK WITH THE ORBITING SPACESTATION SPASTANAME.
J. ...WAS APPARENTLY INTENDED TO REPLACE SPANAME.
k. ...SPANAME1 WAS LAUNCHED TO REPLACE SPANAME2.
I. ...WILL PROBABLY ATTEMPT TO SEPARATE THE NUCLEAR POWER SUPPLYFROM THE MAIN PAYLOAD.
m. ...ARE CONTINUING TO SHIFT WESTWARD AT ABOUT NMBR DEGREES PER DAY.
n. ...AFTER FAILING TO STABILIZE IT.
3-20
3.2.3.2.72 Passive Sentences. A form of "be" followed by a past participle form of a
verb Indicates the passive construction. Passive constructions are "inverted" so that
passive and active forms of the same sentence result in the same "deep" structure. In
the MATRES II grammar, a sentence such as (24) is restructured as In (25), where the
surface subject becomes the object, and the surface object becomes the subject.
(24) THE SATELLITE WAS LAUNCHED BY NASA ON 14 MAY 1973.
(25) NASA LAUNCHED THE SATELLITE ON 14 MAY 1973.
3.2.3.2.73 Ambiguity. Because of the limited semantic domain, the possibilities for
word-sense ambiguity are greatly reduced.
3.2.3.2.14 The Paraphrase Problem. Since we are dealing with a restricted subject
domain, we are not confronted with the whole spectrum of difficulties one would expect
In a system for general English. Nevertheless, we do have to.deal with a certain amount
of paraphrase.
Sentences with eventive nouns as logical subject and with verbs such as TAKE PLACE,
OCCUR, or forms of the verb BE as main verb, are considered paraphrases of sentences
with main verbs which correspond to the eventive noun. Two examples are shown in 26
and 27 below. Note the similarity of the corresponding event recQrds.
Example 1: Impacted vs. Impact (occurred/took place/was...
26. (a) SKYLAB IMPACTED IN WESTERN AUSTRALIA JUST SOUTHEAST
OF KALGOORLIE ON 12 JULY 1979.
(b) IMPACT OF SKYLAB OCCURRED/ TOOK PLACE/ WAS INWESTERN AUSTRALIA JUST SOUTHEAST OF KALGOORLIEON 12 JULY 1979.
3-21
K. Table 3-3 (a) Event Record for Sentence 26(a)
+----------------------------------------------------------------------- ------
IEvent: IMPACTIAction= IMPACTEDI Object: SATELLITEI... Equipment= SKYLAB
... Relative=I Location= IN WESTERN AUSTRALIA JUST SOUTHEAST OFI KALGOORLIEI Date= ON 12 JUL 1979
------------------------------------------------------------------------------
Table 3-3 (b). Event Record for Sentence 26(b).+----------------------------------------------------------------------------
IEvent: IMPACTIAction: IMPACTIObject: SATELLITEI.. . Equipment= SKYLAB
... Rel~ative=I Location= IN WESTERN AUSTRALIA JUST SOUTHEASTI OF KALGOORLIEI Date= ON 12 JUL 1979
.------------------------------------------------------------------------------
Example 2: Reentered vs. Reentry
27. (a) SKYLAB REENTERED THE EARTH'S ATMOSPHEREOVER CANADA ON 21 JUL 1979.
(b) REENTRY OF SKYLAB TOOK PLACE OVER CANADAON 21 JUL 1979.
Table 3-4 (a) Event Record for Sentence 27(a).+------------------------------------------------------------------------------
IEvent: REENTRYI Action= REENTEREDI Object: SATELLITEI... Equipment= SKYLAB
... Relative=I Location= THE EARTH'S ATMOSPHERE OVER CANADAI DATA= ON 12 JUL 1979+------------------------------------------------------------------------------
3-22
Table 3-4(b) Event Record for Sentence 27(b).
---------------------------------------------
I Event: REENTRYI Action= REENTRYI Object: SATELLITE... Equipment= SKYLAB
I ... Relative=I Location= OVER CANADAI Date= ON 12 JUL 1979
+-----------------------------------------------------------------------------4
To achieve this, the sentences with the eventive noun as subject are first restructured
in the ERL component. Thus, in (27b), the eventive noun IMPACT is recogi-,ized as
expressing the main predicative concept and its parse tree is restructured to resemble
that of (a): the surface verb "occurred", "took place" or "was" is replaced by the lexi-
cal entry for IMPACT.
3.2.3.2.15 Functional Synonyms. In each domain there are a number of verbs which Gre
used interchangeably( e.g., "launch" and " fire"; "land" and "impact"). These are
treated as functional synonyms, i.e., they are marked In the lexicon as members of the
same class of event related concepts. Similarly, all terms identifying members of a mis-
sIle class are marked In the lexicon by the superordinate term, and processed by the
template, in this case. the MISSILE template.
Marking functional synonyms in the lexicon as members of the same class of concepts,
allows retrieval in terms of event or object classes, rather than requiring the user
analyst to think of all possible words for the members of the event or object classes.
The two examples below illustrate event records derived from two sentences tsing tte
functional synonyms "launch" and "fire". Note that both sentences are interp,-etcd as
"launch" events. The particular verb used Is shown In the Action slot.
3-23
Table 3-4 (a). Launch vs. Fire
+----------------------------------------------------------------------------
I *>> THE SKYLAB ORBITAL WORKSHOP,I *A CONVERTED S-4B THIRD STAGE FROM A SATURN-SI *LAUNCH VEHICLE, WAS LAUNCHED FROM THE KENNEDYI *SPACE CENTER AT 1330 HOURS ON 14 MAY 1973.I Event: LAUNCHI Action= LAUNCHEDI Object: SATELLITE... Equipment= SKYLAB ORBITAL WORKSHOP
I ... Relative= A CONVERTED S-4B THIRD STAGE FROM ASATURN-S LAUNCH VEHICLE
I Launchsite= FROM THE KENNEDY SPACE CENTERI Location=I Time= AT 1330
-----------------------------------------------------------------------------
Table 3-4 (b). Launch vs. Fire
-----------------------------------------------------------------------------
*>> THE SKYLAB ORBITAL WORKSHOP, A CONVERTEDI *S-4B THIRD STAGE FROM A SATURN-5 LAUNCH VEHICLE,*WAS FIRED FROM THE KENNEDY SPACE CENTER AT
1 1330 HOURS ON 14 MAY 1973.t Event: LAUNCHI Action= FIREDI Object: SATELLITEI ...Equipment= A CONVERTED S-4B THIRD STAGE FROM AI SATURN-S LAUNCH VEHICLEI Launchsite= FROM KENNEDY SPACE CENTERI Location=I Time= AT 1330-----------------------------------------------------------------------------
We are aware that paraphrase rules may lead to enormous difficulties in a system for
general English. However, in our restricted task domain, the problem seems mnnageable.
We also are aware of the fact that there will always be the possibility of a paraphrase
that was not anticipated, as well as other cases when the complexity of the experimen-
tal input exceeds the current capabilities of the model of the system. This is one of the
3-24
reasons why MATRES I is interactive.
3.2.3.3 The Lexicon. The MATRES II lexicon is specifically designed to support the
grammatical analysis procedure, and is therefore intended as an economical rather than
an exhaustive inventory of feature descriptions. It consists of two parts: a listing of
the features or attributes employed by the qystem, and a collection of lexi al ontries in
the form of static declarations of lexical items and their attributes.
The attributes fall into several classes. Examples of each are given betow.
(i) Major Grammatical Category Specifications.
ADVB (adverb)ADJ (adjective)ART (article)CONJ (conjunction)NUM (number)N (noun)VB (verb)
(ii) Examples of Lexical Features:COPULADIR (directional)EVENTIVE (marks eventive nouns)INTRANS (marks intransitive verbs)LOC (locational)MODALPASTP (past particile)PRESP (present participle)ROBJ ("raise-object")RSUBJ ("raise-subject")SUBNUM (subordination number)TENSED (marks tensed verbs)TRANS (marks transitive verbs)1NMBR, 2NMBR, etc. (marks one-digit, two-digit, etc., numbers
3-26
(lII) Event-Related Features:------------------------- +-----------------+------------------------------------
IEVENTS I OBJECTS I OTHER CONCEPTS------------------------- +-----------------+------------------------------------
I ACTIVITY FLIGHT I ACRAFT IINCL (inclination) II ARRIVE IMPACT I MISSILE I REV (revolution) II DEORBIT LAUNCH I SATELLITE I COMM (communication)lI DEPART PNTAE OSE
I DEPLOY REENTRY I II ENROUTE RETURN III------------------------- +-----------------+------------------------------------
An excerpt from the current lexicon is given below:
DELTA-CLASS C ADJ I .
::DEORBIT [ N EVENTIVE DEORBIT I C VB DEORBIT I.DEORBITED EVB TRANS PASTP DEORBITI
VB TRANS TENSED DEORBIT21.DEPARTED E VB TRANS PASTP DEPART
[ VB TRANS TENSED DEPART I.DEPLOYED [ VB TRANS PASTP DEPLOY 2
[VB TRANS TENSED DEPLOY 1.DEPLOYMENTS [ N EVENTlVE I .
DESTINATION C N I1DIVISION [N 21 .;DJIBOUTI [N LOC I .
DOWNED [ VB TRANS PASTP I[VB TRANS TENSED I .
DOWNRANGE E N LOC I .;DURING [ PREP EMOD TYME
The entire lexicon, comprising lexical entries for all three domains studied thus far, Is
contained In Appendix B.
3-26
4.0 IMPLEMENTATION
4.1 Principles of Discourse Processing
From the theoretical viewpoint, the primary goal of discourse processing is to arrive at
the total information content of a text, where the total information content of a text is
taken to be the aggregate of all the information communicated by that text, including
that which is made explicit and that which can be inferred from the meaning of the
words appearing In that text and their syntactic and semantic Interrelations.
A human analyst, however, Is selective. He does not seek to extract all the information
a text may contain, but only that which is needed for the performance of his task.
When reviewing a message, the human analyst uses his innate knowledge of English
grammar, as well as his extra-linguistic knowledge of entities such as spacecraft, time,
location, and actions -- Including all the relevant concepts which can be attributed to or
are Implied by such entities -- and extracts only those information items which are
relevant and useful to the attainment of his current goal.
To distill information elements from a text, the computer must in some sense model the
cognitive processes of the analyst. It must take into account what is known about
human linguistic behavior -- how the analyst interprets text, how he makes inferences,
how he remembers, and how he communicates.
Fundamental to the approaches taken in this field is the assumption that e, psrson
Interpreting natural language text uses much more than just the information ccntained in
Isolated words and sentences. The meanings which people attribute to words and sen-
tences vary not only according to their linguistic context, but also with the subject
matter being discussed and/or the situation in which they are used. A word or phr ise ik
Interpreted in terms of the total context In which it occurs; it is the kno.,ictc_ CC tho
4-1
total context which enables a person to understand language, and this knowledge
Includes all our knowledge about the real world.
Many of the subprocesses Involved in language understanding are still largely unex-
plored and it is therefore not possible to construct a comprehensive model of language
understanding. For example, no specification for the many and complex inferential
processes involved in language understanding can be given at this stage, although the
points in the understanding process at which they operate can be stated.
The OSI natural language processing system is based upon a process model of text
understanding involving four sets of operations.
First, the sentences of a message text are parsed into a set of propositinal structures.
The propositions are linked by various semantic relations which may be explicitly
expressed in the surface structure of the text, or Inferred during the interpretation pro-
cess on the basis of contextual and/or real world knowledge.
Second, the resulting set of propositions are organized into higher-level conceptual
categories, namely, event representations.
A third set of operations links the resulting event representations into a coherent whole,
reflecting the meaning of the message text as . whole.
Finally, when all m.s -ges constituting an EVENT REPORT are processed in this manner, a
set of constraints checks the coherence of the set of messages constItt~tino on EVENT
REPORT at the global level, i.e., at the level of the EVENT REPORT.
4.2 The MATRES II Text Processing System
4.?. 1 General Remarks. The principles of message text processing disc -3, ed above are
partially implemented in the MATRES 11 text processing system.
4-2
MATRES II is the result of the second cycle in the development of a system with full
capabilities for deriving formatted records from the narrative text of intelligence mes-
sages. It represents a considerable advance on MATRES I, which provided only a rudi-
mentary capability for testing algorithms for narrative text analysis.
The two subject domains of MATRES II are air activities and missile and satellite launch-
Ings. While in a fully developed system the unit of analysis would be the entire mes-
sage, the scope of the current system is limited to the analysis of single declarative
sentences.
The definition of the input language accepted by the system is embodied in a transition
network grammar model based upon Woods (1970, 1973). The MATRES II parser has
undergone considerable refinement and expansion and currently accepts a much wider
range of syntactic constructions than was previously achieved.
In the current version of the system, English language words are entered into a linguistic
dictionary, while strings with fixed patterns are recognized at the input stage by a finite
state automaton (FSA) designed especially for this purpose.
The FSA recognizes strings denoting Zulutimes (e.g., 1907Z), geographic coordinates
(e.g., 3674N4261E), integers (e.g.16, 1978), and ordinals (2nd, 3rd, 25th). Such
strings are tagged with appropriate features at the input stage and added to the inter-
nal lexicon.
The major feature of MATRES II is its capability for semantic analysis. This is achieved
by means of the Event Representation Language, which Is a language specially
developed for mapping the syntactic structures produced by the parser Into template-
derived content representations. The basic data structure of the Event Representation
Language Is the template.
4-3
The current version of the system takes single sentences as input and transforms each
Into one or more event records by combining a "bottom-up", data-driven ana!ysis based
upon linguistic and logical principles with a "top-down", conceptually driven domain-
specific interpretation of the structures generated by the input analysis. The "bottom-
up" analysis is effected by the augmented transition network (ATN) parser, which util-
izes a dictionary and a grammar of the reporting language to produce a parse tree show-
ing the syntactic composition of the input string and the hierarchical relationships
between component structures. The output of the parser is input to the ERL "machine",
which uses a set of prestored templates for the interpretation of the input, and pro-
duces event records as output. These event records constitute the primary elements for
the construction of the "extensional" data base, whose purpose is to serve as a support
tool for higher-level analytical functions in a decision-making environment.
Figure 4-1 illustrates how the program reorganizes the components of -- in this case --
a hypothetical sentence to give a clearer presentation of its information content.
The computer program which embodies this approach to natural language understanding
Is written in FORTH, Prolog, and SNOBOL4, and runs on a PDP 11 /45 under the RSX 11 D
operating system. A flow diagram of the MATRES II system is shown in Figure 4-2.
4-4
Input: Unstructured Text----------------------------------------------------------------
I The three unidentified heavy bombers which flewI from London to Cairo on the 30 Apr 1975, refuelled II in Naples at approximately 1300 hours the same day. I+-----------------------------------------------------------------------------------
Output: Event Record+-----------------------------------------------------------------------------------
I Event Type: REFUELI Verb used: REFUELLED
Object:S.. Equipment: UNIDENTIFIED HEAVY BOMBERS
I ...Set specification: THREE...Relative: WHICH FLEW FROM LONDON TO CAIRO
ON THE 30 APR 1975
Location: IN NAPLESTime: AT APPROXIMATELY 1300 HOURSDate: THE SAME DAY
.------------------------------------------------------------------------------------
Figure 4-1. Input/Output Representations
The major part of the system was built in the programming language FORTH, which is an
interactive, incremental system with a low-level semantics which the user can easily and
quickly extend. This allowed the rapid development of the ATN language and control
scheme, as well as the support scheme for the execution of the Event Representation
Language (ERL) algorithms, a formal language written for the purpose of analyzing text.
The ERL algorithms are written in Prolog, a language that is well suited to the specifi-
cation of templates anid the algorithms for instantiating them in ERL. For ease of imple-
mentation, the compiler for the s,'oset of Prolog utilized in this application was written in
SNOBOL 4.
The use of FORTH and the Prolog formalism allowed fairly easy development of the sys-
tem even without the powerful structure manipulation capabilities of a language like
4-5
Toward the end of the I&W III contract, it became clear that the combination of the
grammar and template code compiled from ERL would take up almost all of the available
space in the FORTH dictionary, leaving very little working room for sentence processing.
In fact, it was necessary to partition the templates Into separate files, and process sen-
tences using different templates in different runs of MATRES; even then, only short sen-
tences could be processed.
To remedy this difficulty, a scheme was developed to "overlay" the code for the gram-
mar and templates, so that they would each occupy the same space, but at different
times, since parsing and template matching do not overlap in time. To do this, it was
necessary to design a scheme to allow a portion of the FORTH dictionary to be saved on
disk, and the dictionary to be truncated. The saved segment could later be brought
back into memory at the same locations, and linked into the dictionary in the same way
as when it was saved. Several segments could be created and saved in this way, start-
ing at the same location, and could be restored to memory one at a time as needed.
Several FORTH Words were defined to manage this operation. SEGBASE defines the
starting location of a set of segments; every Word defined after an invocation of SEG-
BASE belongs to the first segment until an invocation of SEGSAVE. SEGSAVE takes as a
parameter the name given to the SEGBASE, and stores the segment on disk starting at
the block number given by the variable DYNBAS, which it updates to the next available
block for the next segment; it also truncates the dictionary at the SEGBASE.
Thereafter, Words are defined, making up the next segment, until the next call of SEG-
SAVE. In this way, several segments are defined. When it is in order to use a particular
segment, SEGLOAD is called with the block number of the desired segment as a parame-
ter; it loads the block starting just after the SEGBASE, making the Words in the segment
available.
4-6 4
Using those commands, MATRES was rebuilt to store the finite-state automaton for lexi-
cal lookup and the ATN grammar in one segment, and the ERL machine and the template
code in the other. A sentence is processed by loading the parse segment, parsing the
sentence to create the parse tree, loading the ERL and temiplate segment, then matching
the temp!ates with the parse tree. This resulted in a very slight slowdown from the ear-
lier scheme, but has allowed more templates of greater complexity to be loaded, and
much longer sentences to be processed.
The ultimate test of a computational system for understanding natural language is its
success in performing some specific task. The goals set out for the current project
have to a large extent been met.
4.2.2 Functional Description. An overview of the MATRES II system organization and
data flow is shown in Figure 4-2. The main system components are: the Lexical Unit
Recognizer, the ATN parser, and the ERL "machine". The direction of the arrows in Fig-
ure 4-2 indicates the general flow of information as a sentence is processed through
the system. The main stages of event record generation are shown across the center
of the figure. Feeding into this are the various analysis components, each compiled from
a source text in a language appropriate to the component.
4-7
i. o U
6..
-C 0.
a I
I-
-4-
4.2.3 The Event Representation Language (ERL). The Event Representation Language is
a formalism which was developed for the purpose of exploring the effectiveness of
"templates" as a knowledge representation technique with which to build systems for
message text analysis in support of I&W functions.
The main function of this language is to guide the mapping process which converts nar-
rative text into formatted event records.
The basic data objects of the Event Representation Language are the templates. As
mentioned in Subsection 2.2.3, we conceive of templates as active memory structures,
which embody hypotheses about objects, facts, events, processes, operations, pro-
cedures and computations required to characterize the links between form and content.
The Event Representation Language is implemented In a subset of Prolog, a formalism
using a clausal form of logic restricted to "Horn" clauses (Warren et al., 1977; Pereira
et al., 1978). Horn clauses may be given both a declarative and a procedural interpre-
tation and are therefore well suited for the expression of concepts in ERL. The basic
computational mechanism of Prolog is a pattern matching process ("unification") operat-
Ing on general record structures ("terms" of logic).
Prolog was Initially developed at the University of Marseilles (Roussel 1975) as :4 practi-
cal tool for 'logic programming' (Kowalski 1974; Colmerauer 1976; van Emden 1975), ard
has since been used in several other centers (Stanford, Edinborough) for writing
language analysis systems (Dahl 1977; Warren 1 977a, Warren 1977b).
Prolog is a perspicuous and powerful language for the expression of the concepts of the
Event Representation Language, and admits of an effective and reasonably efficient
Implementation. Clear, readable, concise programs can be written quickly and with few
errors. Specifically, the following features make it particularly suitable for our purposes:
4-9
a. Pattern matching (unification) replaces the conventional use of selector and con-
o structor functions for operating on structured data.
b. The arguments of a procedure can serve, not only for it to receive one or more
values as input, but also for it to return one or more values as output. Procedures
can thus be "multi-output" as well as "multi-input".
c. The input and output arguments of a procedure do not have to be distinguished in
advance, but may vary from one call to another. Procedures can thus be "multi-
purpose".
d. Procedures may generate (via backtracking, in the case of Prolog) a set of alterna-
tive results. Such procedures are called "non-determinate". Backtracking amounts
to a high-level form of iteration.
e. Procedures may return "incomplete" results, i.e., the term or terms returned as the
result of a procedure may contain variables, which are only filled in later by calls to
other procedures. The effect is similar to the use of assignment in a conventional
language to fill in fields of a data structure. Note, however, that there may be many
occurrences of an instantiated variable, and that all of these get filled in simultane-
ously (in a single step) when the variable is finally instantiated. Note also that
when two variables are unified together, they become identified as one. The effect
is as though an invisible pointer, or reference, linked one variable to the other. We
refer to these related phenomena as the "logical variable".
f. "Program" and "data" are identical In form. A procedure consisting solely of unit
clauses is closer to an array, or table of data, in a conventional language.
4-10
.,*
rim
4.2.4 The ERL Control Mechanism. Prolog provides a remarkably simple form of control,
which suffices for many practical applications.
The declarative semantics of Prolog clauses is such that the order of the goals in a
clause and the order of the clauses themselves are both irrelevant to the declarative
Interpretation. However, these orderings are'generally significant in Prolog, as they con-
stitute the main control information.
When the Prolog system is executing a procedure call, the clause ordering determines
the order in which the different entry points of the procedure are tried. The goal order-
Ing fixes the order in which the procedure calls in a clause are executed. The 'produc-
tive' effect of a Prolog computation arises from the process of 'matching' a procedure
call against a procedure entry point.
Prolog has captured the imagination of many workers in natural language processing and
advanced data base management, and promises to be one of the major languages of the
future.
4.2.5 Advantages of Prolog Representation. This representation has several advan-
tages, among which we might mention the following two: (1) if additional information
needs to be associated with a particular predicate, this can be done simply by adding
another clause; and (2), Prolog provides a uniform way of representing structures and
processes at several levels of grammatical description: syntactic structures, syntactic
normalization, description of objects, descript!on of events, and description of text-level
relations.
Recent investigations reported In the literature show that Prolog is not only used for
grammatical description of structures and processes of natural language, but can also be
used as a practical tool and unifying principle for the description and manipulation of
4-11
data bases.
-* 4.3 Illustration of the Understanding Process
The next three subsections give examples of the various procedural steps involved in
the processing of narrative text, and show how the goals set cut in Section 1 are
approximated.
4.3.1 inputting a Sentence. The understanding process begins when a sentence is input
to the system. This is currently done either from a terminal, or from a disc file. A sen-
tence is input in the format '>> Sentence.'. This is illustrated below:
(a) >> THE SKYLAB ORBITAL WORKSHOP,
A CONVERTED S-4B THIRD STAGE FROM A SATURN-5
LAUNCH VEHICLE, SUCCESSFULLY DEORBITED INTO THE
AUSTRALIAN OUTBACK ON 12 JUL 1979.
The input sentence is received by the Lexical Unit Recognizer, which uses a stored dic-
tionary and the FSA Recognizer to transform the individual words of an input sentence
into a string of lexical units. First, a dictionary look-up process replaces words and
phrases in the sentence with corresponding lexical entries.
Table 4-3 shows the lexical entries for the words "deorbited", "Australian", "outback",
and "launched". The entire lexicon, comprising lexical entries for all three domains, is
contained in Appendix B.
4-12
Table 4-3. Sample Lexical Entries-------------------------------------------------
: : DEORBITED [ VB TRANS PASTP DEORBIT I[ VB TRANS TENSED DEORBIT I .;I
: : AUSTRALIAN [ ADJ NATION I .;
:: OUTBACK [N LOC] .;
:: ( LAUNCH VEHICLE) [ N BOOSTER ) .;---------------------------------------------------------------
Strings which have no entries In the dictionary are passed to the FSA Recognizer, which
attempts to identify them as one of several fixed-pattern categories, (numbers, Zulu-
times, geographic coordinates). In the case of input sentence (a), the strings '12' and
'1979' are recognized as numbe rs and tagged with the features 'N' for noun, and 2DIG
and 4DIG, indicating that they are two-digit and four-digit numbers respectively.
4.3.2 Parsing a Sentence. One of the major problems in constructing automated
language understanding systems is that of transforming the input string of words into a
canonical form which permits semantic interpretation. Part of the transformation process
Involves syntactic analysis. The main purpose of automated syntactic analysis is that of
determining the logical structure of Input sentences (or larger text units). In the course
of constructing a structural description, a syntactic analyzer generally 'regularizes' sen-
tence structure, e.g., It converts sentences and parts of sentences which have proposi-
tional structure into a canonical form. For example, in the current version of the system
passive sentences are converted Into their active form by the syntactic component.
Automated syntactic analysis involves the operation of parsing*. The current version of
the OSI system employs an Augmented Transition Network parser (sentence acceptor),---------- ,
A parser is a formal algorithm which fulfills two functions. One, ittakes a grammar and a lexicon and decides whether a sequence of wordsis a sentence with respect to that grammar and second, it builds astructural representation for that sequence of words.
4-13
and is designed to accept messages from the aircraft activities and missile and satellite
domains. It produces a propositional representation which Is fairly close to surface
structure.
Let us now return to the input sentence under discussion. After lexical lookup, the
string is input to the augmented transition network (ATN) parser, which analyzes it
according to the sublanguage grammar stored in the system. Roughly speaking, the ATN
processor takes the string of lexical entries derived from the sentence, combines them
into phrases, and determines the logical relationships that hold among them. In this
phase of processing the focus is on comparatively local and superficial aspects of sen-
tences such as word order, and the invariant properties of words stored in the lexicon.
The internal representation of the parse tree for sentence (a) is shown in Table 4-4.
Table 4-4. Parse Tree For Sentence (a)
rF:F;E: OLJF UT':L 1 ,ST C) F :
N() 'E : I SN,1 ST OF :
NODEA: 4:DATEI : 652. :1979
1 630. .. .JL..
; 60 ... :1.2E N 1) N 0) I) E
B"6, ..0N
F: N1 I I.. T E" D N 0 D::
N C) D E.' #* F:' F:-NI0 .' F..
I.. SJ " :) I::- N
EI i~j I.. I.'1 NOD 0 y 5 1 N N 0ID
I I :':: I I. :: . . . .I.
EN I N C).') EI IST F)':
I I a I I a 5 4 ") ,, A{J..ITF~I* I... .1 i1 d
1 -N]D I.. US T
NODE: 2:rDFI ;T OF'
. .. . [E' I... :T 4-144-14
i I 1111 ')/\> , TIIEI 0 F I I F. F
* EI N I) I.. I ,I
1 1 1 OF' F
E N D N l c0 I
l". N I N 0 1t U3
I L .I T IF";
I I 9 I ... I N " (
.LI'T)
E iN, f, N 0 I:l'.
V! N 1 .. _. II
N ][ * ,." (11.N I I .0F 1:. ( S I I 1. 1.. D
NI f..
I I . ' .i; *]F:I I" ( " ... :" ' '
E N NCJL I :'E :NC]I)Ei rO,.: 2 2NN:
1... 1 ST C) F:N
N I LIUI I mIS I S I .. ' F:N' 'c:c
ENND.LIST
Ni..IIEi,." 1stNF'FI I I ,I I .I :388,, SATF:N-5
EANOE: 2FN C) IDS T O. I::" N
* * N I36.. AI I I I I I I I I I ::::N I L :":: :
1 1. 'LTIS) OF:
L .JO S) F..lLN ' (EFIIC;I
N_ 0 1) F' CF. 1 13i 1.... 1, (31 -5F
END LIST 1.**ENE'F- NO D*
I END I Tt
34. 366O.M
N Ca IE N P
I IS L8TF F:* *EN *LISEDTLIS
I N0DEN 2 J: E i
N NODE
4- 1f::-
t t I::.E N yL .. I ST
EN N 1NONDED
.. . NDL4-1
E N J) L I
LIST L.ISTNODE 5 INOD
2-', 6 CC)IL>>10 322oSTAG
S: i..LIST OF.:S30() .°'H R
E ND LI..STN 2)E.. A
i L S OF
I I2I4I,, A
I L ISI1::' I Ii;
N:E 2:VG[I O1 N I N )it. leT:
I I I I I < lNI) I.>> DN I I I:.:N L>:
.E N IE N IL S TN 0 I E''0 I:1:1 S T T
NODE;:''+ .4:ACTVI I<<NEILEND NODEo
END NODE
ENrD [.:S3TE ND L IST
tNODE 1,5 C NNOD* <:NI L >
2:12.o. ORBITAL WORKSHOP-11END NODE
L IST O F190.~. SKYL A B
END LISTNODE: 2: DP: LIST OF'#'
1 END LIST
I168.o. THE. .. .LIST OF#*
1 EN ENDoL IST~END NODE
NODE#* 41ACTV
* I * N I XI
* <<NIL>>I...
EEN:D NNOD:EEND NODE
END LIST
4-16
4.3.3 Interpreting the Parse Tree. A parser is a fairly complex mechanism, and it is
therefore unwise to burden It with all the operations required during the syntactic
analysis phrase. Some of the more complex regularizing functions are therefore
offloaded onto the ERL component.
Syntactic normalizing procedures convert intermediate tree structures generated by the
syntactic processor into deep structure trees for logical semantic analysis and event
record synthesis.
The processes relevant here include filling in elements which are missing from the sur-
face structure, resolving syntactic ambiguities, replacing moninalizations with their
corresponding verbal constructions, and generally rearranging the elements in a sen-
tence to regularize its structure.
In the current version of the system, all but the passive restructuring rule are incor-
porated into the ERL formalism, and are intermingled with the other interpretive rules.
The input sentence under discussion does not require restructuring.
As explained in a previous section, the ERL semantic interpretation rules (clauses) are
used top-down, one at a time. Goals in a clause are executed from left to right. If there
are alternative clauses at any point, backtracking will return to them.
The parse tree shown in Table 4-4 is input to the ERL "machine", which uses the pattern
matching process ("unification mechanism') of the Event Representation L .Olage to
produce a set of one or more event records representing the information content of the
Input sentence.
As a first step in the interpretation process, the system activates the system generated
goal 'do', which is currently the top-levei procedure.
4-17
The Top-Level Procedure
------------------------------------------------------
I doU[X,Y,Zl):- build..ER(X,Y,Z,ER), type-.ER(ER).do ([Tree)): _ build-.ER(Tree, ER), type..ER(ER).do([Tree)D: build..YR1(Tree,ER), typeER(ER).
+----------------------------------------------------.4-
The input structure unifies with the head of the third clause, giving rise to two subgoals:
the build-FR1 procedure, and the type-.ER procedure. The 'buildER1 procedure is
illustrated below:
The 'build_ ERl' Procedure for Simple Sentences
------------------------------------------------------ A
1 build-.ER1 (s (Voice, Subjl,Vbgrl,Obj,Compl,Vmods),temp(Name,ER)):-I changel(Subjl, Subj2,Vbgrl,Vbgr2),
find..t.name (Vbgr2, Name),construct (Name, s(Voice,Subj2,Vbgr2,Obj,Compl,Vmods),ER).
.4-----------------------------------------------------------------------------------------------.
Since the input structure under discussion does not require any restructuring, 'chargel'
leaves it unchanged. Next, 'find_ t_ name' identifies the name of the template which is
to be used for interpreting the current input. The relevant information is found in the
main verb of the sentence. Thus, sentence (a), whose main verb is "deorbit", will cause
the DEORBIT template to be activated.
The DEORBIT template as coded in Prolog Is shown below:
DEORBIT Template
+----------------------------------------------
Iconstruct ('DEORBIT',s(Voice,Subj,Vbgr,Obj,Conpl,Vmods),(Verb, AG,011, Loc, Rev, DTG): -
verb(Vbgr,Verb),I agent (Subj, Vmods, AG),
object(Subj,Obj,OBl),location (Obj, Viods, Loc),
I revolution (Vinods, Rev),construct('DTG',Vmcds, 1MG).
.--------------------------------------------
4-18
A--AO84 326 OPERATING SYSTES INC WOODLAND HILLS CA F/6 9W4SATELLITE AND MISSILE DATA GENERATION FOR AIS. CU)DEC 79 6 M SILVA, C A MONTGOMERY F30602-78-C-o274
UNCLASSIFIED OSI-R79-037 RADC -TR-79-314 ML
23 llllljffffff
All templates are encoded as Prolog "construct" clauses. The head of a "construct"
clause has three arguments: a template name, the name of the syntactic constituent
which serves as the context which is searched in an attempt to find fillers for the
descriptor slots of the template in question, and a third argument which represents the
output of the procedure, i.e., the instantiated slots.
The body of a 'construct' clause consists of several 'goals', corresponding to the 'slots'
of a template. In the case of the DEORBIT template, the body of the "construct" clause
consists of six "goals", each defined as a procedure which actively seeks suitable fill-
ers for the descriptor slot it represents.
In the case of the example input sentence, each of the six "slots" actively searches
the specified context in an attempt to find a component which can serve as a "filler".
Since there Is no Agent specified in the input, the Agent slot returns 'nil'. Next, the
'object' slot constructs a record for the subject nounphrase, which It decomposes into
an 'Equipment' component, and a 'Relative' component.
The location clause Illustrated below Identifies the prepositional phrase "INTO THE AUS-
TRALIAN OUTBACK", as the location of the deorbit.
The 'revolution' procedure also fails to find a filler, and returns 'nil'. The DTG procedure
cannot find a filler for Its Time component, but Identifies the prepositional phrase ON 12
JUL 1979 as the deorbit date.
The result of the Interpretive process Is the "Instantiated" template -- or Event Record
-- shown In Table 4-5.
4-19
Rop-
The 'Location' Clause+-----------------------------------------------------------------------------.
IlocationCNP,List, slot('Location=',X)): -II locatl(NP,Xl),I
searchlist (List, X2),I concatenate(X1,X2,X).
I locatiCNP, [NP]):- test...nhead(NP, 'LOG').IlacatilC,nil).Isearchlist([I, .. List], (X,..L]):-
searchloc(M,X), searchli'stCList, L).I searchlistCL, . List], L): - searichlist (List, L).I searchlistL-,nil).I searchloc(ppCLl, Prep,NP),(Li, Prep,NP]): -
I member (P, ('ALONG', 'AT', 'EAST OF',1N, 'INTO', II 'NEAR', 'ON', 'SOUTHEAST OF', 'OUTSIDE OF',II 'WEST OF']),I lexeqCPrep,P), test..nhead (NP, 'LOC').I
+-----------------------------------------------------------------------------
Table 4-5. Event Record for Sentence (a).
+-----------------------------------------------------------------------------------------
I Event: DEORBITIAction= SUCCESSFULLY DEORBITEDIObject: SATELLITEI.. . Equipment= SKYLAB ORBITAL WORKSHOPI... Number=I... Relative= A CONVERTED S--4B THIRD STAGE FROM A SATURN-5 LAUNCH I
VEHICLEI Location= INTO THE AUSTRALIAN OUTBACKI Date= ON 12 JUL 1979+-----------------------------------------------------------------------------------------+
As mentioned before, Event Records are template-derived, event-centered data struc-
tures In which the information conveyed by the Input can be viewed from the perspec-
tive of time, location, type of activity, object(s) Involved, etc. They provide content
* representations for Individual sentences describing atomic events, and form the building
biocks of message content representations which will eventuaily provide answers to
* Question 1 of Section 1: "What Is the content of the Message?"
4.3.4 Identifying the Reported Source of an Event Report. Certain events are officialiy
announced. Thus a Soviet Satellite L~unching may be announced as follows:
4-20
(b) THE SOVIET NEWS AGENCY TASS ANNOUNCED
THAT IMPACT OF COSMOS-954 TOOK PLACE
NEAR YELLOWKNIFE, CANADA.
The system as developed to date Identifies the source of thin, report and Interpret TASS
as the "infosource" of the launch event. Sentence (b) yields the event record shown In
Table 4-6.
Table 4-8. Event Record for Sentence (b)4.------------------------------------------------------------------------4
lnfosource= THE SOVIET NEWS AGENCY TASSlEvent: IMPACTlAction= IMPACTlObject: SATELLITE
I... Equipiment =.COSMOS-954... Number=
I .. Relative=iLocation= NEAR YELLOWKNIFE CANADAI
4-------------------------------------------------------------------------+
4.3.5 Identifying the Reported Status of an Event. Some events are reported as
expected, or as having failed In some sense. The following two examples illustrate how
the current version of the system processes sentences expressing these notions.
(c) THE ORBITAL WORKSHOP SKYLAB WAS
EXPECTED TO DEORBIT OVER CANADA.
Table 4-7. Event Record for Sentence (c)4.------------------------------------------------------------------------
15tatus= EXPECEDlEvent: DEORBITlAction= DEORBITlObject: SATELLITE
j I... Equipment -ORBITAL WORKSHOP SKYLABII... Number.
.Relative-ILocationp. OVER CANADA
+----------------------------- --------------------------------------------.
4-21
(d) THE ORBITAL WORKSHOP SKYLAB FAILED
TO DEORBIT INTO CANADA
Table 4-8. Event Record for Sentence (d)--------------------------------------------------------
I Status=FAILEDlEvent: DEORBITIAction= DEORBITIObject: SATELLITEI ... Equipment = ORBITAL WORSHOP SKYLABI ... Number=I ... Relative=ILocation= INTO CANADA
---- --------------------------------------------------------
As pointed out in the section on the Characteristics of Messages, the key event
described in a message is usually mentioned in the first sentence. This sentence Intro-
duces the TOPIC of the message, i.e., what the message is about.
In order to test and evaluate the capabilities of the system, a set of sentences were
constructed modeled on the syntactic properties of first sentences of the various mes-
sage types Identified In the missile and satellite domain. The sentences were so
designed as to test various aspects of the syntactic structure of the sublanguage.
Examples of sentence types currently processed by the system together with their
corresponding event records are offered In Appendix (E).
4-22
6.0 REFERENCES
Abelson, R.P. The Structure of Belief System-. R. C. Schank and K.M. Colby (eds) Com-puter Models of Thought and Language. San Francisco: Freeman Press, 1973
Abelson, R.P. Concepts for Representing Mundane Reality in Plans. D. Bobrow and A.Colins (eds) Representation and Understanding: Studies in Cognitive Science. NewYork: Academic Press, 1975 Akmajian, A. The Role of Focus in the Interpretation of Ana-phoric Expressions. In S. Anderson and R Kiparsky (Eds.). A Festschrift for Morris Halle.New York: Holt, Rinehart and Winston, 1973, 215-226.
Akmajian, A. and Heny, F. An Introduction to the Principles of Transformational Syntax.Cambridge, Massachusetts: MIT Press, 1976.Akmajian, A. and Jackendoff, R. Coreferentiality and Stress. Linquistic Inquiry, 1(1),
1970, 124-126.
Allerton, D. The Notion of 'Givenness' and its Relations to Presupposition and to Theme.Lingua, 44(2/3), February/March 1978, 133-168.
Alpac (Automated Language Processing Advisory Committee) Language and Macilines -Computers in Translation and Linguistics National Academy Science, Washington, D.C.1966.
Altshuller, Gene & Plagman, B User/System Interface within the Context of anIngegrated Corporate Data Base National Computer Conference 1974
Anderson, S. Pro-sentential Forms and Their Implications for English Sentence Structure.(1) In: McCawley 1976, 166-200. (2) also published as: mimeo, Bloomington, Indiana:Indiana University Linguistics Club, 1972.
Austin, D. Perspective Paper: Library Science, In Walker, D.E., H. Karigren, and Martin Kay(eds.) Natural Language In Information Science. FID Publication 61, Skriptor, Stockholm,Sweden.Austin, J. L. How to do Things with Words. Oxford University Press, London, 1962.
Baker, J.D., Mace, D.J., & McKendry, J.M. The Transform Operation in TOS: Assessment ofthe Human Counterpart Technical Research Note 212 U.S. Army Research Institute,Arlington, VA August 1969
Baranofsky, S. Some Heuristics for Automatic Detection and Resolution of Anaphora inDiscourse. Unpublished MSc thesis, Department of Computer Science, University ofTexas, Austin, January 1970.
Bar-HIllel, Yehoshua "Indexical Expressions" Mind Vol. 63 1954, pp 369-79
Barrett, G.V., Thornton, C.L. & Cabe, P.A. Human Factors Evaluation of a Computer BasedInformation Storage and Retrieval System Human Factors 1968, 10 (4) 431-436
Bartlett, F.C. Remembering. Cambridge University Press, Cambridge, 1932
Bartsch, R. Syntax and Semantics of Relative Clauses. In R. Bartsch, J. GroenendiJk &M. Stokhof (Eds.), Amsterdam Papers on Formal Grammars. The Netherlands: Universityof Amsterdam, 1976. de Beaugrande, Robert. Introduction to Textllngulstlcs (provisionaltitle), Hillsdale, N.Y. Erlbaum, (in press).
6-1
Benzon, W. & Hays, D.G. Computational Linguistics and the Humanist Computers and theHumanities Volume 10, Number 5 September/October 1976
Bever, T.G., Ross, J.R. Underlying Structures in Discourse Proceedings of the Confer-ence on Computer-Related Semantic Analysis Las Vegas, 1965
Bever, Thomas G. & Ross, John R. Underlying Structure in Discourse Unpubliched MITpaper 1976
Bierwisch, M. Review of Discourse Analysis Reprints Linguistics April 1965 pp. 61-73
Boden, M. Artificial Intelligence and Natural Man. Hassocks: Harvester Press, 1677.
Bobrow, Daniel G. and A. Collins (eds.). Representation and Understanding. AcademicPress, New York, 1975.
Bobrow, D.G. & Norman, D.A. Some Principles of Memory Schemata D.G. qbrcwj & A.M.Collins (eds) Representation and Understanding Academic Press, NY 1975
Bobrow, D. G., and Winograd, T., An Overview of KRL, A Knowledge Repre ZcntationLanguage, Cognitive Science, Vol. 1, No. 1, 1977.
Boles, S.J. & Gould, J.D. Syntactic Errors in Computer Programming Human Factos16:253-257 1974
Bowen, R.J. Halpin, J.A., Long, C.A. Lukas, G., Mullarkey, M.M. & Triggs, T.J. Decision FlowDiagrams and Data Agregation in Army Tactical Intelligence Report No. 2570 Bolt,Beranek & Newman Cambridge, MA June 1973
Boyce, R.F., Chamberlin, D.D. King III, W.F. Hammer, M.M. Specifying Queries as RelationalExpressions: The Square Data Sublanguage Communications of the ACM Volume 18,Number 11 November 1975
Bransford, J. and Johnson, M. Considerations of Some Problems of Comprehension. In:W. Chase (Editor). Visual Information Processing, New York: Academic Press, 1973,383-438.
Bransford, J.D., McCarrell A Sketch of a Cognitive Approach to Comprehension W.Weimar & D. Palermo (eds) Cognition and the Symbolic Processes Hillsdale, New JerseyLawrence Erlbaurm Assoc. 1974
Bresnan, J. A Note on the Notion "Identity of Sense Anaphora". Linguistic ;nquiry, 1971,2(4), 589-597.
Brooks, R. A Model of Cognitive Behavior In Writing Code for Computer ProgramsProceedings of the 4th International Joint Conference In Artificial Intelligence 1975
Brown, J.S., Burton, R.R. "Multiple Representations of Knowledge for Tutorial Reasoning"D.G. Bobrow & A. Collins (eds) Representation and Understanding Chapter 11 New York:Academic Press 1975
Brown, J.S., Burton, R.R., Bell, A.G. Sophie: A Step Toward Creating a Reactive LearningEnvironment International Journal of Man-Machine Studies September 1976, pp 675-696
Browse, R. A Knowledge Identification Phase of Natural Language Analysis. MSc the3s,
Department of Computer Science, University of British Columbia, January 1977.
Browse, R. Knowledge Identification and Metaphor. Proceedings of the Second NationalConference, Canadian Society for Computational Studies of Intelligence/Societe
6-2
L _ __ __ _
canadlenne des etudes d'intelligence par ordinateur. Toronto, July 1978, 48-54.
Bruce, B. A Model for Temporal References and its Application in a Question AnsweringProgram. Artificial Intelligence. 3(1), 1972, 1-25.
Bullwinkle, C. See also Sidner, C.
Bullwinkle, C. The semantic component of PAL: The Personal Assistant Language Under-standing Program. Working paper 141 Artificial Intelligence Laboratory, MassachusettsInstitute of Technology, March 1977.
Bullwinkle, C. Levels of Complexity in Discourse for Anaphora Disambiguation and SpeechAct Interpretation. (1) Proceedings of the Fifth International Joint Conference on Artifi-cial Intelligence. Cambridge, Massachusetts, August 1977, 43-49. (2) an earlier ver-sion was published as: Memo 413, Artificial Intelligence Laboratory, Massachusetts Insti-tute of Technology, May 1977.
Bundy, A. Artificial Intelligence: An Introductory Course. (1) Elsevier North-Holland,1979. (2) Edinburgh University Press, 1978.
Burton, R. Semantic Grammar: An Engineering Technique for Constructing NaturalLanguage Understanding Systems. (BBN Report No. 3433). Cambridge MA: E0t Beranekand Newman Inc., 1976.
Caramazza, A.; Grober, E.; Garvey, C. and Yates, J. Comprehension of Arlaphoric Pro-nouns. Journal of Verbal Learning and Verbal Behavior, 16(5), October 1977, 601-609.
Carbonell, Jaime Mixed-Initiative Man-Computer Instructional Dialogues Ph.D. Disserta-tion, M.I.T. June 1970
Carlson, G. and Martin, L. This Antecedent Isn't the Right One. Glossa, 9(1), 1975, 13-24.
Carnap, R. Meaning and Necessity. 2nd edition. Chicago, 1956.
Carnap, R. & Jeffrey, R.C. (eds) Studies in Inductive Logic and Probability Vol. 1 Univer-sity of California Press, 1971
Carnap, R. The Logical Foundations of Probability University of Chicago Press, 1950
Carroll, J. and Bever, T. The Non-Uniqueness of Linguistic Intuitions. Research reportRC6938 (#29749), Thomas J. Watson Research Center, IBM, Yorktown ' s .....York, 10, January 1978. (Submitted for publication.)
Caton, C. (Editor). Philosophy and Ordinary Language. Urbana: University of IllinoisPress, 1963.
Celce-Murcia, M. Paradigms for Sentence Recognition AJCL 1975
Chafe, W. Meaning and the Structure of Language. The University of Chii no P:ess,1970.
Chafe, W. Discourse Structure and Human Knowledge. In: R. Freedle and .1. Crrroll (Edi-tors). Language Comprehension and tile Acquisition of Knowledge. Washington: V. H.Vinton, 1972, 41-70
Chafe, W. Language and Consciousness. Language, 1974, 50(1), 111-133.
Chafe, W. Glvenness, Contr-stiveness, Definiteness, Subjects, Topi:cs ai ! Pointr orView. In C. LI (Ed.), Subject and Topic. New York: Academic Press, 1976.
5-3
Chafe, Wallace L. Some Thoughts on Schemata Mathematical Social Sciences BoardSymposium Cambridge, MA 1975, Preprints Theoretical Issues in Natural Language Pro-cessing R. Schank & B. L. Nash-Webber (eds) 1975
Chamberlin, D. D. & Boyce, R. F. Sequel: A Structured English Query Language IBM
Technical Report RJ1 394 May 1974
Chandioux, J Meteo: An Operational System for the Translation of Weather Forecasts,Proc. FBIS Sem. on the Present State of MT 3/8/-9/76, to appear in AJCL
Charniak, E. Organization and Inference In a Frame-like System of Common-SenseKnowledge Proc. Conf. on Theoretical Issues in Natural Language Processing Cambridge,MA 1975.
Charniak, E. Towards a Model of Children's Story Comprehension. (Technical Report266). Cambridge MA: MIT Artificial Intelligence Laboratory, 1972.
Charniak, E. Context and the Reference Problem. In R. Rustin (Ed.), Natural LanguageProcessing. New York: Algorithmics Press, 1973.
Charniak, E. Inference and Knowledge. In: E. Charniak and Y. Wilks (Eds.). Computa-tional Semantics: An Introduction to Artificial Intelligence and Natural LanguageComprehension. Fundamental Studies in Computer Science 4. Amsterdam: North-Holland,1976, 1-21 and 129-154. An earlier version of parts of this work was published as:Organization and Inference in a Frame-Like System of Common Sense Knowledge. In:Schank and Nash-Webber, 1975, 46-55. Another early version was published in: CourseNotes for a Tutorial on Computational Semantics, given at the Institute for Semantic andCognitive Studies, Castagnola, Switzerland. 17-22 March 1975. A revised but abbrevi-ated version appears as: Inference and Knowledge in Language Comprehension.Machine Intelligence, 8, 1977, 541-574.
Charniak, E. With Spoon in Hand This Must Be the Eating Frame. In: Waltz 1978, 187-193.
Chipman, H. and de Dardel, C. Developmental Study of the Comprehension and Productionof the Pronown "it". Journal of Psycholinguidtic Research, 1974, 3(2), 91 -99.
Chomsky, C. The Acquisition of Syntax in Children from 5 to 10. Cambridge MA: MITPress, 1969.
Chomsky, N. Syntactic Structures. Janua Linguarum 4. The Hague: Mouton, 1957.
Chomsky, N. Aspects of the Theory of Syntax. Cambridge, Massachusetts: Mit Press,1985.
Clark, H. In B. Nash-Webber and R. Schank (Eds.), Theoretical Issues in Naturel
Language Processing. Cambridge MA, 1975.
Clark, H. and Haviland S. Comprehension and the Given-New Contract. :1; R. Fre-dle(Ed.). Discourse Production and Comprehension. Discourse Processes: Advances inResearch 1. Norwood New Jersey: Ablex Publishing, 1977.
Clark, Herbert H. Bridging in TINLP 1975
Clark, H. and Marshall, C. Reference Diaries. In: Waltz 1978, 57-63.
Codd, E. F. Seven Steps to Rendezuous with the Casual User Report RJ 1333 IEMResearch Laboratory San Jose, CA January 1974
6-4
,, _ i l lam i -- . . .. , _ o ; -. -. . ., ;
.1 P p1wm -
Cohen, P. On Knowing What to Say: Planning Speech Acts. Technical report 118/PhDthesis, Department of Computer Science, University of Toronto, January 1978.I Cohen, P. and Perrault, C. Preliminaries for a Computer Model of Conversatkn.Proceedings of the First National Conference, Canadian Society for Computational Stut-dies of Intelligence/Societe canadienne des etudes d'intelligence par ordinateur. Van-
couver, British Columbia, August 1976, 102-111.
Cohen, R. Computer Analysis of Temporal Reference. Technical report 107, Department
of Computer Science, University of Toronto, December 1976.
Cole, P. and Morgan, J. (Eds.). Syntax And Semantics 3: Speech Acts. New York:Academic Press, 1975.
Collins, A., et al Reasoning from Incomplete Knowledge Representation and Understanc-Ing 1975
Colmerauer, A. Les Grammaires de Metamorphose. Groupe d'lntelligence Artificielle, Mar-seille, Marseille-Luminy, Nov. 1975.
Corum, C. Anaphoric Peninsulas. Papers from the Ninth Regional Meeting, ChicagoLinguistic Society, 1973, 89-97.
Craig, J. A., S. C. Berezner, H. C. Carney and C. R. Longyear. DEACON: Direct EnglishAccess and Control. Proc. AFIPS 1966.
Culicover, P. A Constraint on Coreferentiality. Foundations of Language, 1976, 14(1).
Cullingford, R.E. An Approach to the Representation of Mundane World Knowledge: TheGeneration and Management of Situational Scripts American Journal of ComputationalLinguistics
Dahl,V. Un Systeme Deductif d'Interrogation de Banques de Donnees en Espagnol. Grouped'lntelligence Artificielle, QLniversite de Marseilles-Luminy, Nov,1977.
Dascal, M. & Margalith, A. A New Revolution in Linguistics? 'Text Grammar' vs 'Sen-tence Grammar' Theoretical Linguistics Vol. 1, No. 1/2 1974
Date, C. J. An Introduction to Database Systems Reading, MA: Addison - Wesley 1976
Davidson, J. An Evaluation of Conceptual Dependency as a Representation for Languageand Knowledge. Unpublished report, Department of Computer Science, University of Brit-Ish Columbia, February 1976.
Deutsch, B. Discourse Analysis and Pragmatics. In D. Walker (Ed.), Speech Understand-Ing Research - SRI Annual Technical Report, April 1974 - March 1975. (a)
Deutsch, B. Establishing Context in Task Oriented Dialogs. American Journal of Compu-tational Linguistics, 1975, 4. (b)
Diller, T. (Ed.) Proceedings of the 13th annual meeting, Association for Computational
Linguistics. American Journal of Computational Linguistics. Microfiche 32-36, 1975.
Donnellan, K. Reference and Definite Descriptions. Philosophical Review, LXXV, July
1966, 281-304.
Dressier, W. & Schmidt, S. Textilngulstik, Kommentlerte Bibllographle Munchen: WilhelmFink Verlag 1973
6-6
b.1
Dressier, Wolfgang U. Current Trends in Textlinguistics. de Guyter, Berlin/NY, 1 978a.
Dressier, W. U. Textlinguistik. Wissenshaftliche Buchgesellshaft, Darmstadt, 1 C78b.
Dundes, A. The Morphology of North American Indian Folktales Folklore Fellowc Communi-cations LXXXI, Helsinki 1964
Edes, E. Output Coanditions in Anaphoric Expressions With Split Antecedents. Unpub-lished MS, Harvard University, 1968.
Edmondson, J. Semantics, Games and Anaphoric Chains. In R. Bartsch, J. Groz;endijk &M. Stokhof (Eds.), Amsterdam Papers on Formal Grammars (Vol.1). The Netherlands:University of Amsterdam, 1976.
Eisenstadt, M. Processing Newspaper Stories: Some Thoughts on Fighting and Stylistics.Proceeding of the AISB Summer Conference, Society for the Study of Artificiel Intelli-gence and the Simulation of Behaviour, July 1976, 104-117.
Engel, E.S. & Granda, R.E. Guidelines for Man/Display Interfaces, IBM Technical ReportTR 00/2720 December, 1975.
Feys, R., & Fitch, F.B. (ed) Dictionary of Symbols of Mathematical Logic North-HollandPublishing Co., Amsterdam 1969.
Fillmore, C. The Case for Case. In: E. Bach and R. Harms (Eds.). Universals in LinguisticTheory. New York: Holt, Rinehart and Winston, 1968, 0-88 (sic).
Fillmore, C. Deixis I (and) Deixis Ii. Unpublished lectures, mimeo, 1972.
Fillmore, C. The Case for Case Reopened. In: P. Cole and J. Saddock (Eds.). Syntaxand Semantics 8: Grammatical Relations. New York: Academic Press, 1977, 59-81.
Fillmore, C.J. An Alternative to Checklist Theories of Meaning Proceedings of the FirstAnnual Meeting of The Berkeley Linguistics Society Cathy Cogen, Henry Thompson, Gra-ham Thurgood, Kenneth Whistler & James Wright (ed) Berkeley, CA: University of Cali-fornia at Berkeley 1975, pp. 123-131
Fillmore, C.J. Deictic Categories in the Semantics of "Come" Foundations of Language2:219-227 1966.
Fillmore, C.J., Verbs of Judging: An Exercise in Semantic Description, Fillmore & Langen-doen, Studies in Linguistic Semantics, New York, Hold, Rinehart & Winston, 1971.
Firbas, J. On Defining the Theme in Functional Sentence Analysis. Travaux Linquistiquesde Praque, 1, 1964, 267-280.
Foley, J.D. & Wallace, V.L. The Art of Natural Graphic Man- Machine ConversationProceedings of the IEEE 62 462-471 1974.
Fowler, H. A Dictionary of Modern English Usage (second edition, revised by Sir ErnestGowers). Oxford University Press, 1968.Friedman, J.: Moran, D. and Warren, D. Two papers on semantic interpretation in Mon-
tague grammar. American Journal of Computational Linguistics, Microfiche 31, 1978.
Gardin, J-C. Document Analysis and Linguistic Theory. Journal of Documentation, Vol 29.No 2, June 1973.
Garvey, C. and Caramazza, A. Implicit Causality In Verbs. Linquistic Inquire, 5(3), Sum-mer 1974, 464-469.
6-6
Garvey, C., Caramazza, A. and Yates, J. Factors Influencing Assignment of PronounAntecedents. Cognition, 1974, 3(3), 227-244.
Gerrity, T.P. Design of Man-Machine Decision Systems An Application to PortfolioManagement Sloan Management Review Winter 1971
Geach, P. Reference and Generality. New York: Cornell University Press, 1962.
Gelbart, R. Generative Semantics. Unpublished report, Department of Computer Science,University of British Columbia, 1976.
Gimpel, J. Algorithms in SNOBOL4, Wiley & Sons, N.Y., 1976.
Givon, Talmy. The Time-Axis Phenomenon, Language 49, 4, 1973.
Givon, T. Topic, Pronoun and Grammatical Agreement. In: Li 1975, 149-18.
Goffman, Erving. Frame Analysis Harper, 1974
Goldman, N. M. Computer Generation of Language from a Deep Conceptual Base. PhD.thesis/Al memo 247, Artificial Intelligence Laboratory, Stanford University, 1974.
Goldman, N.M. Conceptual Generation. in: Schank 1975, 289-371.
Goldstein, I.P. and Roberts, R. B. NUDGE: A Knowledge-Based Scheduling Program. (1)Proceedings of the Fifth International Joint Conference on Artificial Intet:ijence. Cam-bridge, Massachusetts, August 1977, 257-263. (2) Memo 405, Artificial IntelligenceLaboratory, Massachusetts Institute of Technology, February 1977.
Gordon, D. And Lakoff, G. Conversational Postulates. (1) Papers from the SeventhRegional Meeting, Chicago Linguistic Society, 1971, 63-84. (2) In: Cole and Morgan1975, 83-106.
Gould, J.D. & Ascher, R.M. Use of an (OF-Like Query Language by Non-Programmers IBMResearch Center Yorktown Heights, New York Report RC 5267 February 1975.
Grice, H. P. Logic and Conversation. Unoublished Lectures. University of California,Berkeley, (mimeo). Partly published in Cole and Morgan (eds). 1975, 41-55.
Grishman, R. Impementation of the String Parser of English Natura! Lang-nag c e. n.R. Rusting (ed) Algorithmics Press, NY 1973a
Grimes, J. E. The Thread of Discourse. (1) Janua Linguarum, series minor 207. TI,:,Hague: Mouton, 1975. (2) An incomplete earlier version was published as: Technic-IReport 1, Department of Modern Languages and Linguistics, Cornell Unh'cr- I'y, i P72.
Grinder, J. T. and Elgin, S. H. Guide to Transformational Grammar: History, Theory, Prec-tive. New York: Holt, Rinehart and Winston, 1973.
Grinder, J. & Postal, P. Missing Antecedents. Linguistic Inquiry, 1971, 2(3). ""-12.
Grishman, R. Sager, N. Raze, C., Book-chin, B. The Linguistic String Parser Prc c. 19-93National Computer Conference AFIPS Press, Montvale, NY 1973
] Grishman, Ralph A Survey of Syntactic Analysis Procedures for Natural LNr,",ge eu'.,York, New York: Courant Institute of Mathematic Sciences, New York Uni-c ity R p":rNo. NSO-8 August 1975 pp 94
Grober, Ellen H.; Beardsley, W. fnd Crara7za, A. Parallel Function Sfratr,! Prenlo'rAssignment. Cognition, 6(2), 1978, 11 7-133.
5-7
t~.. - - ;
.1
Grosz, B. J. The representation and use of focus in a system for understanding dialogs.(1) Proceedings of the Fifth International Joint Conference on Artificial Intelligence.Intelligence Center, SRI International, June 1977.
Grosz, B. J. The representation and use of focus in diaogue understanding. (1) Unpub-lished PhD thesis, Department of Computer Science, University of California at Berkeley,June 1977. (2) Also published, slightly revised, as: Technical note 151, SRI interna-tional, Artificial Intelligence Center, July 1977. (3) A newer revised version appears in:Walker 1978, Section 4. (4) An earlier verion of parts of this work were published in:Walker 1976, Chapters VIII-X.
Grosz, B. J. Focusing in Dialog. (1) in: Waltz 1978, 96-103. (2) Technical note 166,Artificial Intelligence Center, SRI International, July 1978.
Grosz, B. J. Focussing and Description in Natural Language Dialogues. Technical Note185, SRI International, Menlo Park, California, April, 1979.
Halliday, M. A. K. Notes on Transitivity and Theme in English: Part 2. Journal of Linguis-tics, 3, 1967, 199-244.
Halliday, M. A. K. and Hasan, R. Cohesion in English. Longman English Language Series 9.London: Longman, 1976.
Halpin, S.M., Johnson, E.M., & Thornjerry, J.A. Cognitive Reliability in Mannel System IEEETransactions on Reliability 1973, R-22 (3), 165-170
Hankamer, J. On the Nontransformational Derivation of Some Null VP Anaphors. LinguisticInquiry, 9(1), Winter 1978, 66-74.
Hankamer, J. & Sag, I. Deep and Surface Anaphora. Linguistic Inquiry, 1976, 7(3),391-428.
Hansen, W.J. User Engineering Principles for Interactive Systems Proceed;ngs of Fa!lJoint Computer Conference 1971, 523-532
Harris, L. R. User Oriented Data Base Query with the ROBOT Natural language query sys-tem. International Journal of Man-Machine Studies, 9(6), November 1977, 697-713.
Harris, L. R. Status Report on the ROBOT Natural Language Query Processor. SIGARTNewsletter, Number 66, August 1978, 3-4.
Harris, Z. Discourse Analysis Language, 28 1952 pp. 1-30
Harris, Zellig S. Discourse Analysis Reprints The Hague: Mouton & Co., 1963 First Pub-lished in 1952
Harris, Zellig S. Methods in Structural Linguistics Chicago: The University of ChicagoPress 1951
Hirris, Z.S. Discourse Analysis Lg. 28 The Structure of Language (Reprinted in Fozor &Katz, eds,) NY 1964
Haviland, S. & Clark, H. What's New? Acquiring New Information as a Proce7.s inComprehension. J of Verhal Learning and Vethal Behavior, 1974, 1.3(, 5:,_31.
Hayes-Roth H. & Mostow D.J. Organization and Control of Syntactic, Serintic,Inferential and Wcrld Knowvcdge fo: Lang_,age Understanding Cooline 10-7 (p-C4 it)
5-8
Heidorn, George E. Natural Language Inputs to a Simulation Programming System 14PS-
65HD, Naval Post Graduate School Monterey, CA 1972
Hendrix, G.G. Partitioned Networds for Modelling Natural Language Sem.ntics Dissertc-tion, University of Texas, Dept. Comp. Science 1976
Hendrix, G.G., Thompson, C. & Slocum, J. Language Processing Via Canonical Verbsand Semantic Models Proc. 31JCAI, Menlo Park, CA Stanford Res. Inst. 1073
Hendricks, W. 0. Current Trends in Discourse Analysis Current Trends in S2',,istics BKachru & Stahlke (ed) Edmonton: Linguistic Research, Inc.
Hendricks, W. 0. Grammars of Style and Styles of Grammar. North-Hollond studies intheoretical poetics 3. Amsterdam: North-Holland, 1976.
Hendrix, G. G. Partitioned Networks for the Mathematical Modeling of Natural La.nguaceSemantics. Technical report NL-28, Department of Computer Sciences, Uriversity ofTexas, Austin, 1975.
Hendrix, G. G. Expanding the Utility of Semantic Networks Through Partitioning. AdvancePapers of the Fourth International Joint Conference on Artificial Intelligence. Tb;isi, Unionof Soviet Socialist Republics, September 1975, 115-121.
Hendrix, G. G. The Representation of Semantic Knowledge. In: Walker 1978, 121-181.
Hewitt, C. Description and Theoretical Analysis of Planner MIT-AI-258 1972
Hewitt, Carl A Universal Modular ACTOR Formalism for Artificial Intelligence Third Inter-national Joint Conference on Artificial Intelligence Stanford 1973 pp 235-245
Hinds, J. Paragraph Structure and Pronominalization. Papers in Linguistics, 10(1-2),Spring-Summer 1977, 77-99.
Hirst, G. J. Artificial Intelligence and Computational Linguistics I1: Methodojogy and prob-lems. Unpublished report, Department of Computer Science, University of British Colum-bia, April 1976.
Hirst, G. J. Anaphora in Natural Language Understanding: A Survey. Technical Report79-2, Department of Computer Science, University of British Columbia, May, 1979.
Hirst, G. J. Focus in Reference Resolution in Natural Language Understanding. Paperpresented at the Language and Speech Conference, Melbourne, November 1 977.
Hirst, G. J. Cohesive Discourse Transitions and Reference Resolution: The Cinema Meta-phor and Beyond into the Transfinite. Unpublished mantscript, 20 December 1977.
Hirst, G. J. Report on the Conference on Theoretical Issues in Natural LanguageProcessing--2. AISB quarterly, number 31, September 1978, 9-11.
Hlrst, G. J. A Set of Primitives for Discourse Transitions. Unpublishable manuscrpt, 1February 1978.
Hlrst, G. J. Humorous Information Processing: Why Al Must Consider Humor. In prepara-tion.
Hobbs, J. R. (1) Pronoun Resolution. Research report 76-1, Department of ComputerSciences, City College, City Univerity of New York, August 1976. (2) An abridged ver-sion was published as: Resolving Pronoun. References. Lingua, 44(4), April 1978, 311-338.
5-9
q R R
Hobbs, J. A Computational Approach to Discourse Analysis. (Research Report 7G-2).Department of Computer Science, City College, City University of New York, Dacc ' er,1976.
Hobbs, J. R. 38 Examples of Elusive Antecedents from Published Texts. ReseirchReport 77-2, Department of Computer Sciences, City College, City University of ,:cwYork, August 1977.
Hobbs, J. R. Coherence and Coreference. Technical Note 168, Artificial Intelligc;ceCenter, SRI International, 4 August 1978.
Hornby, P. A. Surface Structure and the Topi-Comment Distinction: A Developmcn;alStudy. Child Development, 42, 1971, 1975-1988.
Hornby, P. A. The Psychological Subject and Predicate. Cognitive Psychology, 3, 1'2,632-642.
Huggins, A.W.F. Sentence Syntax. On R. Spiro, B. Bruce & W. Brewer (Eds.), ThecrcCalIssues in Reading Comprehension. New Jersey: Lawrence Erlbaum Associates, in prcC.
Hutchins, W. J. Machine Translation and Machine-Aided Translation. Journal of Docuinen-tation, 34(2), June 1978, 119-159.
Huxley, R. The Development of the Correct Use of Subject Personal Pronouns in TwoChildren. In G. Flores d'Arcais & W. Levelt (Eds.), Advances in Psycholinguistics. TheNetherlands: North-Holland Publishing Co., 1970.
Jackendoff, Ray Semantic Interpretation in Generative Grammar MIT Press, 1972
Jespersen, 0. A. Fodern English Grammar on Historical Principles, George Allen and Unwin,London, 1914-29.
Jacobsen, Bent Transformational-Generative Grammar. North-Holland Linguistic Series17. Amsterdam: North-Holland, 1977.
Johnson-Laird, P. N. The Choice of the Passive Voice in a Communicative Task. BritishJournal of Psychology, 59, 1968, 7-15.
Johnson-Laird, P. N. The Interpretation of the Passive Voice. Quarterly Journal ofExperimental Psychology, 20, 1968, 69-73.
Johnson-Laird, P.N. Models of Deduction F. Falmagne (ed) Reasoning: Representationand Process Hillsdale, N.M. Erlbaum
Joshi, Aravind K., & Ralph M. Weischedel Computation of a Subclass of Inferences:Presupposition and Entailment, Technical Report #74 Information & Computer ScienceDept., University of California, Irvine, Irvine, CA 1976
Kahn, K. and Gorry, G. A. Mechanizing Temporal Knowledge. Artificial Intelligence, 9(1),August 1977, 87-108.
Kantor, R. N. The Management and Comprehension of Discourse Connection by PronounsIn English PhD thesis, Department of Linguistics, Ohio State University, 19 7 7.
Karttunen, Lauri Implicative Verbs Language 47 pp 340-358 1971
Karttunnen, Lauri On the Semantics of Complement Sentences, Papers from the SixthRegional Meeting of the Chicago Linguistic Society Chicago: University of Chicago,1970
5-10
Karttunen, Lauri & Stanley Peters Conventional Implicature in Montague Grammar,Presented at the First Annual Meeting of Berkeley Linguistic Society, Berkeley, CA 15February 1975.
Karttunnen, L. What do Referential Indices Refer To? (RAND Report P-3854). SantaMonica CA: RAND Corporation, 1968.
Kartunnen, L. Pronouns and Variables In R. Binnick et al. (Eds.), Papers from the FifthRegional Meeting of the Chicago Linguistics Society, Chicago, IL, 1969.
Karttunnen, L. Discourse Referents. In J. McCawley (Ed.) Syntax and Semantics (Vol.7). New York: Academic Press, 1976.
Katz, J. J. and P. M. Postal. An Integrated Theory of Linguistic Descriptions. Cambridge,Mass., 1964.
Keenan, Edward L., Two Kinds of Presupposition in Natural Language, Fillmore & Langen-doen, Studies in Linguistic Semantics, New York: Holt, Rinehart, & Winston, 1971
Kellogg, C. H. A Natural Language Compiler for On-Line Data Management. Proc AFIPS,1968.
Kieras, D. E. Good and Bad Stncture in Simple Paragraphs: Effects on Apparent Theme,Reading Time and Recall. Journal of Verbal Learning and Verbal Behavior, 17,(1), Febru-ary 1978, 13-28.
Kintsch, Walter and Teun A. van Dijk. Toward a Model of Text Comprehension and Pro-duction. Psychological Review, Vol. 85, No 5, September 1978.
Kiparski, P. & Kiparsky, C Fact. Steinberg and Jacobovits 1971 pp 345-369
Kiparsky, Paul & Carol Kiparsky Fact, Steinberg & Jacobovits, Semantics, New York:Cambridge University Press, 1971
Kirby, K. Structural Ambiguity and Sentence Processing. Paper presented at theLanguage and Speech Conference, University of Melbourne, November 1977.
Kirby, K. Semantic Ambiguity and Sentence Processing. MSc thesis, Department ofPsychology, University of Melbourne, in preparation.
Klappholz, A. & Lockman, A. Contextual Reference Resolution: American Journal of Com-putational Linguistics, 1975, 4.
Klappholz, A. D. & Lockman, A. D.. The Use of Dynamically Extracted Context for Ana-phoric Reference Resolution. Unpublished MS, Department of Electrical Engineering andComputer Science, Columbia University, New York, February 1977.
Klahr, Philip The Deductive Pathfinder System Development Corp., SP-3842 1975
Klingbiel, P. H. Machine-Aided Indexing for Technical Literature. Information Storage andRetrieval, 9, 1973, 79-84.
Klingbiel, P.H. A Technique for Machine-Aided Indexing, Information Storage and Retrieval,9, 1973, 477-494.
Kochen, M. Prinriples of Information Retrieval. Melville Publishing Company, Lrcs Acgelet,1974.
Kowalski, R.A. Logic for Problem Solving. DCL Memo 75, Dept of Al, Edinburgh, March,1974.
5-11
Kucera, Henry, Nelson, Francis W. Computational Analysis of Present-Day AmericanEnglish Brown University Press Providence, Rhode Island 1970
Kuhns, J.L. Answering Questions by Computer: A Logical Study The Rand CorporationRM-5424-PR December 1967
Kuhns, J.L. & Montgomery, C.A. Experiments in Information Correlation AFIPS Confer-ence Proceedings 1964 Spring Joint Computer Conference Vol. 26 Spartan Books, Inc.Baltimore, Maryland 1964 pp 557-586
Kuhns, J.L. Interrogating A Relational Data File: Remarks on the Admissibility of InputQueries The Rand Corporation, 511-PR November 1970 b
Kuhns, J.L. Logical Aspects of Question-Answering by Computer in Software EngineeringVol. 2 Julius Tou (ed) Academic Press February 1970 a
Kuhns, J.L. Quantification in Query Systems Proc. of the ACM Symposium on InformationStorage and Retrieval University of Maryland April 1971
Kuhns, J.L. and C.A. Montgomery, Synthesis of Inference Techniques, Event RecordSpecification System Concept: Preliminary Notions, OSI Technical Report No. 1, R73-008(Contract No. F30602-73-C-0333), Operating Systems, Inc., 17 October 1973.
Kuhns, J.L, Synthesis of Inference Techniques: An Interpreted Syntax for the LogicalDescription of Events, OSI Technical Note No. 2, N74-003 (Contract No. F30602-73-C-0333), Operating Systems, Inc., 31 May 1974.
Kuhns, J.L., C.A. Montgomery and D.K. Welchel, ERGO -- A System for Event Record Gen-eration and Organization, RADC-TR-75-61, March 1975, B003215L.
Kuipers, B.J. Representing Knowledge for Recognition Bobrow and Collins 1975
Kuno, S. Some Properties of Non-Referential Noun Phrases. In R. Jakobson and S.Kawamote (Eds.), Studies in General and Oriental Linguistics. Tokyo, Japan: TEC Com-pany Ltd., 1970.
Kuno, S. Three Perspectives in the Functional Approach to Syntax. In: Robin E. Gross-man, L. J. San and T. J. Vance (Eds.). Papers from the Parasession on Functionalism, Chi-cago Linguistic Society, 1975, 276-336.
Kuno, S. Subject, Theme and the Speaker's Empathy -- A Re-Examination of Relativiza-tion Phenomena. In C. Li (Ed.), Subject and Topic. New York: Academic Press, 1976.
Lakoff, G. Instrumental Adverbs and the Concept of Deep Structure. Foundations ofLanguage, 4, 1968, 4-29.
Lakoff, George Presupposition & Relative Well-Formedness, Steinberg & Jakobobits,Semantics, New York: Cambridge University Press, 1971
Lakoff, G. On Generative Semantics. (1) In: Danny D. Steinberg and Leon A. Jakobovits(Eds.), (1971). Semantics: An Interdisciplinary Reader In Philosophy, Linguistics andPsychology. Cambridge University Press, 1971, 232-296. (2) An earlier versionj appeared in: Papers from the Fifth Regional Meeting, Chicago Linguistic Society, 1969.
Lakoff, G. Pronouns and Reference. (1) in: McCawley 1976, 276-335. (2) Also pub-lshed as: Mimeo, Bloomington, Indiana: Indiana University Linguistics Club, 1968.
Lakoff, G. Counterparts, or the Problem of Reference In Transformational Grammiar. InHarvard Computation Laboratory Report NSF-24, 1970, 23-36.
6-12
Lakoff, G. & Ross, J. R. A Note on Anaphoric Islands and Causatives. Linguistic Inquiry,1972, 3(1), 121-127.
Langacker, R. Pronominalization and the Chain of Command. In D. Reibe' & S. Schane(Eds.), Modern Studies in English. New Jersey: Prentice-Hall, 1966.
Landsbergen, S.J. Syntax and Formal Semantics of English in PHLIQA1. Preprints of theInternational Conference on Computational Linguistics (COLING-76), 1976, Ottawa,Canada.
Leavenworth, B.M. and Sammet, J.E. An Overview of Nonprocedural Languages I BM
Technical Report RC4685 January 1974
Lehnert, W. Questions Answering in a Story Understand System Cognitive Science Vol.1, No. 1 January 1977
Lees, R. B., The Grammar of English Nominalizations, IJAL Publication 12, Indiana Univer-sity, Bloomington; and Mouton and Co., The Hague, 1960.
Lehrer, A. Talking About Wine. Language, 51(4), December 1975, 901-923.
Levin, J. A. Proteus: An Activation Activaticn Framework for Cognitive Process Models(Working Paper WP-2). Marina del Ray, CA: Information Sciences Institute, 1976.
Levin, S. Problem Selection in Software Design TR No 93 Department of Informatin andComputer Science University of California November 1976
LI, C. N. (Ed.) Subject and Topic. New York: Academic Press, 1975. (Some errata tothis work were published in: Lingua, 43(1), October 1977, page 97.)
Linsky, L. (1) Reference and Referents. In: Caton 1963, 74-89. (2) Part of this workwas published as: Hesperus and Phosphorous. Philosophical Review, LXVIII, 1959, 515-518.
Litechky, C.R. and Davis, G.B. A Study of Errors, Error-Proneness and Error-Diagnosis InCOBOL Communications of the ACM 1976, 19 (1):33-37
Lockman, A. D. Contextual Reference Resolution (1) PhD Dissertation, Faculty of PureScience, Columbia University, May 1978. (2) Technical Report DCS-TR-70, Departmentof Computer Science, Rutgers University, 1978.
Lucas, H.C. and Kaplan, R.B. A Structured Programming Experiment The Computer Jour-nal 1975, 19(2): 136-138Lyons, J. Introduction to Theoretical Linguistics Cambridge: Cambridge University Press
1968
Lyons, J. Semantics. Cambridge University Press, London, 1977.
Mackworth, A. K. Vision Research Strategy: Black Magic, Metaphors, Mechanisms,Mlniworlds and Maps. In: A. R. Hanson and E. M. Riseman (Eds.). Computer V;sion Sys-tems. New York: Academic Press, 1978.
McCalla, G. I. An Approach to the Organization of Knowledge for the Modelling of Conver-sation. (1) PhD thesis, Department of Computer Science, University of Britl.h Columbia,June 1977. (2) Abridged version published as: Technical Report 78-4, Department ofComputer Science, University of British Columbia, February 1978.
5-13
McCawley, J. D. Lexical Insertion in a Transformational Grammar Without Deep Structure.Papers from the Fourth Regional Meeting, Chicago Linguistic Society, April 1968, 71-80.
McCawley, J. D. (Ed.) Syntax and Semantics 7: Notes from the Linguistic Underground.New York: Academic Press, 1976.
McDonald, D. D. A Simultaniously (sic) Procedural and Declarative Data Structure and itsuse in Natural Language Generation. Proceedings of the Second National Conference,Canadian Society for Computational Studies of Intelligence/Societe Canadienne desetudes d'intelligence par ordinateur. Toronto, July 1978, 38-47.
McDonald, D. D. Subsequent reference: Syntactic and Rhetorical Constraints. In: Waltz1978, 64-72.
McDonald, N. and Stonebraker Cupid - The Friendly Query Language Memorandum No.ERL-M487 Electronics Research Laboratory UC Berkeley 1974
Maratsos, M. The Effects of Stress on the Understanding of Pronominal Co-reference inEnglish. Journal of Psycholinguistic Research, 1973, 2(1), 1-8.
Marcus, Mitchell A Design for a Parser for English Proceedings of the Annual Conferenceof the Association for Computing Machinery Houston, Texas October, 1976
Martin, James Design of Man-Computer Dialogues Englewood Cliffs, N.J.: Prentice-Hall,Inc. 1973
Masterman, M. The Thesaurus In Syntax and Semantics Mechanical Translation 1966
Meehan, J.R. Using Planning Structures to Generate Stories Americal Journal of Compu-tational Linguistics in press
Miller, G. Semantic Relations among Words, in M. Halle, J. Bresnan, and G. Miller, eds.,Linguistic Theory and Psychological Reality. Cambridge, MIT Press, 1978.
Miller, G. and P. Johnson-Laird. Language and Perception. Cambridge, Harvard UniversityPress, 1976.
Miller, G.A. English Verbs of Motion: A Case Study in Semantics and Lexical Memory A.W. Melton and E. Martin (eds) Coding Processes In Human Memory Washington, D.C.Winston
Miller, L.A. Programming by Nonprogrammers International Journal of Man-Machine Stu-dies, 1974 6:237-260
Minker, J. information Storage and Retrieval A Survey and Functional DescriptionUniversity of Maryland TR 369 April 1975
Minsky, Marvin Minsky's Frame System Theory Theoretical Issues In Natural LanguageProcessing R. Schnak and P.L. Nash-Webber (eds) 1976 pp 104-116
Minsky, M. L. (Ed.) Semantic Information Processing. Cambridge, Massachusetts: MIT'4 Press, 1968.
Minsky, M. L. A Framework for Representing Knowledge. (1) In: P. H. Winston (Ed). ThePsychology of Computer Vision. McGraw-Hill, 1975, 211-280. (2) Memo 306, ArtificialIntelligence Laborartory, Massachusetts Institute of Technology, June 1974. (3) A con-densed version appears In: Schank and Nash-Webber 1975, 118-130. (4) Version 3also appears as: Frame-System Theory. In: P. Nicholas Johnson-Laird and P. C. Wason(Eds.). Thinking: Readings In Cognitive Science. Cambridge University Press, 1977,
6-14
856-376.
Montague, R. The Proper Treatment of Quantification In Ordinary English Approaches toNatural Language J. Hintakka, J. Moravcslk and P. Suppes (eds) Dordrecht 1973
Montgomery, C.A. Linguistics and Information Science J. ASIS, Vol. 23 Journal of theAmerican Society for Information Science No 3 pp 195-219 1972
Montgomery, C.A. and Katter, R. V. On-Line Bugging: Hope for Terminal Cases of Seman-tic Deviance invited Paper for the Gordon Conference on Scientific Information ProblemsIn Research July 1972
Moore, J. and Newell, Allen How Can MERLIN Understand? Knowledge and CognitionGregg (ed) Lawrence Eribaum Associates 1973
Morgan J. L. Some Strange Aspects of It. Papers from the Fourth Regional Meeting, Chi-cago Linguistic Society, April 1968, 81-93.
Morgan, J. L. Towards a Rational Model of Discourse Comprehension. In: Waltz 1978,109-114.
Morgan, J. Pragmatics. In R. Spiro, B. Bruce & W. Brewer (Eds.), Theoretical Issues inReading Comprehension. New Jersey: Lawrence Erlbaum Associates, in press.
Nash-Webber, B. L. Semantic Interpretation Revisited. Report 3335 (Al report 48), BoltBeranek and Newman Inc., Cambridge, MasMsachusetts, July 1976.
Nash-Webber, B. L. Anaphora: A Cross-Disciplinary Survey. Technical Report CSR-31,Center for the Study of Reading, University of Illinois at Urbana-Champaign, April 1977.
Nash-Webber, B. L. and Reiter, R. Anaphora and Logical Form: On Formal MeaningRepresentations for Natural Language. (1) Proceedings of the Fifth International JointConference on Artificial Intelligence. Cambridge, Massachusetts, August 1977, 121-131. (2) Technical Report CSR-36, Center for the Study of Reading, University of Illinoisat Urbana-Champaign, and Bolt Beranek & Newman, Inc., 1977.
Nash-Webber, B. L. & Sag, I. A. Under Whose Control? Linguistic Inquiry, 9 (1), Winter1978, 138-141.
Nash-Webber, B. L. A Catalogue of Pronominal Referents. (Technical Report) Center forThe Study of Reading, U. of Illinois and Bolt Beranek & Newman Inc., forthcoming.
Nawrocki, L.H., Strub, M.H. and Cecil, R.M. Error Categorization and Analysis in Man-Computer Communication Systems IEEE Transactions on Reliability 1972, R-22:135-140
Nawrocki, L.H. Word Abbreviations in Man-Computer Communications System U.S. ArmyResearch Institute (Arlington, Virginia) Unpublished Manuscript 1972
Nelson, R. The First Literate Computers? Psychology Today, 11(10), March 1978, 72-80.Newell, A. and Simon, H. Human Problem Solving Englewood Cliffs, N.J. Prentice-Hall
1972
Norman, D. & Rumelhart, D. Explorations in Cognition, San Francisco: W. H. Freeman andCo., 1975.
Olney, J.C. and Londe, D.L. An Analysis of English Discourse Structure with ParticularAttention to Anaphoic Relationships ACTES de x e Congres International des Linguistes
5-15
Bucharest 1967
Operating Systems, Inc. An Introspective Data Base for an Active Information Model OSITechnical Note N76-017 17 November 1976Operating Systems, Inc. Synthesis of Inference Techniques RADC-TR-73-219 FinalReport Contract F30602-72-C-0441 August 1973
Operating Systems, Inc. Synthesis of Inference Techniques: An Interpreted Syntax forthe Logical Description of Events OSI Technical Note No. 2, N74-003 Contract No.F30602-73-0333 31 May 1974
Operating Systems, Inc. Synthesis of Inference Tehniques Final Technical Report:ERGO-A System for Event Record Generation and Organization RADC-TR-75-51 1975
Operating Systems, Inc. Synthesis of Inference Techniques: An Interpreted Syntax forthe Logical Description of Events OSI Technical Note No. 2, N74-003 Contract No.F30602-73-C-0333 31 May 1974
Operating Systems, Inc. Synthesis of Inference Techniques RADC-TR-73-21 9 FinalReport Contract F30602-72-C-0441 August 1973
Orne, M. T. On the Social Psychology of the Psychological Experiment: With ParticularReference to Demand Characteristics and their Implications. American Psychologist,17(11), November 1962, 776-783.
Ortony, A. Some Psycholinguistic Constraints on the Construction and Interpretation ofDefinite Descriptions. In: Waltz 1978, 73-78.
Pacak, M. G. and A. W. Pratt. The Function of Semantics In Automated Language Pro-cessing. National Institutes of Health, Bethesda, Md., Div. of Computer Research andTechnology, 1970.
Palek, B. Cross-Reference: A Study from Hyper- Syntax Travaux Linguistiques dePrague 3 1968
Partee, Barbara H. (ed) Montague Grammar Academic Press, Inc. 1976
Partee, B.H. Bound Variables and Other Anaphors. In: Waltz 1978, 79-85.
Partee, B.H. Deletion and Variable Binding. In E. Keenan (Ed.) Formal Semantics ofNatural Language. England: Cambridge University Press, 1975.
Partee, B.H. Opacity, Coreference and Pronouns. In Harmon and D. Davidson (Eds.)Semantics of Natural Language. The Netherlands: D. Reidel, 1972.
Perrault, C.R. and Cohen, P. R. Planning Speech Acts. Al Memo 77-1, Department ofComputer Science, University of Toronto, June 1977.
Pereira, L.M., F.C.N. Pereira and David H.D. Warren. User's Guide to DECsystem-10 Prolog.Provisional Version, April, 1978.
Petofi, J.S. On the Problems of Co-Textual Analysis of Texts, Preprint No. 50. Interna-tional Conference on Computational Linguistics, Sanga-Saby, Sweden, 1969
Petofi, Janos S. Generativity and Text-Grammar Folia Linguistica, Acta SocletatisLlnguisticae Europaeae, Tomus V, 3/4 Mouton 1971
Petofi, J.S. The Syntactic-Semantic Organization of Text-Structures in Poetics #3,Mouton, 1972.
5-16
Petofi, J.S. Transformationsgrammatiken und die grammatische Beschrelbung der texte,Linguistische Berlchte 14 1971a idem, Transformationsgrammatiken und eine Kotex-tuelle Texttheorle. Grundfragen und Konzeptionen (Frankfurt am Main, 1-971b Idem,"Towards a grammatical theory of verbal texts, Working Paper, Gotherburg University;German version to appear in Zeitschrift fur Llteraturwissenschaft und Linguistik 1971
Petofi, J. S. and H. Rieser. Studies In Text Grammar. Dordrecht: Reidel, 1973.
Petrick, S.R. On Natural Language Basefd Computer Systems. (1) IBM Journal ofResearch and Development, 20(4), July 1976, 314-325. (2) in: Zampolli 1977, 313-340.
Phillips, B. Discourse Connectives. Technical Report KSL-1 1, Knowledge SystemsLaboratory, Department of Information Engineering, University of Illinois at Chicago Circle,March 1977.
Phillips, B. Judging the Coherency of Discourse. In: Diller 1975, Microfiche 35, 36-49.
Pike, K. L. Language in Relation to a Unified Theory of Human Behavior The Hague Mou-ton 1967
Postal, P. On so-called 'Pronouns' In English. In D. Reigel & S. Schane (Eds.) ModernStudies In English. New Jersey: Prentice-Hall, 1966.
Postal, P. M. Anaphoric Islands. Papers from the Fifth Regional Meeting, ChicagoLinguistic Society, April 1969, 205-239.
Postal, Paul M. Linguistic Anarchy Notes. In: McCawley 1976, 201-225.
Quick, John Dictionary of Weapons & Military Terms McGraw-Hill Book Co. NY 1973
Quillan, M.R. Semantic Memory Semantic Information Processing Minsky, M.L. (ed) MITPress Cambridge, Mass. 1968
Quillan, M.R. The Teachable Language Comprehender A simulation Program and Theory ofLanguage CACM 12 (8) pp 459-476 1969
Quirk, R., S. Greenbaum, G. Leech and J. Svartvilei A Grammar of Contemporary English,Longmann, 1972.
Relbel, D.A. & Schane, S.A. Modern Studies in English: Readings In TransformationalGrammar. Englewood Cliffs, New Jersey: Prentice-Hall, 1969.
Reichenbach, H. Elements of Symbolic Logic MacMillan Co. NY, 1947
Reichenbach, H. Experience and Prediction University of Chicago Press Chicago 1938
Relchman, Rachel Conversational Coherency. Cognitive Science, 2(4), October-December 1978, 283-327.
Reichman, R. Conversational Coherency. Technical Report TR-17-78, (Department of?), Harvard University, 1978.
Reinhart, T. The Syntactic Domain of Anaphora. Unpublished doctoral dissertation,Department of Foreign Literatures and Linguistics, MIT, 1976.
Relsner, Phyllis Use of Psychological Experimentation as an Aid to Development of aQuery Language IBM Technical Report RJ1 707 (26140) IBM Research Laboratory SanJose, CA January 1976
6-17
.1i
Reiter, Raymond Formal Reasoning and Language Understanding Systems TINLP 1975
Rieger, C. J. III Conceptual Memory and Inference. In: Schank 1975, 157-288.
Rieger, C. J. Conceptual Memory, Conceptual Information Processing R. C. SchankAmsterdam: North-Holland Publishing Company 1975
Rieger, C. The Commonplace Algorithm as a Basis for Computer models of Human Memory,Inference Belief and Contextual Language Understanding. Proc. Conf. on TheoreticalIssues in Natural Language Processsing Cambridge, MA 1975
Rieger, C. J. III The Importance of Multiple Choice. (1) In: Watz 1978, supplement. (2)Technical Report 656, Department of Computer Science, University of Maryland, 1978.
Riesbeck, C.K. Conceptual Analysis Conceptual Information Processing R.C. SchnakAmsterdam: North-Holland Publishing Co. 1975
Roberts, R. B. & Goldstein, I. P. The FRL Primer. Memo 408, Artificial IntelligenceLaboratory, Massachusetts Institute of Technology, September 1977.
Robinson, J.A. A Machine-Oriented Logic Based on the Resolution Principle. Journal ofthe ACM, 12, 1965.
Rosenberg, Richard S. Artificial Intelligence and Linguistics Proceedings of the FirstAnnual Meeting of the Berkeley Linguistics Society Cathy Cogen, Henry Thompson, Gra-ham Thurgood, Kenneth Whistler and James Wright (eds) Berkeley, CA: University of Cali-fornia at Berkeley 1975 pp 379-392
Rosenberg, S. T. Discourse Structure. Working Paper 130, Artificial IntelligenceLaboratory, Massachusetts Institute of Technology, November 1977
Ross, J. R. On the Cyclic Nature of English Pronominalization. (1) In: Reibel and Schane1969, 187-200. (2) In: To Honor Roman Jakobson. The Hague: Mouton, 1967, Volume II,1669-1682.
Ross, J. R. The Superficial Nature of Anaphoric Islands. Linguistic Inquiry, 1971, 11(4),599-600.
Rouse, W.B. Design of Man-Computer Interfaces for on-line Interactive System Proceed-Ings of the IEEE 1975, 63 847-857
Roussel, P. Prolog: Manuel de Reference et d'Utilisation. Groupe d'lntelligence Artificielle,Marseille-Luminy, 1976.
Russell, B. On Order In Time, reprinted in Logic and Knowlege (R. C. Marsh, ed.), GeorgeAllen and Unwin, London, 1956.
Rubin, A. The Relation Between Comprehension Processes in Oral and Written Language.In R. Spiro, B. Bruce & W. Brewer (Eds.) Theoretical Issues In Reading Comprehension,New Jersey: Lawrence Erlbaum Associates, in press.
Rumelhart, D.E. Notes on a Schema for Stories Representation and Understanding: Stu-dies In Cognitive Science D. Bobrow and A. Collins (eds) New York: Academic Press1975
Rumelhart, D.E., and Norman, D.A. Active Semantic Networks as a Model of HumanMemory, Advance Papers, International Joint Conference on Artificial Intelligence, Stan-ford, CA 20-23 August, 1973 Stanford Research Institute, Menlo Park, CA 1973 pp450-457
6-18Olt--
Rumelhart, David E., and Norman, Donald A., Active Semantic Networds as a Model ofHuman Memory, Advance Papers, International Joint Conference on Artificial Intelligence,Stanford, CA 20-23 August 1973. Stanford Research Institute, Menlo Park, CA 1973,450-457
Rumelhart, David E., Understanding and Summarizing Brief Stories Center for Human Infor-mation Processing CHIP 58, University of California San Diego, April 1976.
Russell, B. On Denoting. (1) Mind, XIV (number 56), October 1905, 479-493. (2) In:Bertrand Russell. Logic and Knowledge: Essays 1901-1950 (R. C. Marsh Editor) London:Allen and Unwin, 1956. (3) In: Herbert Feigi and Wilfrid Sellars (Eds). Readings in Philo-sophical Analysis. New York: Appleton-Century-Crofts, 1949.
Sag, I. Deletion and Logical Form. Unpublished doctoral dissertation, MIT Department ofForeign Literatures and Linguistics, 1976.
Sager, Naomi, The String Parser for Scientific Literature, Natural Language Processing,Rustin, R., (ed) Algorithmic Press, NY 1973
Sager, ;daomi, Syntactic Analysis of Natural Language Advance in Computers, No 8 F. Altand M. Rubinoff (ed) Academic Press, NY 1967
Sammet, Jean E., Programming Languages: History and Fundamentals, Prentiss-Hall Inc.,Englewood Cliffs, NJ 1969
Scha, R. Semantic Types in PHLIQA1. Preprints of the International Conference on Com-putational Linguistis (COLIN-76), 1976, Ottawa, Canada.
Schachter, P. Does She or Doesn't She? Linguistic Inquiry, 8(4), Fall 1977, 763-767.
Schachter, Paul, Constraints on Coordination, Unpublished paper, Language, (in press)
Schank, R. C. Understanding Paragraphs Technical Report 5, Instituto per gli studiSemantlci e Cognitivi Castagnola, Switzerland 1973
Schenk, R. C. Using Knowledge to Understand, Proc. Conference on Theoretical Issues inNatural Language Processing, Cambridge, MA 1976
Schank, R. C. Identification of Concetualizations Underlying Natural Language. In:Schank and Colby 19i3, 187-247.
Schank, R. C. Conceptual Information Processing. Amsterdam: North-Holland, 1975.
Schank, R. C. The Structure of Episodes in Memory Representation and Understanding:Studies in cognitive Science D. Bobrow and A. Collins (eds) New York: Academic Press1975
Schank, R. C. and Abelson, R., Scripts, Plans, and Knowledge, The Proceedings of theFourth International Joint Conference on Artificial Intelligence Tblisi, USSR, 1975
Schenk, R. C. & Abelson, R. P. Scripts, Plans, Goals and Understanding: An Enquiry IntoHuman Knowledge Structures. Hillsdale, New Jersey: Lawrence Erlbaum Associates,1977.Schank, R. C. & Colby, K. M. Computer Models of Thought and Language. San Francisco:
W. H Freeman, 1973.
Schank, R. C.; Goldman, N. M.; Rieger, C. J. III & Riesbeck, C. K. Inference and Para-phrase by Computer. Journal of the Association for Computing Machinery, 22(3), July
5-19
1975, 309-328.
Schank, R. C. & Nash-Webber, B. L. Theoretical Issues in Natural Language Processing:An Inter-disciplinary Workshop. Cambridge, Massachusetts: Association for Computa-tional Linguistics, June 1975.
Schank, R. C. & the Yale Al Project. SAM -- A Story Understander. Research Report 43,Department of Computer Science, Yale University, 1975.
Schank, Roger C., Primitive Acts of Conceptual Dependency Theoretical Issues in NaturalLanguage Processing, Cambridge, MA 1975
Shapiro, Stuart C., A Net Structure for Semantic Information Storage, Conference onArtificial Intelligence, London, England, 1-2 September
Schmidt, Charles F. Understanding Human Action, TINLP, 1975
Schneiderman, B. and McKay, D. Experimental Investigations of Computer Debuggingand Modification Proceedings 6th Congress of the International Ergonomics Association,College Park, Maryland, July 1976, pp. 537-563
Scragg. Greg W., Answering Questions about Processes, Explorations in Cognition, DonaldA. Norman, David E. Rumelhart and the LNR Research Group, San Francisco, CA. W. H.Freeman and Commpany 1975 pp. 349-375
Sgall, P.; Hajicova, E. & Benesova, E. Topic, Focus and Generative Semantics.Kronberg Taunus: Scriptor Verlag, 1973.
Shanon, C. E. & Weaver, W. The Mathematical Theory of Communication. Urbana: Univer-sity of Illinois Press, 1949.
Shapiro, Stuart C., A Net Structure for Semantic Information Storage, Conference onArtificial Intelligence London, England, 1-2 September 1971 The. British ComputerSociety, London, England 1971 pp 512-523
Sidner, C. L. See also Bullwinkle, Candace L.
Sidner, C. L. A Progress Report on the Discourse and Reference Components of PAL. (1)In: Proceedings of the Second National Conference, Canadian Society for ComputationalStudies of Intelligence/Societe canadienne des etudes d'intelligence par ordinateur.Toronto. July 1978, 206-213. (2) Memo 468, Artificial Intelligence Laboratory, Mas-sachusetts Institute of Technology, 1978.
Sidner, C. L. The Use of Focus as a Tool for the Disambiguation of Definite Noun Phrases.In: Waltz 1978, 86-95.
Silva, G. and Montgomery, C. A., Automated I&W File Generation, RADC-TR-77-194, June,
1977.
Silva, G. and C.A. Montgomery, Knowledge Representation for Automated Understandingof Natural Language Discourse. Computers and the Humanities, Vol 11, Pergamon Press,1978.
Silva, G., D.L. Dwiggins, S.G. Busby, and J.L. Kuhns, A Knowledge-Based Automated Mes-sage Understanding Methodology for an Advanced Indications System. RADC-TR-79-133,Rome Air Development Center, Air Force Systems Command, Griffiss Air Force Base, NewYork 13441, June, 1979, A072395.
5-20
*1
Sime, M. E., Green, T.R.G., and Guest, D. J., Psychological Evaluation of Two ConditionalConstructions Used in Computer Languages International Journal of Man-Machine Studies1973 5: 105-113.
Simmons, R. F., Semantic Networks: Their Computation and Use for UnderstandingEnglish Sntences, Computer Models of Thought and Language, R. C. Schnak and K. M.Colby (eds) San Francisco: Freeman 1973.
Simmons, R. F., and Bennett-Novak, G., Semantically Analyzing an English Subset for theClowns Microworld, AJCL 18 1975.
Simmons, R. R. and Amsler, R. A., Modeling Dictionary Data, Directions in Artificial Intelli-gence, Natural Language Processing Courant Computer Science Report #7, August,1975.
Simmons, Robert F., Inferential Question Answering in a Textual Data Base, Proceedingsof the Annual Conference of the ACM, Houston, Texas 1976.
Simmons, Robert F., The Clowns Microworld, Mathmatical Social Sciences Board Sympo-sium, Cambridge, MA June 10-13 1975, Preprints Theoretical Issues in NaturalLanguage Processing R. Schank and R. L. Nash-Webber, (eds) 1975b pp 17-19
Simmons, Robert F., Bennett-Novak, Gordon, Semantically Analyzing an English Subset forthe Clowns Microworld, American Journal of Computational Linguistics Microfiche 197518 1916a pp 67.
Smith, S. L., An on-line model of traffic control In a Communication Network TechnicalReport MTR-2813, MITRF Corporation - Bedford, MA March 1974
Sondhelmer, N. K. Towards a Combined Representation for Spatial and TemporalKnowledge. Proceedings of th Fifth International Joint Conference on Artificial Intelli-gence. Cambridge, Massaphusetts, August 1977, 218-282.
Sondhelmer, N. K. Spatial Reference and Semantic Nets. American Journal of Computa-tional Linguistic, Microfiche 71, 1977.
Spencer, N. J. Differences Between Linguists and Non-linguists in Intuitions ofGrammaticality-Acceptability. Journal of Psycholinguistic Research, 2(2), April 1973,83-98.
Springston, F. Verb-derived Constraints In the Comprehension of Anaphoric pronouns.Paper presented at the Eastern Psychological Association, 1976.
Steinberg, D. and Jakobobits, L., Semantics: An Interdisciplinary Reader in Philosophy,Linguistics, and Psychology, New York: Cambridge University Press, 1971.
Strawson, P. F. On referring. (1) Mind, LIX (number 235), July 1950, 320-344. (2) In-Caton 1963. (3) In: Antony Flew (Ed.). Essays on Conceptual Analysis. London: MacMil-Ian, 1956. 21-52.
Strub, M. H., Evaluation of Man-Computer Input Techniques for Military Information Sys-tems, Technical Research Note 226, U. S. Army Research Institute, Arlington, VA May1971.
Strub, M. H., Automated Aids to On-Line Tactical Data Inputting Technical Paper 262, U.S. Army Research Institute (Arlington, VA) February 1975.
5-21
Taylor, B. H. A Case-driver Parser. Unpublished MSc thesis, Department of ComputerScience, University of British Columbia, May 1975.
Taylor, B. H. & Rosenberg, R. S. A Case-driven Parser for Natural Language. (1) Ameri-can Journal of Computational Linguistics, Microfiche 31, 1975. (2) Also published as:Technical Report 75-5, Department of Computer Science, University of British Columbia,October 1975.
Thomas, A. L. Ellipsis: The Interplay of Sentence Structure and Context. Lingua, 47(1),January 1979, 43-68.
Thomas, J. C., Quantifiers and Question-Asking. IBM Thomas J. Watson Research Center,RC5866, February 18, 1976.
Thomas, J. C. and Gould, J. D., A Psychological Study of Query by Example, Proceedingsof the National Computer Conference, 1 975, pp. 439-445.
Thompson, F. B., Lockemann, P. C., Dostert, B., & Deverill, R. S. REL: Rapidly ExtensibleLanguage System Proceedings of the 24th National Conference ACM, September 1969p. 69.
Triggs, T. J., The Role of Protocol Analysis in Understanding Human Cognitive Processes,Bolt, Beranek and Newman (Cambridge, MA) Report No. 2616, Contract DAHC-19-72-C-0019, Department of Defense, July 1973a.
Van Diijk, lhwe, Petofi, J. S. and Rieser, H., Textgrammatische Grundlagen fur eineTheorie Narrativer Strukturen Linguistiche Berichte 16, 1971, Another detailed version ofthis paper is Zur Bestimmung narrativer Strukturen auf der Grundlage von Testgramma-tiken Papeire zue Testlinguistik/ Papers in Textlinguistics 1 1972.
Van Dijk, Teun A., Petofi, J., Janos, S., and Rieser, H., Two Text Grammatical Models,Foundations of Language, B 1972 pp. 499-545
Van Dijk, Teun, A. Some Aspects of Text Grammars, The Hague: Mouten, 1972.
Van Dijk, T., Recalling and Summarizing Complex Discourse, Mineo, 1975.
Van Dijk, Teun A., Complex Semantic Information Processing Manuscript, University ofAmsterdam, 1 976.
Van Dijk, Teun A. Text and Context, Longman, London, 1 977.
Van Dijk, Teun A. and J. S. Petofi (eds). Grammars and Descriptions. Berlin/New York: deGryuter, 1977.
Van Dijk, T. From Text to Interdisciplinary Discourse Studies. Paper prepared for the LaJolla Conference on Cognitive Science, University of California at San Diego, La Jolla,August 13-16, 1979.
van Emden, M.H. Programming with Resolution Logic. Report CS-75-30, Dept. of Com-puter Science, University of Waterloo, Canada, Nov, 1975.
Walker, D E (Ed.) Speech Understanding Research. SRI Project 4762, Final technicalreport, SRI International, October 1976.
Walker, D. E. (Ed.). Understanding Spoken Language. (The Computer Science Library,Artificial Intelligence Series 5), New York: North-Holland, 1978.
5-22
Walker, Donald R., Automated Language Processing, Annual Review of Informatin Scienceand Technology, Volume 8, Caudra, Carlos A., Luke, Ann W., (eds), Washington, D. C.,American Society for Information Science, 1973 pp. 67-1 19.
Waltz, D. L. (Ed.). TINLAP-2: Theoretical Issues in Natural Language Processing-2.University of Illinois at Urbana-Champaign, 25-27 July 1978.
Warren D. H. D. Implementing Prolog- Compiling Predicate Logic Programs. Dept. of AlResearch Reports 39 & 40, Edinburgh, May, 1977(a).Warren D.H.D. Logic Programming and Compiler Writing. Dept. of Al Research Report 44,
Edinburgh, September, 1977(b).
Warren, D.H.D., Pereira, L.M. and Pereira, F.C.N. Prolog- The Language and its Implementa-tion Compared with Lisp. Procs. of the ACM Symposium on Al and ProgrammingLanguages, SIGPLAN/SIGART Newsletter, Rochester NY, Aug 1977.
Wasow, T. A. Anaphoric Pronouns and Bound Variables. Language, 51(2), June 1975,368-383.
Wasow, T. Problems with Pronouns in Transformational Grammar. Unpublished ms.,Department of Linguistics, Stanford University, 1976.
Watt, W. C. Habitability. American Documentation (Now Journal of the American Societyfor Information Science), 19(3), July 1968, 338-351.
Watt, William C. Late Lexicalizations. On: K. Jaako; J. Hintikka; Julius M. E.; Moravcsikand Patrick Suppes (Eds.). Approaches to Natural Language: Proceedings of the 1970Stanford Workshop on Grammar and Semantics. Dordrecht: Reidel, 1973.
Watt, W. C. The Indiscreteness With Which Impenetrables Are Penetrated. Lingua, 17,1975, 95-128.
Webber, Bonnie Lynn. See also Nash-Webber, B. L.
Webber, B. L. A Formal Approach to Discourse Anaphora. (1) Report 3761, B. Beranek &Newman Inc., May 1978. (2) PhD thesis, Harvard University, 1978.
Webber, B. L. Description Formation and Discourse Model Synthesis. In: Waltz 1978,42-50.
Weizenbaum, J. Computer Power and Human Reason: From Judgment to Calculation. SanFrancisco: W. H Freeman, 1976.
Weischedel, Ralph M., Computation of a Subclass of Inferences: Presupposition andEntailment, Unpublished Doctoral Dissertation, University of Pennsylvania, 1975.
Welchel, D. K., Synthesis of Inference Techniques, Basic System Design for EventRecord Generation and Organization (ERGO) System, OSI Technical Report No. 2, R74-
003 Operating Systems, Inc., 17 January 1974.
Whitley, M. Stanley Person and Number in the Use of We, You, and They. AmericanSpeech,3(), Spring 1978, 18-39.
Wilensky, Robert, Using Plans to Understand Natural Language Proceedings of the Annua!Conference of the ACM, Houston, Texas 1976.
Wilks, Y. A. Decidability and Natural Language. Mind, LXXX (number 320), October 1971,497-520.
5-23
Wilks, Y. A. An Artificial Intelligence Approach to Machine Translation. In: Schank andColby 1973, 114-151.
Wilks, Y. A. Preference Semantics. (1) Stanford Artificial Intelligence Laboratory MemoAIM-206/Stanford University Computer Science Report CS-337, July 1973. (2) Alsopublished in: Edward Louis Keenan III (Ed.). Formal Semantics of Natural Language. Cam-bridge University Press, 1975, 329-348.
Wilks, Y. A. An Intelligent Analyzer and Understander of English. Communications of theACM, 18(5), May 1975, 264-274.
Wilks, Y., A Preferential, Pattern-seeking Semantics for Natural Language, Artificial Intel-ligence, 6, 1975.
Wilks, Y. A. Methodology in Al and Natural Language. (1) In: Schank and Nash-Webber1975, 144-147. (2) A revised version appears in: Yorick Alexander Wilks. Seventheses on artificial intelligence and natural language. Working paper 17, Institute forSemantic and Cognitive Studies, Geneva, 1975, 5-12. (3) Another revised versionappears as part of: Yorick Alexander Wilks. Methodological Questions About ArtificialIntelligence: Approaches to Understanding Natural Language. Journal of Pragmatics,1(1), April 1977, 69-84.
Wilks, Y., De Minimis, or the Archeology of Frames, Proc A.I.S.B. Edinburgh 1976.
Wilks, Yorick, Frames, Scripts, Stories, and Fantasies, Preprint, Coling, 1976.
Wilks, Yorick, Natural Language Inference, Stanford Artificial Intelligence LaboratoryMemo AIM-21 1, April, 1973.
Wilks, Yorick, Natural Language Understanding within the Al Paradigm, Stanford, CA:Stanford University Stanford Artificial Intelligence Laboratory, AIM-237, 1974, Decemberpp 40.
Wilks, Yorick. Processing Case, AJCL, Microfiche 56, 1976.
Wilks, Yorik. Methodological Questions about Artificial Intelligence Approaches to Under-standing Natural Language. Journal of Pragmatics 1, North-Holland Publishing Company,1977.
Winograd, T., Frame Representations and the Declarative Procedural Controversy,Bobrow and Collins (eds), 1975.
Winograd, T. Procedures as a representation of data in a computer program for under-standing natural language. Technical Report 17, Artificial Intelligence Laboratory, Mas-sachusetts Institute of Technology, February 1971.
Winograd, T. Understanding Natural Language. (1) New York: Academic Press, 1972.(2) Edinburgh University Press, 1972. (3)Also published in: Cognitive Psychology, 3(1),1972, 1-191.
Winograd, Terry, Five Lectures on Artificial Intelligence, Stanford, CA: Stanford Univer-sity Department of Computer Science, (STAN-CS-74-459) September 1974.
Winston, P. H. (1) Artificial Intelligence. Reading, Massachusetts: Addisnn-Wesley,1977. (2) An earlier version was circulated as: Notes on Computer Intelligence, May1976.
5-24
Of
Woods, W. A. Procedural Semantics for a Question-answering Machinc. AF7PS Confer-ence Proceedings, 33, FJCC 1968, 457-471.
Woods, W. A. Transition Network Grammars for Natural Language Anclyo . Communica-tions of the ACM, 13(10), October 1970, 591-606.
Woods, W.A. Progress in Natural Language Understanding: An Applicaticn tc Lunar Gol-ogy. AFIPS Conference Proceedings, Vol. 42, National Computer Conferencc_, 197,3.
Woods, W. A. Lunar Rocks in Natural English: Explorations in Natural Lar_1:: e,, Ques'ionAnswering. In: Zampolli 1977, 521-569.
Woods, W. A., Kaplan, R. M. & Nash-Webber, B. L. The Lunar Sciences 1Jt:::J LanguageInformation System: Final Report. BBN Report 2378, Bolt Beranek and New;,- -n In.c;., Cam-bridge, MA, 1972 (NTIS No. N72-23155).
Woods, William A., What's in a Link: Foundations for Semantic Networds Rcpresentatior,and Understanding Daniel G. Bobrow, Allan Collins (eds) New York, New Ycrk AcademicPress, Inc. 1975
Zampolli, A. Linguistic Structures Processing. Fundamental Studies in Computer Science5. Amsterdam: Noi-th-Holland, 1977.
5-26
APPENDIX A: Sample of Sanitized
Message Text
A-1
znJ XJ~ U-. rc u Ix
zi> EE I 0~ C L
m C OZU. IxCLU N'~- <1 C wZ
rX F ~LJ4L. MJLLJLJi 0 z
0cI- w --4-. E: >-~f u-0 ~ E U) xWzL =1 rJ~ .jm X_. z (0 >(-z
0.z j E ffl Z,,- ~ W.~ El X U ..4.- : zJ ~-C LJ xXE LL J)0
Vn -J4 Z r~ .. j-I - N U.1 Z C rc I M
x a c a: LL.J UE E Li0 IL! Z z >_-N Z ZC E ZL! C q, c'4U.J<C
C Up-z z= Z LO z 3~ X: cu.XC _j Z"CL! cH .C 0 --!z Z~ ZO L
Z _ E -i c w ~ JiIW Wi L) xz '- "xi C 0u c r _ L ClIC C ZL 0i
IT4~ i_' _j z DC u CZa- Ji n 0 X I- Lii Cj ZCL- L o z c_- WC :U~-_
ce ci -u ZCm w k JLWC -X ZLid (n C. z Um x cL CC 7- :rc . z '- DL! IZC U3rmCzU <-: CL~f L- iiL X .D C i z DX z_ E uJ _ I .
I- _ r_ W 0.. <~ii- ~.Z E i ... Z 0
X:4C- 1 c 0 CE.- Z f- :3 CL WCED ( E L L ZcC C4 xiiw0 * L C _jL. L aL A1I
0 Z 0 L.C W. 0 .e LI ; - :r r-LLJ fl L-- W Z !Z 19
4-CC >- M C i- a4 . C x r C-i XU 4- C <r CL!a
N. z 0 CC N_:w U=Ud ZLiMU ITU T Q L)X Ii A-2
w x Lii -C w
C Z
C Id 0M >- En U) LZz C C- -T4
I-. i- 3 CLw wLL I- w 00 Wn 174CL
* C '4 **I. p 4 >-0 cl: F- m C4: cc z 2:4 )
0t wi m 0 2: x w z rm->- -CDxz 0 z-j 0x w 0 L
aZc Id i-x ci En Ixi WLd F-( :x wc: z U) C U) I U) 40% " - E
OE LL. C <Ew > <I * (n 17~4 LIxi I-- ZI Q~ Z I W) 4 1= in -i c
C inU pI- w C E w 0 * wC4 1-4- jIxwC a_ a w - 1-4 MO E *L u xcc: iX 0 x 1-4 WL-m >- I- z 3 -j * c:C <E : Li U 0OIL J MIi-
-- -j -i>- w 0rdz X4 :3 2: J.- * >- Idi -C<1w-C 4x W - X .JC W ;EC 4 C wd-JC x UCi:2
M - *l...I Li Q i-i CL C ZL Z... C -i -T ME:I- L <W Id_ji -i C: U)'-" d 0 3 Z C-4JC IT -J 1-4 3 ZEXH-3N)S4 Q c Z E))C <1ZC C-j IC Id Ci:I Li- I
0 Z 174- W Cd m - U) I- 13 Li WdU)nZ9i F =0- L.)F- -TC zU) 2 z1- -4 M: C - - jCL -i4TU) I- 2: c ZO 0 Z EU lL 0 -faa- m:0 =) M
I M 1~i-4 W )144 - T .ix U) U) cn- N W 3Ud w - .JE OXU) 0LL U- Ix< 0 IdMEEr Z 0XC4~ i: * Idc:z -m-jw I M:-~ It-)- *
CC <1rZCL Z ZCc: 0 Id i U).J M EL) CL. C ww F- Ecci: E7Z IdOz w . Q4D ii z c >- m m = CU)nWC < CpuL 1-N X: -q C = x C(1-4 0- WZxi- ;E a:F-C<Li Q- Ma dC Ci:C W = DI W M En Eni' ) U) d3 0 0 dl-4.jCU) inJ d~ X2:- : 7 0 "-iZ ix ... i 0z I C:- -j
-j z~uI Z a zix a: D0 w :3 02Z Z LL .U 1-4EL~ ir- C F-<inw w U) w L U C zW Md W~U X ECC
Id Ew 0 i U)W EnZ wZ <r I-U) z :Qz a0 ma
Z c0Li0 mL .i.J z =) li-4mQ.
ckZ CE) u I-La U) 1Zi c ni L)L) >- 3LM CZU) CL) W " m
W c _ ILL CL. I- i0 <1 a- i-q Idq 0 ci: -ic)
MI a) U) ULi 'C ~LL ai- - ZE1- F- P"ji-M Id ru F- LaQ L Wd m-l- Zn U2z 0Z Z) WUi- w )IXdI Pz inZW LI. -jw w WC- *-Z C 0 Ci.. -j ZU EnIZ ~U) CL " tellZI)>- C = IL C X * 4 0Z aiaZL Q LL~ M I u " T4 U) ~C -< C-NOU)>- 0W W-X o 1±1W N x.JEIX Q-j-=3-
- U) F:- w x E Ce-4 F :2:2 z Id Li O:ii w e w UU) Xi:..1lo >- = )Lf) CL) < iU') 2z d Id w>- Li zU) Id l- C..-j z m -
U) Z Q a~C -j i Z Iii.j CCW LI.. C i-ic Z X - CL. <CEWLLC WW U j Ci CiE, xExamZ LW CL- Li-I " -I)I Z n L1 -3U M Z I- L IY4 L) 1- l2- U) I-- =I m co m 4
4z zCUI U IT~Z (nU ma z i-iiQz 1 1-i ( "QZ Z H U) 4 (
Id*zI Cr-< - z C C W 11 17 X 4 Z* 14I CE aa I J wI I
< WL0 ) >- Li O 'J U)i:i- 140-- N X CX 0 w Cl: Li)i 0W)d -Z ONz;E -)0 - X4 :4C -4 W"- i-i . U) 0 H-a " w -i -4
.aLzU)c^cL--ZU) OZO0U)0Ei- UCCL <WJ UZMU)EEZd-4CLZU)
U) a- LoCH !U '- r)z ) 0 M ) Z Ci-Z C
U)U)E-i--i-i i- LL m * EClCHm CL "i-Za U 0C4*C CL-UC4 X
A-3
LU -aLU UPQa~ Lu Li i-i M C$n p
C C 40 U H coM1 r- 1..'4 C c xr D IC
LuI X ELL J" M Lu Lu Z Lu cC Z" 0 C) U. C 0D n. M- -az u -I, o f~r o c- 0) r 0-C4c a L>-H u
nr P IM, U)C M .e C to Ca W. )-u (nL) Li L) -j UU io Lu $Y ~Z L
_'4 C LUI~ -. a- z D a 1:4 C Of c(P -H :c -Uz C C. C- x 3u Z 1 (a-
ZE. al Ca.. C M ~ Zd i) i74H Cz i LC i
U P~ - Lu m Lu (P ZZ Z ED HWZI-<- L c zi H- CI C1 UO =i L)u En :
C z U C L)' Lu. Z Ca Zi CC (r)C LL U - LL W; z 0 C&I JH U-
CC~ Z...j C ILc C C W 0 0 C, iX MaCL.(En CW, ze - 0Q i CL :3(1 il ) .*H- LUC : C ! E UJ2:X4i. ni-i X4 ME Lu: r c~ ~ui 0'H zCEC I.
HE !x HLuI r u- E Lo UuCL cD mcCP Cm- 01) (4 j Liz1Lm2PfM Z u "CD C M D~ H4 W Lu 0c
174-!'-aC W - z Z>Lu H 0) <E 0C !XCL H -LL LuILU HLU 0=)1C O UH)~( _-, CL UC C c ) LuL)Z--E =aUmE M Cm U) z L r-E CaC Z- Ix z :4-Z 0 DQOZWO C)- H1 :J~ 3C 74 0- ;T Z -i= F- Z Wu I.H
0 I 01- . i, 1: PLU Z u zC 0 U, ak 1 cc x -i z ca Cz z -- WIm 1z L ).Zri Et CuOC C) HL-,: f C-
*ZWOC CL-.JCA al C ci ZWL Li Zu coZ~ C) UJI 0 <
Xu LJI''C EC =3i>C C LJ L :3 Ld f-a-ii- Z LiE ~ ~ z <E IT4 Z C~ W Ld -C i>3 I)~ I ~
C:uCC)Z(NJCZ* -i CJa. Za-a .Uj CW H.ju C) - 0CL
ccm)J a CI . LLJ H - HC)- I--4H- m ix:L Z 0~C w, .JH- Lu :3H1-- U C ZH < LLi H. Z C)LL aI)
M :4LuA-1.4Z z x W LuZaa z '< P C)~L =1 I C a w- Z :3E LuM ~u Ix 0 L C Z Luai-aWi c Z. 1-.C _ HIx a Z C X~r - LuW 174 W W " -H Z Li MP C) H D LUC) 1'-1 C ILIZ ixC)HF- ac i Lu W 4E-1 Z C.1 CaE Z CLCL CX4 L
M 3 0- .. cz 7jM C P a: Ci Z L WI< -C HC > c CZW Ca : ~U. C4 WL~Z W CO L aC1 C) CI U-~ L Ca cEH m -ZZ >- >- W . wzH - - L. rCL Waa Lu J L u( -a CaW nD ci -j iamULo LUC -(f) Cm H 0 LWZ a-"iMCa-a C- mC.JL
~JZ LC C) C) 0 1: 1w C3 CL: amC Z CL U- u Z u zwP~~ua~~ cc _ ZC ZI- Zi C)uZ Zl Lu: 6-CZ EWCH n
LuI DILE ixzZ N H-a LuI uu aL UUH -jCw M wD w. m )-Limo C)Cz"<C-i Ca 1a-4 L CD LuC.T j T U- C H -4 Caz 3 1H2 . -im L (nH Lu >- -J X4U) -iLLu I U)- WPu 0u.-m Zw m= UJ Z " x EnZ M ..J " Le CaH Ca CL(to >-H 2>L. M*H->- Z)Z Z Zc)ZCWuZZWua-a.j U W M 0 W Za- z LU Lu _j "W H~LL.CC C)0cZ X: N~JC M Hm-a 1 z E "Z C) -~ E ~CA: cz u U-4 ILC .ZO aZ1nZWC H-ZN " Caaa-a 0D Li W4D 0 ZD z Tz c Lu ce mC 174 -1L)M Z Z C X X:W..j >- -i LU f-aX Z Z C-CL aI iui(P e Z
C ZEComoiHCfjJ '-a1 17 JI -4( COOC (P U-CC C Lu o -w<cmH-Z-"a"4UC << Z C- (Pi aww o" t -4 1-i>- W LuLUL 3 z0 u
<...iZ cwxe>om( Ce-CH-a D. 1.40 1*CL Wu -i X c Cj LLuHCZO)CZWZXI LaUch a- Z Z (n rC4 En * M a-4Z W C4H- Z tnwxaaw
(n C CD .Jz I-L. Li CUL M- mjj m a- U- ILWJ - c a-i)>-Z zWWuCLWOUC)LiC)E< Cau-aWa-im Ma mW-j..-j~ .JaC LuI MXCwMOZ..IO*.$H*JIIjeCXC*WCe *CXCWCZ0 IX.zz-
A-4
L- 1
W 2: X i 'W wd
0 EW -l) ia 4T1.! ")2 W : 9- Cr C>
ixZ c z 2: x t- tC I. >- 0 CE CE m i zr U~ U)
X: <2 W 21 2: Z - ". m < -4z 0'- z: !J Ln Cc-- a) U. 2
X ~C3 CU iLJ X4 Ud CI
ILI C U) Z t a: U-1 P-- <1 C z1
JUI U-jX c LaL wJ =- -j M CiZCC a: Qo x i - z- < ui LiU IC CLUcn oi LU LL C.. -j a: IX 0Wu u. w J~ I-- Cj IM 4T -- w 0 w!X14 LU.1 2: z .J >- c 0D
I =)- uJ Z: X - :3* U Cs-0l r F- I- < Ccz: c':
LL1i Y4 3 2: z: zU c UC Cz w 1 0CEO 0 0 Ci 40 $ZU L) -j
x z:2 -- U$C:U * ~ < zmUL) C -~i Cc =I: LZ : :3 uiw U ) W *C< T4 CA 1 sXI-, J -1U U) U zU 2 t4. r- Ix "- ED- m)!'L ' Lo x U) Z Z 0:~:UU 2: N NNhi LVs. i E c U)-1 ICc -1E~ F- Z-i z Qz " Q
W C4 WLUC- W- W ' j--i2 Z -- x U 0 D1= - F- U Z * O 04 CT to 0 WL CwIC mrew =Cw C LU - Z U LA--3 a sZ MU) n )( X Ww W- =c U-<C
-C r En WJ mE z x )'- I-CC >- WL Ui±.2 U Z n cC U mE< C I!: A-- -j Ix 04iU~ E Cn-iM C O ZZ :2 -:2..J P * F- LI.
.1 W 3 CMi tN 'J*U?-OC >-*2: Z
-j CJi-"i U J I-0 - 40L CIx CpI l w w ce 4W)U LI E 0-C~C *. -j' CIL. W Ix <.j " w 2: UC4 >-(
2: 74-J xC. Q_ -U U)U)U)U)Uwi WWD. W<CLiJWJC 0 NC<r 3u LU LL.UUW wWWLU LLIX E -"E - WE W -j C- '-j Q
CC': W 0 oUWUWWWW2: <CI-O'-UWWiWI-U U) -ic Im Lo ~U~- I rexlxx Z>UXC z-3C 2: "i-iOEx x <C I-WCXOCD0i0 <E 4TCU : U) C L3 C
L Z C CL X WU L-LLOfWWUWWI G-F-Z .JO>-.JMi 0 x~ I- IxM C X CUC I- " W P OiQ OOQ U C4 = -JC Z W F-- (nW W
uj 2: LO .. J .j 0 I <1wb..Uot-2:Zx:DxEl-C< OW I-l- " ~ z: ce -cCCC : x W W M: ()UMZ - CL A- c ) >CIxW
D-f C: r-Ia. I-- M i MaiIX4 W I- i L U U)'-" W 0W CE z -1 ZU)-0 Cr 1- U < x XExEr-XCr : CZ C Z W M 74 -4W XW (n 0 -j <
m C CO C4 2::ZZ2:2:m > W CCMU.U)I"4ce CL X-' "CCL-X CO WIF- 14mmmmmmmj "x4uCU2:=*UO'C- W F- ZU))<xU :3 1-- C I--O CA': '- -CW.J=3CW )LC I- C 1-4 UZ L .- < : -4 z 0 1 1 1 1 1 <E LZU z )C u) U)U ixb..3c LL.2:
Ow I Cei- m " 3 LL. C I- C W Z I- 0 -- O~iW >-I-" 0C cc: C3 1I-h1i I I I N <CL.C <IIW zfl c) H wrr ( z c If
.12:0x *00 U I-'-CZU3 >-M I-- < - -X-
W " Z WI-ZI--J WUWWW WWL.WZ-WMWWf'- c~e Wsr Ci2Z)E'C x X X xX x X -4 W xX 0E W W XZ- U C CO xQLz 0)
IZ I-C4CL U LL. CCCCCCC.JECI CCE0C: OU C 0 " zUU)Z "o- ZU)ZZZZZ2:ZL.. Z: Wz:wzI-Qw 04 2: x
U'-4UOIWO~CCCCCII-->-<CLs-CZ i--U I WC'cwI-*
A-5
t~ X z z w-
w 0 '- - <: Cl) LO
L) <E 0 =)4 1 0 < Q4z I W fO ( ii 0 W Sj ce Ix: <E Xr
=$1 L CWI- " 4: ZE I- 0 z I- 4i<1 wL CrxCIx 0 W Cl) Li < <r 4M4i w :> 0 w I- WZ Ixt LI z: Z
I- < F- CE:CL Z X4 0 w W Li CJ- Cc) F-i Z: wi 4: Ci. CD m~ 41 0<: 4 s IiD.i m i I-l (n Li w i W Cn
3 4 >- Lii=4:OX L 0 4x 0 x F:IC:-
-. *- LI J r az Cr: 3 I- '-4 07
ri) M ML- w uI F i (n I-- En zwC 0 M En4C r-UXi LLJ 3r c- Cl) <Ei Z LL. LLr:CZ IF-Lj
_jLL W)- '- 1-4 W 3 Z:O U) 4I- 4C <U.Z'-"C0-L) N4 41 4:0) E cc: 1-4 '-i io (n) Ci r-wCL) -- i4-j -I- "-4 CD I- w .- C4 '-4 U) CL)
L-44 )-'E - Li. XC X Li -jw w mcez IxIi Xr'- Z -i (n 14: < u z w wL r- Ot- ,4 CWima 0 4 -z < W M Z'- " 0'- : z Z4:Z Ce l-
bZ a:- 0 L 1 C Z) >- Z U) Z f. <~ ( -4
cc. m -m F- U x tO >- z: 0) (.0 X: LO OT <4:CWZ-4 '-4 C 4LC. i ". 00 Cr I W OZ U)64 n W-Z
IT U - MLii < i Cr ix Cr !X z Li. X z ZI-CN >- IT(nL)" Cf'4I-0- Ui 6 4 -4 LiLi w i WI- 0i 4: 0D 4 a
mO 0Z z - -Cwc ti xl Ii E ;EW Z U iM Z iX 0WI-- L -4 DZ U)Cia. Z z z (n1 WJ I X - <4: w0 W O :>Z m r z w < =- C1. 1-4 Z -j CL '-I Fi C- Li. 4X
u-i 0 ZW m I- w z LiJ= C X~ z C drl) <E~
SL L- LJ1Z L) I ILL Li "l -< Dii ~ C Q ~:4 OfL-x
zU to i- W Ci-41 W a, W tI X :>' Li.W m 1 0 < 5w 12I-. ~LiII4-I 4: Cr: 0)1-ICL. 4: - Ci. < 4LLIO
r Z LL ZZZI-I--ZZ C: Z Z wu Cj:o Z W C 4: _':Z Wld 4: .O-1 Z U -4.4: c- CtULLd CI m -tC 4 Z) L
Z:WL C Li .t t Cl) Cr r: LiXQi j cl 4C Q U <L): wrW L t iO Z 4: 1=1 Cr, Li I-Cr L -C LL ILr: 4:i>C C
Z4W X-W CL.W U C4- I -- 4 cr- W '-.L1 I-4
-Cz i. Z zi CiU C. 1=4 Z) C Ew - : , U Z W14 <E C -Cf W -4U(n-wI. di <~- U) 4LiiW)-4: z +r 4: O>-ZJ Z:i zi N~ -0) W-CLd-:> -4l r4I L .4 CK- a: - M 14 * L WL iu Z - 'i " X U) ICU)2
'-4. > <--Li C. CO-ICr>- Z LL Lu M X E) I- LL 4X M > --0)4 -jU)_- 01 0 wrCr Li m Z4l-ii X:4 -Zii4 O
-W4: Z 0 LL 4C CDZ" 4-: 4: I .W ->- _ Z Cr-LO L W l LL 3 4 4 O -L.U Cr 0 I W- I- ZU U)>-4*:Z c z u
Ln Ui M Cr: I i. m 4 W-4 LOCr:4-:Wa-'W1= 4:4(nW34 w - < <r : 0 -fi44 --- C IU)cX0- ZUw C u LO1-:L Z) Z~~-3 L40)-z W
r- w ZiU 3 IX-I <0 -0 L r Xi~ Z - z m CJ - W I
or_ W : 4IL ZL.-. WL rLL -j iiI-OO<W 4:ZCn4IX4Z0ZWL. d i .I I- M -4U) 3:~ -4UCr LL C-C:: C. M- 0 Z.0 W
L C 0 -a!CU W t .- EW I--m >- w WzL C L I 4 C4:LW D <L4: _j z x m E) m*. C4< . Uw<uwCW-L ) x A Z Z0
A-6
.............
i- C7 u4
Ix, za Z Li (l 03 0 Ix Efl Z Li fo
ix C -a C .I-
w < Xz E I- z Q
C- *J WJ ZE El- CLi 3 <V Lii I- Li CwL 0 U)0 CL1- X: m xa 0 LiEr ) Li -i .J.J U i E C r c 3(n X z~ m L z Z i . z mCE < "-'ce N M: I-q N~ c x.
Z cO Li w >- zU En Li a C..i X-- Z i u OU Zn zLi Li E 0Cx 4C CK F-C 4 - Z '-" Lit I 4x m z zz VZ Lx Cx W~ F- X X w-Im0 i Za c-L C cr I-V.*-- w
w M 0 a4 XW C :IxOLZU) zI- L) x '- _ WE C oox 0 i: -
r-l - x i iLO (n U) ( C z r .- ifzZF- 0014 0 F-4 :3 Li U)E W L Z IL U-
Li Cr. ci u Z lm Li w 0w r wE r-C4 m~~ w L *LwZ S C4 Zl Li L C, I Cz Z i a: Li I-C IE xzLI--z Li0 Z i3i >- 0 I u~~ I- n U-I 4x mI m C4CL -4C4 -j -q M Z Li W C4 IXCL- w u OECr- <U) I CLI X: E)C C" ox 1-i z U) Z -Z U) EZ LL . CL ZC L WI- L.L W t-I = -4C 1 F- *V.C q Lij LoC z.U X: E"- m OC- WL = WX z Zr- U) Li C.J LwI-- I - _j X -I- 0 IxI- "-4ce 0 a: p)Li 0i O Li Li UWC
(nl 0 &I- I-- )C E 0 C CL c mw Zj W ).E -jV~I wu Li Lz ZC<)- i 0 0 0-- ECL-i V 0
Z 1: 3i- r- i - I- (nL w LIx r e C C3 I ) i -jC Li- 0 UiW 0 Im ILL I- Z MLWCD" Lr '- 0 CL Z c: C-Ci z LL zc Z CU)i J-Z xE Cz 0 CiZCL 0 Qri 0 w U) =1 =3C xO) i X -4 r-) U) I-O wL V.L U)CWNZCC-J CLi LiW, < xI Z m C xZ M: _. jJ -ZC .J
U--4 LIX C-4 w m I X,~ W ~ UV ri'-E3 L) Dx Li J1 Li wZEX CJ z T E En )U)C 0' -' LiP '- Z L) CL .!x )- w ML w E= i xOl -4 (P x C4 I- ixE ZO =1 ID m - C I-VW *w<z 3 !x E W4 C~Z X4MZ z- z>- 1-W Li -J V.- c j U) LL XEw EU) Li<ClA -- mJ :: Ix - E LLI -J cl:z Z
Z ui u CIL Ce Li m w U)EZ LLi m 1ce Li I~i< -C4 jJWi LL.Z a-L 0 !a:- Li = -T UC)F- I-Li U) CC D iL-I '-ccCL w n : P-f ~z D. F-i=4Z 1114 Z I- (nU)Z zCco m < w m Li>- OQ Lo-m * ZE VT WC X: EWEE:X: Li LiZ..JUi) M cl-
I- XJ Ni FZ -V-) U) 0 w w n~~I W LCX4 C
m z x :3 w 3L u 4 - 3 .4C C Q.XJM ice -4 "'-4-<I- I- CO <u CC M 0HU m CLEE W x E : -=
(n Z . CL x Z CL 0OLL 0 >- D ZZZZ w m m w w =O Cl Li w U) 0 Z.J--I- -1 0 . mm "~C U- I-XC~ MZ
m> X~ U) Li uLi Z z wZIC:3 x Z =IEzcrzxLzc CCr M Z MI4 iZ Crz " CL0 N~ " z j 0C~ Lim
*i ~J CL 0 - so -z O N " )nCL :3 =i.muwzx~iuxm LnC LLC a- -L C 174 "- F UCL- F-0 -4n-)J u X (ZCz CC c r Lit co Z rJ C) I- D-4 -i" 0Z3 ZO ce -LLIOi ME 0 = Ll .J Li C zECW1Li W z CC
fI- u L ; (n E E X: j !Y i- M Z ) .. C -CW~WO-ZJ) Y4CL t Li MIW= Ei CLZ) .-Ic C"I'UZO -~CE ULLWO~Or JCC-qIU Ltru (JI Llji ZZU z iL W)C ZCIXC ED""-,j c~rzlz cwc<-W IZV.CCU) Zwip-w WC=-wcrzwcjrU*.~CLC-CLXCL. ce
4~ ICL LA- <40V-L!L1-- 0IV.CLZI:ClC -I-ICLLiWjZ*4C Li)<EOEWOVCl- c>CE nI .nr L E L L 7 - L
A-7
I. z i'-4 0 Z
Et z m LJwl
Le c -.z I-C SY4 U-1 z r xC C
E m (Ds Z4 a: Z oZ m
x) N 3 0 <C X : X 4, I Z ii LLI 0r- IC Z Li - Z'- Lz Z w 1=1 z: I Li !XL 0 LL 17(1- U-1 -4-- C L U J r4 F r- OCCL:U 0 W L
L3 0 MIUf) t. N Z L E ZC L) L) LC- ST4 <C$ +0M C 0f .j e ) r-U !X "14 CL
-i - E -J C F-I -j z Li - z CL: U) LL W LI =W A~ . C iZ w D1 WL Li . Z rJ-ZLtCE 0 ZI :3 C- -j Li, rJi -j WCEZO-CZ X 0 Hm X Z cn W 4-i _ T I
C cc CeJ-~ m U) L C) L.EC iC O :WL I.--i LL )zX4iiI-i 1-4 H Z3 H - iW xZ m )CWHU)W L
CL Li j C)CN1Lli Z Z EL E z >- cj C C1Li LL.x L. ULi W Z 0 0 Li WCx I >- Z 17
0 U) .- Z L.) ) U i m UU) MU Hn O- WC-1izH z C z U0>- H C E CL: C rE- AZ c ~C:W 0Le
, zZ -j c C L Li - XZ ITI WL ILL M LiLi rz: Q- i-j N Q U H -i Z H I ZH mx i Li U) WC 1-4 z E EHI 0 xCL:Hce :HC H r CW Ix 6 z Wi ?- C C" MO x-0U)Z wi U- HCX:ix:0 C r Er M CL) U ZE W0 rzX 0Li C c C=fl I) C- xi c U)4 Z xD Li Z 0ZU :3 C)H1 U- u CL: z cE z C-e X M-4 W L-C 4- c UC4r L e- Li 0 Z M C x L U C Li.1 (n mW wiWi0 x m) 3 N F-..J X4 W Z 174 M!X
11L WHE J-Z LL 1:4 iE Li U) WU LOr-Z) - Z I)HC ZC SU)M I-- H*Z LO ce LiiW ZW w <<cCxIx
r- .JxEZ C WU -4 ZU 174 Li IU) ml$~ torNmz H-z Z0 U) Li -JilM'"L: LW W.. C 3 HOC 4 U*Li H-Z-'ix n i U)L 1-4HICL: CCr.Ul (n 1Z Mzz zxuwwLi. "-'mC c <C >-I Z ) U) 0 14 z (n O=HCEi 174M
r-. XH -j :; ..) 0 Z '-4 Li 0 '-4 a CZ _j C ... X 0U)1 Z CL C = - - (J) '-q co F- C:E -z _j r).1 CL Z U M r ZLi x r-4 0 m i CLC C -j i H H+ (i Li tU) W Wm m) Z .U ... M Li 3 L: Li$' CI Z _j ' ( 1 IUC: -Z Z4CC4u rZO OHZ U) u >- Z MLU> C4)H Wmmu mm U)EC:C N.1z ZO 0 -4- LiW cc CrL Li =-WC Z ZEHF Li U)
- p~ m) U3C c C: x1 D. x''W >-E WiZ F- Ml z mmwcwC CL:L '- " ..JW a:Ca Z -U L- XC Ml C M NJZH E-F--j C4-) U)4 - ( 3E 1 3 zZ CZ - - CC _j M ) ''Cc :HUp
OEO - rk:CL 3H -W 0 "Wi~ ZCr:>-Li-j Cl U0 LizEn EZC -i m mW WLL J 14 =I 90 Cz- Ix r- w U) -)1 0U1:z w -iw0) 4* ~C, cE xxZO0W LiEx-ED)- :3 EU)U W Z 0 CC:-c x co cc -ZZ U) mL .. ~J)- zHW 0<wI :NWU CU)CLC:r
<. E>j- C -,4 W < QE.x-j M UsE.J _ J-4)I _I"M j CL)* x-jW W L U >-LiW H, ~Wi-Li W-4 -j UXE)LiHWUU)
w * zW W >- U)UI- -jU u Z WO ZO XMI-- M) zum CL Mo" -CrLO i)- -J CL:OD . j z-C X U " Z CLHF COM'- 0"WcUHE >- W 0 X-4EUM "4XCU a XEMC C CZ O Z E 01-OH 4- H N. CEEx 0 0HU- "C-
EE-. "Z ZUCXMXM 00 *U)Z DW 4 CH U)LO Ix zoLiZ "MI "WwH- 014UWCWC<CL:CW NWZ)>-CL-CL).JWiECX~X~ Ic EE<*"=- =EcL:LLjx~i>W:3>**NiU) U)DX3 0.1cl cC1 4T D. CI~CDC:L I-U)w~ wL~ xx 1C-4 MI- c rL- -CLCXL:EZ xIr _L: .Z ZH-4 W Wufl <CM)' ZECZMMU a m)MU)H CL W WU)t) MU43 WHJ)i'4LE ocfl WI-IEL WE1-Z *L '-4'-4Z"4W *U)ZZ.JZMOZ *Cf:Z OI-zI-WmZxZHzZ
EE~H~LEEHC~l'-ECH-i'--l EC~H~HOMl
A-8
H WL U) H ..
C U) 0 C ~ U_z z r,. CL( UW c) M)
-u W1 C 0 C LiZ=. '-i CZ E- L) 2 LL~ a: J U-1C <: -x
< C W- yQ X C x 01 M UZ E: Li : HE ct 0 a. CU)
x~ U) '-l L) U XE L -lZ H- xCLr : C LIW. H% E- CJ Z LL. fu ;E E
r ix D CK U) Li --i CZ H i 0 m m Z U L) I~ 2: 2
w\ I=i '-X z3. mr r !X 0 Lid E. 3 W CEX: U) NW LL. : H Zi W : Z U
Z C r C(4 Z C-4 ULr I -'X 'Li"L. C CO Z: LL r_-. " M-i~ H : n 0L .i Wd
x~ ix 4xH Lil LDE CM i IL 3:(n Cl a) cr: cc C U W W U.. Li L CL: U)ri E4 -j *E Ma XE -I 0 H 13 r 2:
H Z LUI Wdf = W ll Cd EL < C -C Uii 17E 0i-r r- EN *E x: 2: 2: CX - H i cj CZ Cr
E) 1= H C r: cU) W l N Hu F i .0 -1 L.W ZH F- li C U U 1 UMW i L w U
LL E X: WC Er U- 3 C lz IC H 3 Z i W. IZCr: L) O UJ4 0: _)zu C j w C !XC X Li EU) z~C * C Ix -T <L LT 4za UI L TrE H H C ZI zC
D W L -jW U) cr Hn U) 0 X~ w u 4E: CE r CL 0 ICmLL U) H H) X C) Z > 20 -j c. = C H c C 0 CE ED 0 2 LL. Maf ZW CL zE m Hp. -j - C (n~C HpI *Li J d : 4 WW CW E0 U Li) Li CC CL u Ub C: -J t% z 20
1:UU WE C:- Z"w U) H LiL H CL E: SXIUr) m rECLU) O z zO cc2 ZC Cr :i u 1- C c w r W EH'W
Z i = U) M ~E U) : ri) 174 Z .i Z I CQ'z ix W) HC- 0- E I WrW C d <E4Q r IT7
U) 4-i .j HW 2:W C- D LL H L m. IC W .: rE 0 EH FCE -JI- C N'I =r 2: ix U U:1- L) LO2:~r Z2:4_j 0CU).En UO = E U) W w) ZW W 00- D Zi -U) 0L C 1=1 W2:N CL UO =E W ' EC WE rU) w :3 W Z x =O 4T E E: CC c .. iE2 =1z z4I I J WH H- -i CCH Z:i 0 H 0 jZ H W < C . ' DEE: M:-4W U L) Z -C ix x X,: (n i:_ w M WEZ X2:
HU)En 1 C 74 U L Ii Li.. Cr LL (n)U Z H- &) E )O z M VW <1 CL W u (nW CLEx f4 cC " I C CC " -0 (nE: mIU- z X:EH C CE r : Z LL. C :3 - Li. z:~ mCI4CT F--XU) 4L)U -.. 3 C 'gq W W CL 4 CD 14 0 -1 W .JZ z U)'- C CL a. i IJ U- ZI Un L -I W I- ZI2 m Z L)U) E X: M)CL W W .JI U C 0 1 U)H "~ X: l -4H- " LOE
cc: WE EE zu " --H 2: Z CC CZ- 0 ce wC x -4E c z2 .- 4 0 CC U)Z 0 H F- ..D 0i E: CIx E:C-- ;E
m w <C '-4 ar:.. U) Cj cCX T o C U)W U CH2:C0 =) I
U) = Uf)W LL U)CL --i 4J C -j CO H U H- ui U rC u c ;)ELi ZH - U)n i-i E : e CL " -4 0 U) &T4 x: u W z= C*J W '--4 " 17-1 : U) (n EU)x C W T- En m 0 mNWu x E * xU)M Li C im x U) H n U) . M-'> Z ,Z1: U rZ 0 ZO-' U) mCWU- LiZ C U) C W~ 17 j W NCKJ-Z Z Z Cz4= 0 x .- w 4-q U Z i-C N .i C1 -1 X:i w P -i 1Z WU ) 02:4C LL. WEE Z - LL u .. J U C) - CW W .. J-iWEX:WC C Cx
w W W-EC =1W~ 1 -44 Z ) co U) ZO0 I4 3:E ~.ZI~ M - iW WEUEC~C CE ir) 0~-i WIz)3 -1~ 0C < 3H--4H O EU):t
Li '4 C Z WWCJ _ - HCIlZ 64 ~ H(J x:m DP qr- "CL cIx z W C J "i-Ix HwU) U mU WW EZ2 E - w r i a- W Z HUIZ4 nUMM) E WCOMM) UCC(DWm0C rwzrw C4ZE(W u~ co -'OHOz 0 -1 <W < j zUwr j CLU)-44)~ W)~2 "HC " 3ETx 3 U E z HHC w m i =jiCL WEc mE xzHECr U ce C cr C 'U -10 X: EH=I-.ZNUH-CL U ZN"i D00 ZI m Zm Li...ICCL (n z t)cL "-NE1
JE U) wCU <1 .JU) E(n) -14 ( 0wr HCL. H H44 44'C- 00Z44HIjH-20:Z2:.U)z U~zr-L.jCImIw m 0
A-9
Cj x ILI z
ZCn UCL C4 JZ EW'1 Z I- ZN
rj rc c x-- C.zccZ' u- uj L a ZC f C- C
Ci z- Z) r. cXz Z ss0 W
C 4 in zr t- qr c
<L Et tr- i:1 C44* llW i
i-4 Z'~ II rc-,x zz E;
0r 'i- M rC 7) OU! E
En~ C 4 .r
Aj Z S. U C ,tr -)IuL. 5- n'~-7C ~ 7.,2 zs Lz~Z
f4 n j d MS M~ CCCcr,~~ U CdXziiL U) -3 -zwrD -x
-lCZ~ Orlri-" -xw zC ZC zOC
7C-"r C L iLrCEU CJ- Z L, i Z X:SW wC E z4
Ct -j r- i-4p C c 4L j L L w -i
7 F U! Z En C CZ C- CC rU- Z ~ L
Ld__
LI '
LI C.4 D EU- tfl Z _
Li U z '-4J
Ef ix cc E X 1 P4 L i C1U LI U ) E H i Z- 2 z U-4
eYni .01 x$ Zf CI r LI2 z L $ C-
ZCZ X Z $ L $ f zH - Z ZLI-T r C- wrr M ~ H C I L
$E~~~~ r(c i cc ,W rr* -4- >- r L.. S C2C ZCJ 2 uZz . L-
LE$- En CC ZC C:, m CDZECCZ uC DD rtC CC C zC IC 11 ZC)
X:r~ Q ! C I C1r zz j zz 1ZS .U)S 0;1 Ll) Z $
C!Z - ZZ 1: ZCr LEE 0 W1r2 z C CLXI C cC .0 Il
z r - C- 52 z zZ r-- 2u r,5L14 zC IZ 0 i 0L " Z- t ZE- !X 2Z
CC ~ ~ -r I- X ~ .± C: L zr .L)l C C ZC- 4 LL z-' DC M
(EWLL -I Cix I- 4 .i -C L x
it.Z o rL Z~r 1;4 C. 10 rr X
U ) C 0CX C1.4 £ t i e
tit Z, C. -m ~ r 01E fC EDZWl - L r
Y~ -C ,l C zi C. C CxW c)
i.. zC =a 5 l ICI- S C
Ze C1 'C CT.. L-L CCJ uZ.LSc ~ ; Lw 0"Jj - IS ~zU0,75L w3 CN z xC) Z u
Lr hiUI-- .'C 3. CC-C.c CI- I cf! -,C .n a
C~~~~L -t C L"CuC-t .ccCZ, . ,-. i -z z z
f: U.t E-L.-lf,.h C r sC1j Sx -! 1- :: r . - z i '-27 _ _ _ _ _Z
A-il,
rz Z ir
Z z c
E-± -- L -zz :m Lj
Cz aC C
L- JC -lC 2L- CU ECCC
!X -H rC 4c z
rc -iOSS<-i
3E e: A-12
owl MM- M PM'CMR-
-. C C
7-. C
<CT:
C z -
7 7
zC
C 77
C' '
-- :E z C tC 5
7 -: C
C '..C
5 ;, C D "2
Ci 77
?-C 71 O Z ~ W w I'Z u
Zz'- L'.~ C3 ... C C-"' Li C~~~7H~~L cZ~h- f,, 7,
C L- C0 .
Uj~' -a ;
C :- CN' C x- 4 L c ,r2 z .Cio - <1 'c
A-13
MW
LIi ;C U;4- c er
77 3: _,e
57-
ZZ4
C~CS
I C7
7 7Lz . --4C
r. -C-
z~ z
C~ CC~
C CD
Z L iS:>'~~
77c .0 _7
Z
wa -CC
Z>- -- z Z-
77.
c Z c< K IZ z 2:
A:Q
Z :.!:E *4z
-t.
w x
w r
CC w 2 >- C3 CCor w Z fic z w z i:4 LI
LI 3i uI Cin 0rr r-- i-
:> WP - 3 L LIu) LOI <E 2" 'CC a CDwwL-i>-* M: o eo no wc w Ej iamz 0
m-- 0 Lo H HCn 4z H01l) CO CC H zU Mi mn C mC <au W oLUw x r -- f >-~Cro r )C z -J ( H.-I >-CO C C 03 -j L-1
02z= Cr 0 ZI 1-4 LL Z(ZH za i a- ix CE LaOCCD m z C C 2 3 C[ E5) LI ZLa z w(in
HLCLI~ n :4 2Q 0LZ D (2U.4C) L-J C~ W'-COZLLLILt) LLL (PC LIM
wP LiZ mJ WI LId Hh wr C n ~w r- lz-WCCL_ T- L C C Z 'i :pCI. r~bf iehl =Z L-O.-O L C
F-s L I C HZ z Z W l = 0 m z:C C.::
LaSO g) LI = I- wr- j~- LI LIC: WLi-. .. - r !=i rX i *.
zH C oX J Hz- CDHC Oz 3 CCCC .-e 2CCC~L X, cO~ 0 SI z~t z- C.i
U) z 0C cf W! Z' )- 0 3 2>- ccaED C - c - -L3 h C C.
W4 C4ZUI) LL 0 > d iz<..C U CxZ :3 C Z 0 Z 21j C LJ ncCLIC:4zOC C z M U)
- m '-C0- L-i W.z r--
7- wf .- mz.-- -S c - -- :-,--~r i---' z *w cr..ZZZU,ZC~-. L'.-4 it3 -r- 0E
Zr- m Z LI >- Z- -
:3iw 3- L) 2 N--'---CC 7d ZZ~ S CZ
C :> ~ Cvs -i CU C- C.ZZZZr.W
US ft -- )w ;:IcD m&4=
C cLz zLI :
d C 2L 3 3 - Z :3L)3u i
w- zCM -SZU -:Ya
z LLI ~i 4 "CZ <- 0 m-AU.. ..1
CD -Z*- U- CwZZZuZ
Lu Z -, LU :A-15
-"7771
C.>
C M
zz Z:
C~ <C
Z iz
2:C ~ Cot C z: ~ CC 27~
>t -
z z -E j-CC r 7 tWW
C.-.- C d3~ C Q~~-~~'c~~'-~ CC4:C:'2 -ri >
ZZZ: ~~ZL)Cj z Mf~ Z M-.-? Z~ Z ~ Z
~~..LLLLLL~ > C,± W! ~ ) i. ~ LZCCc X-)DC C-.. P u *j C-4Z iL
W 4 ~w~~Ej2~ Z-ruZ:~C~ : -..- C.~r--s- -z. <:~ m
Ci = _ C, -qix S n r, - foA-16 czn N Z Ld4
:> W-. 7_
r 0 d0 p
c) W wC w: W3c r ;
>i; i Cz 3-l L -.0 z:- C C £M - _ :- LUUl
m -r iPC ,Lt-P - K
C 2. CD cr 3 c -c
z - 2z 0X: Z -) Zt
-~~ ;Q -,-r-c - CLa: Z ~ z ,~ x .L C: :D. -. Z
c r1 0. * <- * N
= .. - .. i z Z ~ -- -
2:i "r - 7C- Z -.1 C C .cZ
Z!. C_ - U-:u -s -4,.iC W := m . )- zi i.
<: Ii A r LJ c :zzm - - w -i =E :
CutT]m z r Z- '--a:0 E:. Z - r7 ) 2OL :
i-iss CA.s.C- -J 3U2. u := u ) ; c;
:a- L: E: zEizz2c: a 0 1 m 1L
zC z- ZUZ ZZ -C -- i, : z -1
z 0-1 c -r c 0 IL
:A- -i , CC :-j io C
S' <:3 C. m3m C2:- LU '- IZ LU - U D C'- il 2: 1 -CZ L..a C i 22 LJ-'2:
z LiZ-. (J) X, C:: -LL.s2 CL.L-i La.U WWC-W~~~~flC -c- O2LJU.,~:c Ct CrIAJ-
m Lzcn Z ;LUC2:ZCC is.,tW. C_7..'Z LE:2 i- JLU Z 30 s-'J..LJ L CJ0 inn.=ZsWL Li-L± 0 :.. WwC -i2
Zc C". -C tl. L3 C af'- U W 00 0~ co jj , j 3C a - J i J A a - 2 $ C 2 W
5*- rzz 7uww - c a 7- .2or *
z -- cLU ~i < r L Z 57:' "4- 3x
Lu Lu
- X:z 7
..- : -C . LL. t'-~ *2Z Zz 1 C i L C _ D2 , Wix Ic2: Z~ w wu _ I- D- l C>-
0 2: E -i j=0 w Li ---tC--C XN C
C L:L 4 >- C M) M DU 7
2: uz CD hii i7 j-X4 C : CC - ZC2:n Z :3 ~i C. ~ C E7
ED W- r , C.......i7:r -7
c- M ro0-
zi CL:3<:
ztz
z 21,
z rj
,-C
Z>
J. c6
7. C: z
k<7 i2_ 7 M7
'Z TAr_8
2: C>
* r in Ld wl :C3 z: E C z 0
UCU) 2 0 0i 4- =1 C D Lj zH Z44 N in Li z i r z - X
FL * CC C4< C: C- - Z
zi : E# r- -W U1 -j CC *
LJ 1-3 i-4 C. CC CiUE L U iiI CD m: iLi e X- ** z- I Li W n1 X Le- c: Z
Z-'C .. C Z 2~ Z~ U: L i en X u _ C.W.C f ';
D. ztr c w. F- L-1 H m Licr n W W i 3,i < i4 ~~ L>-J C :~ UJ 2i 6- Id CJ Z- S
CD CW1
d i-' * .dS uI 2Li ~ ~ C i-i-4 L.J J Ej II CO N i n S i
Z1. WC 2 i. i C -1 0~. 2:
z) z - zn u~ S CN E j0 C ti Z Cr r 7- 2: Z nt
7z z.wz Z) z ..) r 'r i
Z : Z >- 4:??. s) - ~ C Z. Z n
C'--O -D 2: U-1.3r an : h
z-. 2: 11' 1 c c 2: Z-n fl z hil-fl
Z- CL4 ~ Z r C, J<I c Cj C -C" -.
:3 C r 53 0 >C. z~~~5: >C i t i Z-n
2Tr CJ. ZL- Li 2n C
7- E7 E M.
c ia zC CM
sc aCE r L 1i zc r IC
Z O : z i r..cttZ V'- A.0 XO U L.Cw- Cc C z
2: ul zC in5 Z . Z mr Z:0Z
£,-.~C :.: 7-CL m-5 u C
U .C XLii -
5" TP CZiL C-) :Dj-
_j Z IS- ;'. -,-Li. i- -q 5
A-19
:>i- -1zLz._ CZ rC cr a) M3 )<!, z C. L W L
>-: z m ~ C- X4 LC ~~~L) -. C>--C
Z) 0 z- z.C.)
CC
Z Z>- >-:3 -. ,
m L - 0j 2 7C-M. zD r C
z >-C Z
z 0. C. 0 x~-
z LE -j .. L .C 'C Z CL-1 z-. :0 <1 -z z r~ E.
c~~~~> Z: z w
Z~ ~ i2: Z Z - ~ ~ ~ C ' J ~
CC ~~ zz ~ 0 0 z C. ~ Z S ~ ~ jfCll Z 0I- c ...- zl. U*L -z IIUL-X
IE L)z>- w C- L.-I r L- L. ZtflzfL. -E c " 'e . = 0~C - L ZJ XC 0JW
- i J C - LL -. -i C.j C -0 -
7-J ZU~r'~ CJ" W Z)
-1W~CLJL~iiiuJ Z Li CU C WU.ZJWU C P mC-4~ Z L - r - r)
M Z~CW- z~Z~ C)C% z < Ui
CC -j-'Z M C0 - -4W I u i 0Z LI -~- U.ZI
:D ZJ0iC U-Z3 x ~ ~ -- z r ' ZXZ C,4 -i~ Z LL-1cfl iulZ X ZuJ0 Z i-0 EC CC0Z t -.1CL 0 e N
rC~ Z~i ZNE -0 z .4 CrW)u
=>' c* rII # ~ U~ CD 0Ic L I~ Ll * l
.0 n ZLe)0 I 1- Cz ;4 -AM P0
l~-C
7-z -C
71Z-' -J
* .2'
7L 7:.zr77. 21
L7LM -C ; 1, 1,
z Q)~ z z 40 -JX 4T- Z W I T 0z E
C: 4: C ' in >- x
o w z >-La W (D 4 "x Ii- X #-4 0r x ~
0 z w X:z Ui. Ix w w c: z 0Z= - CD <: =) x x 0 cc>- '~r Z Z < Ld. w4 <: c sQ: ~W--i w z >-
z x Un -1 0 X: X <r w m. LC >o C:). E4: -j 4:x. a <-i U " :0C
li-o Z Z LL 0 m) t- 0ZM >- XL.. 0 >- Cre m U) 4x 0mz CA: ZC cz Xr 0 -.J -4: wu I- w4z )-1 - r- 0 40 CL X:W--4 Zr.. x 1 <E 0 cr CC M- U )
)d3 4T= i XL LL. w Zm Loxa 0lU 0 4: )- c 0 4:mZ U W 10 0iz Cu 094 x3 -j
x L) 0 Lo z w W 0 Z UZ w4: < z 1 >-I 7: 0 a-a 4 z N
C4 ' LLJ m-*3 C4 Z I- 1: >0 CL cc WO C) : 4x Ix 01- -X: -w X:r _ : <c 4: a a 4 m xC w
omw 04: X i X 1 Lj. i w~ r- LL i..J>- '-E r -Z z0 >- x Ji z 4:
cr- I cc 4:: C ZLL 1 U) a-iLO0 <:;cz 0 C4: 04: F- C: L) w wi- "LO WO0LJW OWLaJ C3 X 3 mA x mW (nLi. >-c u - Ut U- u. xa 4: Ui" c 0
<: W 4 4z L z L N L) D ~Cl- CLu EEU r- ~ J Cj :jUx0 c: z u c:0 CLWi z w - z:Z 4I- 00 cc il 4T Q: .J"-- Ia-a "-i X.JM LL.-. CL Z - I-Ct) M- 1-- X)' CL 4: (n LIi 04:E 4:Ef) ZC tflZ W 000- 4: EJ W -U1-1 0aZ X4 0C~ UW " 0a Ix w 3W W
Ui 0 11 MU C4:-4 .Z M W E '- rw =) LO4
a:Zfl- -E >- Of - .4 W z CL U N -- i " "*4uCLLU) ZN >-fl tN 4T z z- < >- ZCANWCL
-14 00l-1-- f (flW$0W0 ZCL$ Z 0 0 -aCLtf CALs Z I L OWCw D-F
LLI~~~~~~~ -22LJLiL J jii ww i > e wWL
r - 1 rr -- j r 4T" o - m XiWm mO
APPENDIX B: Lexicon
B-i
LEXI CONFEATURE IDIGFEATURE 2DIGFEATURE 3DIGFEATURE 4DIGFEATURE 5DIGFEATURE 6DIGFEATURE ACRAFT
FEATUPE ADVBFEATURE ADJFEATURE ACTVTYFEATURE ALTFEATURE ARRIVEFEATURE BEFEATURE BEFOREFEATUPE BOOSTERFEATURE COMMFEATURE CONFIRMFEATUPE CONTINUEFEATURE CONJFEATURE COPULAFEATURE DAYTEFEATURE DEMONSFEATURE DEORBITFEATURE DEPARTFEATURE DEPLOYFEATURE DIR
FEATURE ARTFEATURE EVALFEATURE EMODFEATURE ENROUTEFEATURE FLIGHTFEATURE GOFEATURE HEADFEATURE HAVEFEATURE IMPACTFEATURE INCLFEATURE INTRANSFEATURE LOCFEATURE LOCATEFEATURE LANDFEATURE LAUNCHFEATURE MISSILEFEATURE MFEATURE MODALFEATURE EVENTIVE
FEATURE NUMFEATURE NUMMODFEATURE NATIONFEATURE NATO
B-2
FEATURE NFEATURE OBSERVEFEATURE ORDFEATURE PASTPFEATURE PMODFEATURE POSPROFEATURE PREDETFEATURE PRESPFEATURE PROFEATURE PRTCLFEATURE PREPFEATURE PENETRATEFEATURE QUANTFEATURE REENTRYFEATURE REFFEATURE RETURNFEATURE RELPROFEATURE REVFEATURE ROBJFEATURE RSUBJFEATURE SATELLITEFEATURE SCONJFEATURE SERVICEFEATURE SUBNUMFEATURE SUrERFEATURE STAGEFEATURE THATCOMPFEATURE TJPEFEATURE TOCOMPFEATURE TYMEFEATURE TENSEDFEATURE TRANSFEATURE UNITFEATURE VBFEATURE VMODFEATURE VPASSIVE•t ( AIR FORCE) E N I .3:: ( AIR REGIMENT) E N I .;:: ( AIR SPACE) C N LOC I .;:: ( AL JAGHBUB) C N LOC I .;.: ( A MINIMUM OF) t NUMMOD 3 .:- ( ARABIAN SEA) C N LOC2 .;.: ( AS FAR) C PREP EMOD I .;:: ( AS MANY AS) CNUMMOD 3 .3ti ( AT LEAST) C NUMMOD I .3zi C AT MOST) C NUMMOD 3 .3•: - (BARENTS SEA) C N LOC 2 .;• C (BOMBER CORPS) [ N 3 .;•: C BUFF A) C N NATO ACRAFT 2 ..2 ( BUFF B) C N NATO ACRAFT 2 .3
C C BUFF C) C N NATO ACRAFT 3 .3( C BUFF D) E N NATO ACRAFT 3
s BUFF 0) C N NATO ACRAFT 2 .3(s C CAPE VERDE ISLANDS) C N LOC2 3
B-3
i: ( COMMAND AND CONTROL) C ADJ I: ( COMMAND AND SERVICE) [ ADJ ] .I: ( CONTROL AND REPORTING) C ADJ I .
: ( DEORBITED AND RECOVERED) L VB TRANS TENSED DEORBIT II VB TRANS PASTP DEORBIT I .,
( EARTH'S ATMOSPHERE) C N LOC I .;: ( EAST OF) I PREP I •
( GULF OF ADEN) C N LOC I .3( GULF OF AQABA) I N LOC I .;I HEAVY BOMBERS) C N ACRAFT I .;( C HEAVY BOMBER) C N ACRAFT I .( IN CONJUNCTION WITH) C PREP I .3
: ( IN CONNECTION WITH) C PREP,] ..is( !N REACTION TO) C PREP 3 .;
( INDIAN OCEAN) C N LOC] .I
: ( LACCADIVE ISLANDS) C N LOC 3 .;• ( LAUNCH SYSTEM) C N BOOSTER I .3: LAUNCH VEHICLE) t N BOOSTER I .3
( MALDIVE ISLANDS) C N LOC I .3( MEDIUM BOMBER) C N ACRAFT I .3MEDIUM BOMBERS) C N ACRAFT I ..MIRAGE II1) C N TJPE ACRAFT I .3
: NATIONAL GUARDS) C N INATIONAL GUARD) C N I .3
: ( NORTH OF) C PREP 3 .3: ( NORTHEAST OF) I PREP I ..: ORBITAL WORKSHOP) C N SATELLITE 3 .3
( UTSIDE OF) C PREP I .;C NORTHWEST OF) C PREP I .3
: ( ED SEA) ( N LOC I .;: SAUDI ARABIAN) I ADJ NATION I .;
: ( SEA OF CRISES) C N LOC 3 .;: ( SEYCHELLE ISLANDS) C N LOC 3 .I
: SEYCHELLE ISLAND CHAIN) C N LOC 3 .IC SMALL SCALE) C ADJ I .;: SOUTH AFRICAN) C ADJ NATION I .;( SOUTH OF) C PREP 3 .;( SOUTHEAST OF) C PREP I .;( SOUTHWEST OF) C PREP I .( SOVIET UNION) C N LOC NATION I .3( ST HELENA) C N LOC 3 .3( SPACE CENTER) C N LOC3 .I
• ( TEST CENTER) C N LOC I .3( TEST RANGE) C N LOC 1 .
:( C TOOK PLACE) C VB TENSED INTRANS I .3( TAKE PLACE) C VB TENSED INTRANS I .3( WEST OF) C PREP I .;
( HITE SEA) C N LOC I .3: A C ART I .;: A313 C N SUBNUM I .: AA C ADJ I •;: ABOUT C ADVP EVAL PPEDET I C PREP EMOD I .3
ACFT C N ACRAFT I .;
B-4
:2 ACHIEVE I VB TRANS I .32, ACTIVE E ADJ 3 C VB ACTVTY 3 .32: ACTIVITY C N EVENTIVE 3 .32: ACTY C N EVENTIVE 3 *;
ADDI'TIONAL C ADJ REF 3 .3ADX I N EVENTIVE 3 .3
ADZ t N LOC I .;2: AFRICAN t ADJ I.;:: AFTER C PREP TYME SCONJ 3
AGAINST I PREP EMOD I .3
AGENCY C N 3 .;AIR-TO-SURFACE I ADJ 3 .3
AIR C N 3 .;:: AIRBORNE I AJ I C VB FLIGHT 3 .
AIRCRAFT C N ACRAFT 3 .3
ALEXANDRIA E N LOC 3 .3ALL C QUANT 3 .;
• ALONG C PREP EMOD 3 .3: ALTITUDE C N ALT 3 .3:, ALTITUDES C N ALT 3 .3
AN C ART 3 .;AND C ONJ 3 .3
:: ANNOUNCED C VB TRANS TENSED THATCOMP COMM 3C VB TRANS PASTP THATCOMP COMM 3 .3
• APOLLO C N SATELLITE 3 .;." APPROXIMATELY C ADVB EVAL PREDET I .3:: APR C N TYME MO ] .I2: APRIL C N TYME MO ] .3. ARE C BE [ COPULA 3 C VB TRANS 3 .3
:: AREA C N LOC 3.;:2 AREAS C N LOC I=, ARRIVED C VB PASTP ARRIVE 3 C VB TENSED ARRIVE 3 .
AS C PRTCL I .;?2 ASSOCIATED C ADJ 3 .;
ASM C ADJ I2. ASTRONAUT C N I .3
ASW C ADJ 3 .3:2 AT C PREP EMOD TYME 3 .3
ATLANTIC C N LOC I .;•: ATMOSPHERE C N LOC 3 •
AUG C N TYME MO 3 .3
:: AUGUST C N TYME MO ] .I
AUSTRALIA C N LOC NATION I .1:: AUSTRALIAN I ADJ ] .I
AITYILIARY C ADJ 3 .3AVIATION C N EVENTIVE 3 .3
. A-4 C N TJPE ACRAFT 3 .3•: B-7r C N TJPE ACRAFT I .;.' B-753 C N TJPE ACRAFT I:B B-60 C N TJPE ACRAFT I ..
:: B-60'S C N TJPE ACRAFT I .;B B-61 C N TJPE ACRAFT I .3
B-5
:: B-63 C N TJPE ACRAFT 3 .B-63'S C N TJPE ACRAFT I .3B:: -635 r N TJPE ACRAFT I ..B-675 C N TJPE ACRAFT I .;
:: B-80 [ N TJPE ACRAFT I .;:: B-TYPE C N TJPE ACRAFT 3 .;:: B-TYPES [ N TJPE ACRAFT I .;Z: BACTERIOLOGICAL I ADd ] .;:: BAIKONUR E N LOC I .;:s BARFLY C N NATO ACRAFT I .:BASE C N LOC] .;
:: BASED C PASTP 3 .;
:: BC254 [ N SUBNUM I .;:: BE [ BE I I COPULA 3 t VB TRANS I .3:: BEACON I N NATO ACRAFT 3 .3:: BEAGLE C N NATO ACRAFT 3 .3•: BED C N 3 .;:: BEEN C BE PASTP I C COPULA PASTP 3 C VB TRANS PASTP 3 .3•: BEETLES C N I .;:: BEFORE I PREP TYME SCONJ ] .;
BEING C BE PRESP I C COPULA PRESP I C VB TRANS PRESP 3 .3:: BETWEEN C PREP EMOD TYME 3 .3•: BOMBER C N ACRAFT I:: BOMBERS C N ACRAFT 3 .3
BOOSTER C N BOOSTER I ..:. BORDER C N LOC] .;
BOUNDED I ADd ] .I:: BUFF C N NATO ACRAFT I .3•: BUJUMBURA C N LOC 3 .;:: BUTTER C N NATO ACRAFT I .3:: BY C PREP EMOD TYME I .3
CANADA C N LOC 3 .;:: CAPETOWN-BASED C ADO 3 .I:: CAPETOWN C N LOC I .;: CAPSULE C N I .3a: CARRYING C VB TRANS PRESP I .3a: CENTER C N LOC 3 .;: CENTRAL C ADO 3 .Ia: COAST C NLOC 3 .3.: COLLECTION C N EVENTIVE 3 .3:: COMBAT C N EVENTIVE 3 .3
a: COM73ATANT C N I .3:: COMBATANTS C N 3 .3a: COMMUNICATION C N 3 .3a: COMPLEX C N LOC 3 .;a: CONDUCTING C VB TRANS PRESP I .3a: CON'DlICTED C VB TRANS PASTP 3 C VB TRANS TENSED 3 .3a: CONFIRMED C VB TRANS PASTP CONFIRM I
C VB TRANS TENSED CONFIRM 3 .;a: CONGO E N LOC 3 .3a: CONTAINING C VB TRANS PRESP I *3a: CONTINUED C VB TPANS PASTP 3 C VB TRANS TENSED 3 .3
B-6
5CONTINUING I VB TRANS PRESP DIR CONTINUE I .2CONTINENTAL C ADJ I .:CONTROLLER I N I .:CONVERTED C ADJ I .:CORNER C N 3 .;2COSMODROME E N LOC 3 .:COSMONAUT t N].I:COSMONAUTS E N I .:COSMOS I N SATELLITE 3I.2COSMOS-605 C N SATELLITE I .:COSMOS-629 C N SATELLITE I .:COSMOS-706 C N SATELLITE 3 .:COSMOS-722 C N SATELLITE I .:CRAFT C N SATELLITE 3I.:CSM I N SATELLITE 3:CURRENTLY C ADVB REF TYME 1 0;:DAMASCUS C N LOC 3 .:DATE t N DAYTE 3 .:DAY C N DAYTE I .:DEC C N TYME MO 3.;:DECEMBER C N TYME MO I .:DEFENSIVE[C ADJ 3 .:DEGREE C N I .:DEGREES C N I .:DELTA-CLASS t ADd I .
:DEORBIT C N EVENTIVE DEORBIT 3 C VB DEORBIT 3 ..:DEORBITED C VB TRANS PAST7 DEORBIT I
I VB TRANS TENSED DEORBIT 3 .:DEPARTED C VB TRANS PASTP DEPART 3I:
C VB TRANS TENSED DEPART I .:DEPLOYED C VB TRANS PASTP DEPLOY 3
C VB TRANS TENSED DEPLOY I .:DEPLOYMENTS C N EVENTIVE I .-iDESTINATION C N 3 .
:DIVISION C N 3 .;:DJIBOUTI E N LOC I .:DOWNED C VB TRANS PASTP I C VB TRANS TENSED 3 .iDOWNRANGE t N LOC I .iDURING C PREP EMOD TYME I .
EAIG ERC~ C ADVb TYME RE .3D
:EAST C ADVB DIR 3 C ADJ I .3
EQUATOR C N LOC I o;ENROUTE C VS TOCOMP ENROUTE I e;
: ENTEBBE E N LOC I .3: EQUIPMENT E N 3 .I: ETBF C N LOC I .3
t ETHIOPIAN C ADJ 3 .I: EXERCISE C N EVENTIVE 3 .3: EXPECTED C VB TRANS TENSED TOCOMP EVAL ROBJ I
C VB TRANS PASTP TOCOMP EVAL ROBJ 3 .3
:: F-4 C N TJPE ACRAFT I: F-4E C N TJPE ACRAFT 3 .3
:: F-SE I N TJPE ACRAFT I .3: FAILED I VB INTRANS TENSED TOCOMP EVAL RSUBJ I
C VB INTRANS PASTP TOCOMP EVAL RSUBJ I .3:: FEB C N TYME MO] .;:: FEBRUARY C N TYME MO 3.°:: FERRY C N I .s FIGHTER-BOMBERS C N ACRAFT 3 .3:: FIGHTER C N ACRAFT 3 •:: FIGHTERS I N ACRAFT 3 .3is FIRED C VB TRANS PASTP LAUNCH I I VB TRANS TENSED LAUNCH 3 .I:s FLEET E N I .353 FLEW C VB TRANS TENSED DIR FLIGHT I .3:: FLIGHT C N EVENTIVE FLIGHT 1 .;:: FLIGHTS C N EVENTIVE FLIGHT 3 .3is FLOGGER C N NATO ACRAFT 3 .3i FODDER C N NATO ACRAFT I °,is FOLLOWING I SCONJ I .3is FOR C PREP I .;:s FOUR C N NUM 3 •:: FRESCO C N NATO ACRAFT 3.3i FROM C PREP EMOD TYME 3 .3is GENERAL C ADJ I .is GROUP C N I .3:: GUAM C N LOC I .3:5 GULU-BASED C ADJ 3 °:s GULU C N LOC I ;:: HAD C HAVE I [ VB TRANS PASTP I C VB TRANS TENSED I .353 HAIFA t N LOC I .3is HAS C HAVE ] C VB TRANS TENSED ].:: HAVE I HAVE I C VB TRANS I .3-: HAVING C VS TRANS PRESP I .3: HEADING C VB PRESP DIR HEAD 3 .153 HERMETICALLY C ADVB 3 .Iis HOMEBASE C N LOC I .,;s HOUR C N TYME UNIT I .;
i HOURS C N TYME UNIT I .:is HR C N TYME UNIT I .;,: IL-28 C N TJPE ACRAFT I .;
s: IMPACT C N EVENTIVE IMPACT I C VB INTRANS IMPACT 3 .3i IMPACTED C VS INTRANS TENSED IMPACT I .3:: IN C PREP EMOD I .;s INCLINATION C N INCL 3 .3
B-8
-wo
Ir
i-t INDEPENDENT E AD) 2 .Iii INDICATES. E VB TRANS THATCOMP I .3ii INFORMED E VB TRANS PASTP THATCOMP COMM I
E VB TRANS TENSED THATCOMP COMM I *Iit INITIAL C ADJ 2 .I
INTELLIGENCE C N I .3
a: INTO C PREP EMOD I .3
t INVOLVING C VB TRANS PRESP 3 .3:t IRANIAN I ADJ NATION I .3
IS I BE ] E COPULA I I Vb TRANS 3.1it ISRAEL C N LOC NATION I .;a: ISRAELI C ADJ NATION I o;: JAN.[ N TYME MO I .3
JANUARY C N TYME MO I .3it JUBA C N LOC 2 .;: JUN C N TYME MO 2 .Iit JUNE C N TYME MO 3 .3a: JUL C N TYME MO 2 .3
JULY t N TYME MO 2 .3a: JUST C ADVB PMOD 2 .;:2 KALGOORLIE I N LOC 2 .3a: KATHMANDU C N LOC 3 .;it KB252 C N SUBNUM 2 .3:: KE843 C N SUBNUM, I°;it KENNEDY C N 3 .;i: KENYA C N LOC NATION 2 .: KENYAN C ADJ NATION 2 .3
KFIR I N TJPE ACRAFT 2 .3t: KILOMETERS C N LOC UNIT 2 .3
KINSHASA C N LOC 2 .3a: KM C N LOC UNIT 2 .I:a KMS C N LOC UNIT 2 .3
LACCADIVES C N LOC 2 .3
it LANDED C VB PASTP LAND 2 C VB TENSED LAND 2 e;it LANDMASS C N LOC I .3t: LAST C AD, 2 .I
LATE C ADJ TYME 2.3
t LATER C ADJ TYME I .a: LAUNCH C N EVENTIVE LAUNCH 2 .3a: LAUNCHED C VB TRANS PASTP LAUNCH 2
C VB TRANS TENSED LAUNCH 2 .3ai LAUNCHER C N BOOSTER 2 .;:: LEBANON C N LOC NATION I .3it LIBYAN I ADJ NATION 3 .3i: LIVING C ADJ 2 .;i: LOCATED C ADJ VB TRANS PASTP LOCATE 2
C VB TRANS TENSED LOCATE 2 .3:: LUNA-23 C N SATELLITE 2 .3ist LUNAR C ADJ 2 o;i: MALAGASY C N LOC I .;it MANEUVERABLE C ADJ 2 .
it MANNED C ADJ 2 .I
B-9
MANY C QUANT I .3it MAR I N TYME MO.] *,: MARCH 'C N TYME ..,0 3 .1:: MARITIME t ADJ 1 .Iit MASSAWA I N LOC I .3i: MAURITIUS C N LOC I .3it MAY I MODAL I C N TYME MO 3 .3:: METERS C N LOC UNIT 3 .Ii: MID I ADJ .3it "!G-17 I N TJPE ACRAFT I .3: KG-21 I N TJPE ACRAFT I .3:: MIG-23 I N TJPE ACRAFT I .3:: MILES I N LOC UNIT 3 .3it MILITARY C ADJ I .3it MINUTE C N TYME UNIT) .3it MINUTES I N TYME UNIT 3 .3it MISSILE C N MISSILE 3 .3: MISSILES I N MISSILE 3 .3is MISSION C N EVENTIVE 3 ,3:: ML-28 C N TJPE ACRAFT 3 .3:: MODIFIED I ADJ I .;it MODULE C N SATELLITE 3 .3it MOGADISHU [ N LOC 3 .;:: MOMBASA-BASED I iDd 3 .3:2 MOMBASA C N LOCJ .Iit MONTH C N TYME I.3it MOON C N LOC 3 .;:: MORNING C N TYME 3 .;it MOST C QUANT 3 .3it MUSHROOM C N I .3it NAIROBI-BASED C ADd 3 .3it NAIROBI C N LOC I .3it NASA C N NATION 3 ,3t: NATIONAL C ADJ I.;it NATURE C N 3 .3it NAUTICAL I AD. 3 •it NAVIGATIONAL C ADJ 3 .3i NEAR C PREP EMOD 1 .3it NiEPAL C N LOC NATION 3 .;it NEWS C N I .3Ii NIGERIAN C ADd NATION I .3it NINE C N NUM 3 .;is NM I N LOC UNIT 3 3it NMS C N LOC UNIT I .3it NO C QUANT 3 .3$t NORMAL C ADJ I .is NORTH C ADVB DIR I C ADd 3 .Iit NORTHEAST C ADVB DIR I C ADJ 3 .3is NORTHERN C ADd 3 .ts NORTHWEST C ADVB' DIR 3 C ADJ I .t NORTHWESTERN t ADJ 3 .1
B-10
4.
2t NOTED I VB TRANS PASTP OBSERVE I C VB TRANS TENSED'OBSERVE I .3iZ NOV I N TYME MO ) .;:: NOVEMBER C N TYME MO I ,is NYANJA I N LOC I .;it OCCUR t VB INTRANS 3 .1:: OCCURRED C VB TENSED INTRANS 3 .3s: OCEAN C N LOC I .;t: OCT C N TYME MO Iis OCTOBER C N TYKE MO 2 .Iis OF C PREP I .;
it ON C PREP EMOD DAYTE 2 .3:: ONE C N NUM 2 .;:: OPEN C ADJ 2 .2.: OPERATED C VB PASTP ACTVTY I C VB TENSED ACTVTY I .i
: OPERATING C VB PRESP ACTVTY I .3t: OPERATIONS C N EVENTIVE 2 .3:: ORBIT C N LOC 2 .;it ORBITAL C ADd I .: OUTBACK E N LOC 2 .3
2 OVER C PREP EMOD I .,22 PACIFIC C N LOC 3 .3• PENETRATED C VB TRANS PASTP PENETRATE I
E VB TRANS TENSED PENETRATE 2 .32: PENETRATION r N EVENTIVE 2 .;:t PERFORMING C VB TRANS PRESP ACTVTY 2 .3:: PERIOD C N TYME I .,:: PERTH.C N LOC 2 .;:s PHANTOM t N NATO ACRAFT 2 .I•: PILOTS C N 2 .;:: PLESETSK C N LOC 2:t POINT C N LOC 2 .3• PORTION C N 2 .;22 POSSIBLE C ADJ EVAL 2 .I2 POSSIBLY C ADVB EVAL 2 .3$2 PRECEDED C VB TRANS PASTP BEFORE 2 1 VB TRANS TENSED
BEFORE I .3it PRECEDES C VB TRANS TENSED BEFORE 2 .I
2 PRESENTLY C ADVB REF TYME 2 .3
: PRETORIA-BASED C ADJ 2 o;2 PRETORIA C N LOC 2 .;$. PREVIOUS C AD REF TYME 2 .;it PREVIOUSLY C ADVB REF TYME 2 •ul PRIMARILY C ADVB EVAL 2 .3
PROBABLE C ADJ EVAL 2 .3PROBABLY C ADVB EVAL PMOD 2 .3
2$PROCEEDING C VB PRESP DIR 2 o3RATS C N 2 .3RECONNAISSANCE C N EVENTIVE 2 .IRECOVERABLE C ADd 2 .I
is RECOVERY C N 2 .;2$ REENTER C VB TRANS REENTRY 2 o3so RE-ENTERED C VB TRANS TENSED REENTRY I .3
B-11
K REENTERED C VB TRANS TENSED REENTRY 1 .3
t REENTRY I N EVENTIVE REENTRY 3 .a: REGT C N 1 .3a: REGIMENT E N I .3a: REGIMENTS I N I .as REGION I N LOC I .3a: REMAIN. COPULA TENSED I .3t: REMAINED C COPULA TENSED I .3i: REPORT C N I .;a: REPORTING I VB TRANS PRESP 3 o;: REPRESENTING C VS TRANS PRESP 1 .3sa RETURNED C VB PASTP RETURN I C VB TENSED RETURN I .3a: RETURNING C VB PRESP RETURN 3 .3a: REVOLUTION C N REV I .;:: RGT C N I ,;:: RIYADH C N LOC I .3as ROUTINE C ADJ I .3:a ROUTINELY C ADVB I .3a: S1234B C N SUBNUM 3 .I::' SA554 C N SUBNUM 3 ,3a: SA622 C N SUBNUM I ..
i: SAFAF C N NATION SERVICE 3 .3a: SAFLT C N NATION SERVICE I .3:t SATELLITE I N SATELLITE I .3=s SATELLITES C N SATELLITE I *3. SATURN-IB C N BOOSTER I .;
it SATURN-5 C N BOOSTER I .3as SAM-3 C N MISSILE I .3a: SAME C ADJ 3 .3:: SC462 C N SUBNUM I .3it SCRAMBLED C ADJ I ,:: SEALED C ADd I .;:: SEP C N TYME MO 3 .3:: SEPTEMBER C N TYME MO ] .Ita SEYCHELLES C N LOC3 .3
it SIBERIA C N LOC I .;a: SIMULATED C ADJ .3 ;a: SINCE C PREP EMOD TYME I .,a: SIWAH C N LOC] .I
i: SIX C N NUM I .;it SKYHAWK C N NATO ACRAFT 1 .3t: SKYLAB C N SATELLITE I ,$a: SOFTLANDED t VB TRANS PASTP LAND I C VB TRANS TENSED LAND I ,3as SOMALIA C N LOC NATION I .3a: SOME I OUANT 3 .;as SOUTH C ADVB DIR I C ADJ 3 .Ii: SOUTHERN C ADd I .;
a: SOUTHEAST C ADVB DIR I C ADd 3 e3a: SOUTHWEST C ADVB DIR I C ADd I .3t: SOUTHWESTERN C ADd 3 .i: SOVIET C ADd NATION I .3it SOYUZ C N SATELLITE I .3
B-12
C _ A_. .. t. ,kz .. .. . . .
SOYU.-22 C N SATELLITE-] .3SOYUZ-28 C N SATELLITE 3 .3
s SOYUZ-TYPE I ADJ SATELLITE IISP2 65 [ N SUBNUM 3 .3
is SPACE E N.LOC 3 .3s5 SPACECRAFT I N SATELLITE I .3i: SPACEFLIGHT I N FLIGHT I .3it SPORES C N I .3s SR-71 C N TJPE ACRAFT Is.3is SS-I C N MISSILE I .3it S-4B C N I .3it STAGE C N3 .
is STAGING C VB PRESP STAGE I .3i: STATEMENT C N 3 .;:: STATION C N LOC 2 .3ts STRATEGIC C ADJ I .tt STRIKE C N EVENTIVE I .3it STRIKES C N EVENTIVE I •it SUAM C N LOC 3 .3it SUBMARINE C N I .j55 SUBORDINATE C ADJ 3 .3:: SUCCESSFULLY E ADVB EVAL 1 .3it SUDAN E N LOC 3 .3:: SUDANESE C ADJ Iit SUGGESTS C VB TRANS THATCOMP 3 .it SUPPORT C N EVENTIVE 3 .;:: SURFACE-TO-AIR I ADJ 3 -1it SURFACE t N LOC 3 .;:: SURGUT I N LOC I .3it SURVEILLANCE E N EVENTIVE I .3is SYRIAN C ADJ NATION 3.3
:: TAIPEI C N LOC 3 .;TAIWAN C N LOC NATION. 3 .3
:: TASK C N EVENTIVE 3 .3TASS C N I .3
TEN C N NUM 3.;is TESTING C VB TRANS PRESP 3 .3it THAT C CONJ I C RELPRO 3 C ART REF 3 0;tI THE C ART 3 .;is THESE C ART REF I.3is THEY C PRO 3 .;ss THEIR C ART POSPRO 3 .3it THIRD C ORD ADJ I .;
Is THIS C ART REF DEMONS I .3is THOSE C ART REF 3 .3
is THREE C N NUM 1 .is TIME C N TYME 3 .;
:t TO C PREP EMOD 3 .;
it TOBRUK C N LOC3 .3it TODAY C ADVB REF TYME VMOD DAYTE 3 .3: TORORO C N LOC 3 .Iit TORTOISES I N 3 ,
B-13
aTRACKING C ADJ I .:TRAINING C N 3I.aTURNED t VB PASTP DIR 3 C VB TENSED DIR I o;aTURNING C VE PRESP DIR I .:TU-95 C N TJPE ACRAFT 3I.:TWO C N NUM I .:TYPE E N 3 .;:TYRE C N LOC I .:TYURATAM C N LOC I:U-43 C N TJPE ACRAFT'3.aU1009B I N SUBNUM I .aU1211B C N SUBNUM I:U1232 C N SUBNUM 3 .aU1324B I N SUBNUM 3a:UABC C N 3I.aUAF C N Ia:UBBC C N I .:UG254 C N SUBNUM I .aUGS36C C N SUBNUM 3I.aUGANDA C N LOC NATION).I::UGANDAN C AD) NATION 3 .aUNDERWAY t VB FLIGHT I .aUNDETERMINED C AD) IaUNIDENTIFIED[C AD) 3IaUNITS E N I .;a:VARIOUS-C AD) 3 .aVEHICLE C N SATELLITE MISSILE 3I.:VICI11ITY C N 3 .;:VIOLATED-[ VB TRANS PASTP PENETRATE I
C VE TRANS TENSED PENETRATE I .aWAS C BE 3 C COPULA I C 'JB TRANS 3 .:WEATHERCEN3 IaWERE C BE 3 C COPULA I C VB TRANS 3 .aWEST C ADVB DIR I C AD) IaWESTERN C AD) I .;aWESTWARD C ADVB DIR 3I.aWHICH C RELPRO PRO 3 .:WOULD C MODAL I .aX C N LOC I .a:XB4a42 CI N SUBNUM I .aXB262 C N SUBNUM 3 .aYEAR C N DAYTE 3 .;a:ZEILA E N LOC 3- .j
ENDLEXMC R'BYES
B-14
........1.Jim,
APPENDIX C: Templates and Auxiliary
Procedures.
c-i-C-1
...........
* d. w
0- di cu 0
i- t' U0 I. 0IxE0 Etu- WU 0dalE 0
F-' C w 4 U a
<E Z U I 0 0C-. .C. - - .C r -ci u.W C +) V .Ix+. a 08 (C JCk: - w 3-. 0) 3i I . ~
MLu C - 4.) i *J . 0.
i uIW w a -- - ~ 0i ixV *a) I U N Ix -0 i Ei .0 m4 CDdi 0)d C C W-'<1 c 0 --i -Z I--+) 31. Ix 0). . )<. -s i 0x U- -0 Z.1+ 4)V-4 ) (D - O . +;t E .- 4 0 ..
cc Z~ C -~J "'- w '-) 0-a. ) > 1>X"-4 -' d '-)- fl~ di L >rzC .0~-a di G
C:L 0-0 -4 - m .,I . -q '4N . WCC 4O.Q fl--4t' M 0 f- r'- 0 a'- m C
toW . . . 4- 0 0j 4T t-CJZ Ze E - : C
-l c +w-i, n+ -'- :1 >+ U) 00Z - 4 1
c ai X r..-C.. iV LDd -), 9 i di ' I n di
m ~ ~~~~~ ~ ~ -2-Z- > + ( )N-ma
r4 XI
0 '0 0 U1.0 U.
I flO 0- :> I
U. D. 0 sO04 -0 -
ftU-r '- CJ1ft00 -P J- 01.00f.- C404~~0 Ln >0-40
00 04 00"F- V-4 0 -a . -:
i z E 0'a C-3=
z zw wi
w +)4-
.. -> ... .I *T U)
w -' 01 C-4 0
U) I ~ -j IX
01 w -4 a) 1-'a
CO:r 0 a" ajU
4- pC4u H. a l 3U -4- .'a] -
0 _0ti C- w. '0 m ro O' + Nu L Lw- W _j- C _ LL "- w)
4i 0' N' -' Z-'1i 0 (Fiji *-' CiDi m0 H'0 -40- 0' m' .Ja 174 4-) c xU 4=>-j ci0 *-c a*jrC w -
*1w-; 0 1 01 .iD" VI inU t C*4 <
174 -- E~ C0 .. I I-r >O'i E.
- Z -0 CD f- 4-) W 4-4 t-l 1-4 Z- -f~
0al4 0 DW0IZ -CJ IL-j -aI0Lf '-t L mX-a I.4- ;- ( -En W -i-C4 -1 W 4 C 0 3
6T4 M-(n C U CZ f- zCICI W (n >CwJ- oi (a (D :3U
LJ.a> o D txJ +) -4' . )'I-- sc- LLt -a -0 X I0LiO., .- 7 1 4 ' alW1 O-i0 +) ". - Dii r D - i- w ~ i
9-'r 0n a4'01U') C - -'1U -O.. W +) Cocx ICOIai -I cj 00- u U)a u CJ U2 U o- % -4
=0 a) m- -4 0 ME(Dr4 E0)14DM
Uc + -c al I-c - (5 mD > 11 . w (D 4- m 0 4. Dt )( % . 4% w c .r c U f:
f-i - U 4. u U U U LI) --
4- D D Z 4 4 NN a
0
1 09-4
0)~~~S Dr- - o
Q M- Ln f-ri n.. I F- D 1
- U -1-. 0
(n 0 0) 0)
ci C. .-4i M . c-1 ..-Z -. : LL l -oA - 4. 0 .-- z r c +) r 1 -
C +) .4u + * ) 0 + 0 -1C- 11 0 +). z9 0 ' c II.-
U)> n : En Qi )I "9 iI a 4)
-~- -~ Q' 4-) C' C H-0 +.- L3 l (1) c 0 0
It ') .. -4 -4 x. E)L I -
+) n) 0 1- C- 4)J 0 AIC:3Z ft r-1 n Mm (
0i w1 .. '4- I- U)W0 ' 00 w 0Go " ' lm _j X: v*4 r- i4 C
C 4. Z II th ri M'- *)M_ r-I 0 l'-
ci o -*'iC ' .. .(V w
..4 -) r-l *'' > C; 4)J ri- Q' .. wC 4 1) -' 0 -4. 0l % 0
rq 0 w M 4-) +) 41- 4-) 0 J 4.) :<' 0 0-, n s * ) -
_j 0 - 1.) D M -to M1 - .*- ' M
+)+) I-4 -0' tU) w 1 () C q 174 4) 11J 4)<t
w IJ t't I+. c - -. 4 C' -J I) tz W I -4 C C ir I iL-M %--im 4 ) 4 .L 01 ) 0 Go M - 0 C- .4 0 J(% C IW
-i Ir " Or' U) 0 n ). CO .E - 1 4 -9 .J3 LLX -I 01 9-4 4- C (- )I n 7 tn 6*U CA- , -4-) cc- I)> CI r4 t to ., -i- r. m. 01 CL' 0 a
4J% s ~ -ir ..4 to.' In4) LA '-9 CL 4) ., .I) a)(-IC C - - 0 c c c J0 :+)-p j to -C
do III & u 11- %..U)D A O .'VD 41) 0.t0to0g 3 o
CZM .mAe- IN.4) c I m m " m...m.- fil 0)- ,.b . 1 o 4C' to.- S.J->S.0 S.S. I -"Ci .4) >- io -ia 3
-- OLA 4)-.4-3C~i3 I -5 ) - L . S.1-L .
LLI ra
z L
M u
n 0 --q
*S En Li
C. Is- LnU Z 0 1W = w Z cr
aU)- m J a, 3 a1 H w -0 a
m IuL rC< a I LL '. az .. En 0 r.0
.4.W) 1 x 0 X 0... 40 Ce-4Cfl.w)u0
a ) 0 co rfl 0. .Q. -a. .- ) ws wI w~. a ls XJ Z 0 im l.
+) X --- W~ ML' m- _3 ,
.. .4) 10
_j~4 3 a)(2 ~0 0 D ka '.4 -. 4 +),.+ . C ma C 0) CL,- 0.s Z .. _j -l a- .
+. ~ ) ~) I-0 +) 4-. 4- . 4 I - _4U.0 0~ L'2L I.4. C ' r03 WU WC.J~- u m C C L - U S - - Z Xm IH '-. 4- -4 H
2: +C. +).)'- CD 00 "J - &CL a 00 .0 0O-U US0 0 0LtE -4 k 'a I 4 )' D'
c Zi1.-1 .4 -4 C .Q L0 -4 -4 En) I0 - -- 4 r-4 -4- -> > J L) u- M I-4 4 4 HZ .C C zI~ CC. ., , lW ZZ 4) 4-)
+) 4N) U Cu u U * -) a.0. tiu u u u um mm (- 9-f- 9- 9- m m S C0 Ls asU u u m mrm m- M) -1 4 4.i )0 000 0 300 w20 1w0 alt a) al w m
.4~t to4U toSU US-USS '- 0 0 0
C-6
4-)
U) 0 W1
a 0a
r* :) 1 CL m
Z 0 :) 0 - -4c e(f
0' D~l4Z I O- m fl . 0- w aIO *.l- a - 4-). '0 2:
.C. Z -1 LL z 0 0 0 U1~ 014DI- ,-4,,-p4CL .J r- .(D - - .11 .11 -- - D_
o 0 m m CA- C- 01 (DM P j+ al - a- v ) _
> 0 CU U wZP - - ILZ4.) - M - *.I + 1 -) M 1
Li C l l 14 .~ -s to' M -a) I*0)q1.rf - "C j
.4 r 4 0 r 01 "-4 0. 0. 0. a# Z ) 1 f 1-4 -.- -L tftfl4f4 IL 0 Cf CXCL_ jCL W+ D
C 4- - k' -4 0.C C 0 -4 -H5 La- -1 01- u W E -jm-
C &o r- w C4WZ Z C CJ I J - I -0 % 4, 4,& 4. i 4 (- E n( .- +-10 1 '4CLW( 4WW ,jr 4-) 4,) 4A 04, 0 , >-- LL X M P
> - C- +) 6 t 0 -4 0r-4 ) rI 0 Z
+) I al cI 4WrI l I A 1 ic
a L)
C W
Io '-4 t
41 ~ - 0.
CJ LL 0 'rp- 4-)t
I-o-j ce Li 4 t
I W. Qs Z .4 a) .(D f- CLW Si. 3'- I 4-F. CL z CL 0
CiL - 1 a1 ME .,a 5 1 V 3 F-a *S I Si
(Da~ *.4 Si .- 14 * Z 11 aL~j+)Sij r- C Si 0 -Si r cu .9)
Ln i- C.4 C 11D 0 ~ C0 4CL E M%
ZU1_ In-W 3SZ I
+) L l a .. ) C-O aI (D -j o
a - t . . * a x X . a I- .. ) a
CU43- CU ~ - i I s L. I. JL z za3ci54 C.- W e--'i al :3 -. a) ZJ a 3W -. I
4. CZC*. (LL- LL W'E w>---i. W W )( C0i~4 aZL ) z r U- Si II- -a -0.43Si-
WL 0.3 %~-a4. % 0 '0103-a 5- >& i
IX aL-a I -C 4j _jW~ Z 6M+ Icc'
- LX 9 a L.'aCLIf C Si w a1) M 00-0.'- L43WmLL4ISM LA C U-MIc :
5-).5-SW I a ~ C4 :3 3. kI *. a i0
M .5 -Si 3') LAa IS D -i I LX Ix r0n a e)
- 6~ j- .1 . )4 .0 +.) )CL a c aa CC. iW i CD ~ lJE ii (D - A- =mt 3 iC a i 0) I -'-43Wa- E +) '-SiiW- (D D Q-- -s - Li Li '0 S..
-0 +) E 3-P E4- 4-4 4 3 + 4- aa ?Do Io 0 0 0 Si C c )x SiQ C i
r-I '- -4 -1 Si 1 1 '- - zC r Si0f 4- 4 0 t..- f C Cic
II II I I I Si i I 1 0
'.4 -l C) C wIlI C
C-8
p.-.
w
w Q rw
LL a t -%
EUssU
0-j a) Ix .
CL) 0) Ixw.a
M aw xa
CEI4 a -
Us w- .. a
s co.. a. LL 1 64 U 'cX I I LlJ.PaCfts (D - I L (DCDc
+)+) f a , ) Lu . W U(D m a..-C w T a. II1Wx*I 4.) Ix ) Of.
cc+)~ *.a *. L 1 a4- l .- f)~a 4-)w)T 4) n < (
a~ ~ <.. +. -~ .. a r
4j .. a) - x W aj a) Z
I~ ..x X ~ QJI I IXOJ ) ( ..- a a 1 '- I. a. X. a$ :n 31 z
-'P ~ 31 (p .. i Ix4) a. Dft >t
I w I +) 1 11 E
aaaaaaaaaaaa. IX U 4 1 a.
CC-
APPENDIX: D FSA Listing
D-1
APPENDIX: D
PATTERN NO:S NC
:A 0123456789 => Ml so
:S NItA 0123456789 => N2 .. ,
:A S ORDS P itA N -~ORDN so:A R -~ORDR Ps$A T 0 ORDT s it F * BLANK* C N NUM I DIG I sotF -,/I C N NUM IDIG I s
:S N2tA 0123456789 -> N3 ,,
:A S *~N2S soiA N uN2N so:A R all ORDR sosA T -2- ORDT so:F *BLANK* C N NUN 2DIG I 3:F -,/: C N NUN 2DIG 3 1
:S N3IA 0123456789 a,> N4 sotA S ORDS so:A N i'ORDN ,
tA R -~ORDR ,tA T a:, ORDT .tF *BLANK* C N NUM 3DIG 3 o:F -*1: C N NUM 3DIG 3 so
tS N4SA 0123456789 u), N5 soSA Z a'TMF sotA S a' WS so:A N a'NAN *,IA Ru ORDR so$A T => ORDT so:7 *BLANK* I N NUN 4DZG I s:7 ,/ C N NUN ADIG 3
D -2
ISN$A 0123456789 a N6 asA S =* ORDS asA N aJORONsA R *~ORDR a$A T *~ORDT asF *BLANK* I N NUM 5DIG I a:F -&/: E N NUN 5DIG I op
tS N6:A 0123456789 NF NFIA Z =D- TMF a:A S n2- N6S asA N *~N6NsA R *~ORDR aIA T 0 ORDT aOF *BLANK* C N NUN 6DIG361.sF -*/tIC N NUM 6016 3 s
tS.NFsA 9123456789 -1 NF P
:A S -:P ORDS osA N OR OR Ps
:IA R a'ORDR a.tA T *~ORDT .:F *BLANK* I N NUN I JSF -#/: C N NUN I a
tS ThFF7 *BLANK* C N TYNE I a
tF -*/I I N TYME Ijo
tS ORDS$A T * ORDF .
3)
tS ORDN ORF,
D-3
sS ORDR:A D *~ORDF
:S ORDT:A H -~ORDF
tS ORDFiF *BLANK* C NUM ORD 3IF -.,/t C NUM ORD I o
IS N2SsA T -3- ORDFiA 0123456789 u~2LN1
IS N2N:A D -3- ORDF:A 0123456759 *~2LNI
IS 2LNIsA 0123456759 =3- 2LN2 s
IS 2LN2tA EW -~2LF
iS 2LFIF *SLANKl* I N L.OC IIF ms/I C N LOC I
IS wASsA T -x- ORD?sA 0123456789 m*o ALNI ~
D-4
iS N4NsA D -~ORDFsA 0123456789 4 LNI ,
IS 4LNIsA 0123456759 *~4LN2
IS ALN2tA 0123456759 4 LN3 ,
IS ALN3:A 0123456789 4 LNA ,
IS 4LN4sA EW -o4LF
IS ALTaF *BLANK* C N LOC IIF -*/I I N LOC I
sA T -x. ORDFtA 9123456789 *~6LNI ,,
IS X6UsA D *,, OR DisA 6123456789 w*o 6LNI .
IS 6LNI:A 6123456789 030 6LN2
D-5
IS 6LN2tA 0123456789 *~61J3 .
IS 6LN3$A 8123456789 *~6LN4
:S 6LN4IA 9123456789 *~6LN5 ,
:S 6LM5sA 0123456789 *~61246
IS 612&6SA EV w~l 6LF a
IS 6LFIF *BLANKl* E N LOC I ,IF -o11 C N LOC I
ENDPATTIERN
D-6
APPENDIX E: Examples of System
Input/ Output
E-1
-UJ
z
LO r.i r- L D
Li.. i
, - . i
U C7 Z i - Cr,-L
Dcf: U < 'CU 0' O o L- 41
C M U)1 II C O0C
Z~
-n i:Qi iz Z
m' :-c zz wL L~iCE 'ZL CL -'Z4Ii C C
LLJ~ - >- : . IU
7,s1- C ~cC c 3-I4 dIU
L~~j CI, Cu. aCxul ~Z s~-s .Ldc~ 5 s CfW 3E z N') Li64L.
z; f.-O flC OJ 0 LWY) 0 X -QU ~ I ;E : j <1 ly-
c.j -Zf zC OC")f C L) C.P. a rr, cr d .1i U :- =i; -a --a m z U) W 4. t
>-c.LZ 0WC< -.- i.L -- H .5 *1 C C a A C 0 CO
E .-2
ML'
C
t zCCz C
CL L d C. W z L L
CD u 0 M: 0 L T .
m Cu ml- L
i:Cj -U L) ii, L I OCj x L)U)
o CL N 0 Ua =1 l I) - 0 CL W) O
En £fl U) 3 0 :3 cWO 0CQm LU Z: m LU N
-j I >- >- C ZOj- - C. ij'.'
U) U-1 re) i i4- WU, m * r .1 LUN m'- 2 --I z mN 03 z- m 0),:Q ~ r 1:4 z U) Z -4 C-4 b-) i0 Z' CNj
,Y U .1U I ~ I- x I I LUJ I0 ' Z 0 )-Z 0 >
< =I 7-(, C Z:-N w <r m ~N Ld
- C, u) <1-C0< D,) C -1 < NULU> :3 Q-4 Ill >- M 4 UWIl>-X:-:)
L)r U - e z " 1=4 I -)..4fl rWC"
~~.'4 0x < jU I LU.) 0 i Uc 0 <l ' I4 U a u x 0 ' u Lm H W UZ -j I! WO 1H:4 -- L. LLU i- Cj o W~ c w -I
u ) i m * Ir'J-~ .~ U *LE+' =4W- i C uz w) * f 0Uz _ J c (jI m *~4TmD zron C w*It C 4T m >ur c * I L V
M < ) CO( I - 11 _" ~ L C al H J- W U J > a! < nra f5
C L C E -4 P q L n c e ..-q .) L () e c--4-*
2:
Li 2
Lii
ci OD coN" ,-i rN If)
z i ml '-I
0 CO Lii mN0
ui C'fl0 WLil) -i
'OWC NW
Cr -. -A-cH
LC x Z;7-~ ~ ~ ~ L3ZL) L - r 1::, -
uZD a; ) Zjr-cL W ::
-4 C I~ M F- .-. SU)AC u
U (j &U 5 Z iCj Li
U)a+'2.J U
E-4
LUC i4
C
Orr w Cx 0O C WUi L
<r D. zU~ i~-O'0 C w
a41 j N L) 1-4 Mc 3 zCE Os x N LiW -
XrC- WO' W 13, d C>z =;i- D. -4 .*W CcIzj. C-
(n- I- l-W w 0 I-- (n)C <:m j 1-0 cIEnl'- < ~ u =)_ UL >- W -il-- C (n Ji U. Z
a: UC Z1 j- Lw.. " .J. -j z n )z c 0 0-:1C Ix cc- 'j I-C. L- cC C 0 C-.
C C Z 0 CfC oc ;E cic 0 i I- i-U) CLI Ix CJI0 LId .J Lj I- Z WI- U) U) CLL 1: W-qfL :30 cJ r L 0UL EnC u0mc Cm C
L) 0' - _JI ED 1 C 00Z M I 17-w 0 CE w LICE r- ccnZC a z -i CC -j *) C -4 m CL IC 4x i
CL 0 CL- LI 0 1-'C < Oz a CL U) I- Z -q 4Cz x 0 :3 :4 zO- 0 "'- " C _CJ 2: z:-
LL =W La .ICi 174 ' OCC %4 = ) ce C LI2 0 C Z IxM 0 Z'I"iC w L WLJN -C 0 OW 0 ' C 0,r,! ZC CWX&' 0
WI, - CX J JW YC *0 C. r- 0 6r. i-- WOW0 LI - U) * I- -N,mW CCx - xC C I-Ccc-o I-- C Lor- Z-- L::z LIOa, - U) z U) 0
-j 0 En1740 U0M1-- 0 O LZ CI: ,L- :3N ) 1--n4LI -41-1 3 w c1:4c u w 0 - M LO --t CC0 ='O L) 03)y = i.- LDLW hTiL) C I- C: C LIJCL u Z-4.) n )
xC JCU--IO ZI- w J 3 Z -i *I - C U W 0 i-i ZO.ZcC Ci-CL% ) m -C Z *)- i- u 01m >CN ,X,-i M
z c0 -IL C0 XI- X CLd k- X4 -LdC' I- m ~ .W I' *) iw ol<a_ cI C-A-- CL: CCC "E X0~-4 )CIC C - >- C :3 Of O4 Z
C4 W -LE z-j0 L L-J- L UJ1 - 0C E .L -1 U CU< 0 - - ICLiI cc 1 0 'L->- Z -*e- UCc: L C>- X X4 I-4 Z &T40 )J W>- xcCCCI-
CL L = 0 I L .- LI-i Z -,X L'CL 0 0 X4leLIC< - 0 C - M:z 3 -j L LiC'NEZz0 IW-jU)M-N 3D M0 I-U)OCL CIx w CQ LO -C 0[ CL I
M 11 X LO j Lf, .Jc U) o'0U, ) C -M c -I- U WNU) oN in< r4 Z n0 4C C UC0C U L Z)UU Z J U0 LLC- =i"0 CWLL - Zi C r-z
*Z - iqW 1- -- 0 0-4 CC 0030)oaL jJxC0 '4 30 0-Z C >-W1U -M:MXCrZCL Z uUO InClmC LI W- CC
IU) CW Imz w w cC Cl C CLLILIC<CL W wIC CL zW aZ LL 0 LCL r-C
Z CI -I - ZI -I-- - -' -- " - Z EI- i41- 0 LI- 1- I--u) 0 U ) =1IO) U) CC- ix LI U U) L.) InUj)
E-5
I-j
C L)U) z
LJ
LLU
I LC, Z C0; C - La X
tL-' Cl L C
CLO Lo z -z u % a N =) 1
-j' ITI.-x ) 1;" cmIlz 10 w 1w 0
Cc, 3 * 3 Cii:- oLU LO -j 4 -LL.,LC C 4x a<1X~)-~ LU>-U. c
'*'-:~C 0
C- C U > ~
Li> <rC~
U) C U
L- a- 2~ C- E-c - Nm CL ME izmC CLE>
M I .* i4fM 0 U) 1. .40 M )0C OC L 0 j ; -I
x ~ ~ ~ ~ + mU z L 4- a 0 UJ Z 1 W) -P 11 U W 4) 0 u cJ (5 D j -4- LU -) 0U Z 0) 4- 11
E -6
C C
z U
C: C0 w r
CLL
-4C4 < -4 L)N n:
ol- w ww CzIm wL 0
CLd0X 0 z2 00
V) X4 00 Co r% 4T U
Il c 10 0010CC <r i . o -Cl'
00
M <rq >-W>-W-Z(fn- i4 w -
u<z u< U CZ UTWJ1ZF- < W 4- &4 C F
CLE CLE X 4 CL.E a- x x al> a. - < a~j > o- XC -C C . & IfC ZC L I L --) CC c- - 0.4-) ~ C C
."- ( -.. C CZ4- H4 ( C- :I C-~ 4 o z'A -4 C -(4100 W +- + l:HW 4-) 0 ) W) 0IU 0 a)OUI +)0 U ;G .. , 4-) W 10 +)
W L)UU 4) l =+) U + U al4,)'-) -* *UL) i -)- t a) L) C-)4.>J WCJf0 4C~C t0CJ 0 Z* >~.I 0 * m ! ( : >U i- 0= >u i >U
E-7
I I
F--o
CL C
F- Z-C U) C
w U)w
w F-z
LLi
LiL
CLL
o CDU
CzI.-
C (n
>- LI c
XL C Cwi C) M4 N) w ~ X4 ,
>~ m - IT m C0 C- -3 C1 O ' F- V)Li 0- C 1 C C CLi Lii
CL N Z C C '0l z c X, 10' <c ' CL: N 0) 4c 0'
-I - "4 0) F- N- ~ - (Ici-Cx C4 -i (n U) '- e) i Z N
X: i Cix L)C x
W 0 zz 0' C.L rli C z W i4WL
LI 0 m '.C U) Lii ) (P: M- CPM:I -4 =I Lii- - =1 C- z
>- F-Z U)ViP
-'.~ L" 4 - u i-1 -3N L- F-C<N Ej X4 fuCC~)CC c C'-C. L)-jPi ccI- u 0C-i&C
L)C l i0iC F- u 4Z c -- u c w -) H I jcx FC n 0) zamw) c -LiwUCCALL - £ iCu CL cLi
Z~ C 1-4~- E-"
-4 WZ jo* ,' 0 >-.. c- 0 L
w a' w w 00~ L~ *j 3 01 i C, ii 0 Li
E -8
I
CL~ 0'N
UL 0'0
14M 4T
Z0 u :
w cz (n
0 ;A C ~ 4:'
CL U Z:
LL0 03 Im c cc crinI-j3
: .. A.LL1 0 4
Ul C0') C-4 Z) mL:x 0% -> ' > r0
0 C: ONO z0 3 2T4Oq 00' 14' 0%
C-N Cl .0 0 ~ o 1- 0 x ,C0 N04 1 L 4: I4 0' w 01 1
,a ix 0L !a : l L1-0 4:c m : 14 N4 (1Ld 4
C 14 w 0 NNZ~~~l~ Ld >- .J)
F- 1 1 0 1- i- N- C C
CZ 1-- "-0 wz 00
i: _ _.c )3CO~-HU)CU1- OJL..J *.C:,- _'j -1+~l --
Z~~~~~ ~ ~ ~ -9aC w >-r Z U i
U.'
0'0'0'
0'
z -
C1. La C
-) I C1
N ~0 '
i- C1.4
CNcc Li O
HW+)i I-H-+ CIqMIIOHiaZW-j.-0Z H uWHFOCU)
WW CO > ie CWCOI EnZJ :
w -~ -I CI - 02<.4) Li 1! 5 F i-~ 0 Ud li M: oJ+Z *.- C W c0 _4 Mj Z W C a) z M~'w *U.4T C-4 j +) < ) H > 0U L
-W It~,u W Cc*EWCO 11 *.J -Ix2~ *1 a) .. JHC
-H r 00 n 1 .- A.0 o 0 U'
:1 r4 -4 W C -) 1 E i E-1O
+) 0 Ci 0 aj -) I w o ci13 : a)+) 1 Ii
z
C If LL
Z. j C W C 0 u0 0u Ol Q a-4 L) -0 2:<1N -1 C 4T 0 LL Tr- 2: < >- -
c C Cr_ W- u~ V) fC V 0 ii 0 c ~a: 2tij 2 0 >- =I" 1 0 ->rz wd W
0 0 : -1 z CC Z 74 Z CL- Z " Z17 1- C Ix Lo > )I-4 -I LU4404 OW
LL) LLI 9:4 C4 LU 2 0 Z- L- XZW2:. - C 4-It 202 OW U .1 < a:-" CW Cj1: 0 x -f2 n : z -1ZC <E C z a:- z Cr >- En X Er) Ld IT-4 < )=)0W Z 0CC <C Czx 4 N <-. D- - 0-4 L) -1 000 00 0 =) - -e17 0 1: 64 C <'- P4 j 0-- 0 0l--2:0 2:0 to CT Z C 0 m 1 -2:z z- X .2: U ZZ.C 22 "- 1:4 CQ ;E - ECLX C Z pq 0Z LLU WI-c Z C C) LL C C 4T x 0 02 00c LU I-A -4 n> Cc
Cr c . ) = X* Ix C 3 * L-4 - XW LU' - C LU-a)~- (l- O'0 LUJN 0' C4 IX C CL LL CLLLL4 174 iM =) A z V
En17 U W-j I X Cc r" = 0 174 0 WW .W _jU ~ 0--CN
-T-- --C C , O- C-40 C V) M E WLO 4 -j . I-qI ccX4 2:3 0-tO 3 0 M :C-f C.J -i m 14WWX. C.01 .. I LJZ
>-w Z >- C0 I- - 0' 01 Ir-X T C - xc C - x x = cc i- 0U 2:0 UC "-04 Z -4_jM M' 0 U) ck 0 to i- 00C 0.0) C CIZ F- Z 2CL F--C - < < n < mC< m <3 0 :3 U) -WWL UJ CnL WW U U WlEnWLUI- - )- -j -I-L 9,4 w~ o.J QJ al ol Z OImz4 oD
0 0 r 4 -J >- Lu w>- >- Z>- Z>- i C c.4 j IxU.N LlU 0 w w -<EF J< E0 eN- 2:0 0~U CCr4I~ -CA -t-C~-C: M CIU-,M MZ64
04a CLLmz _ CLC LWLM L L X4 : m (n- M F- " 0. -CL~ r-,W C cc~
o2:w 2: z Ie - 0 0 C 0 'T I - *C I---I0tT 2LU F- - L 0~ 0 0 - I- -> - :IN IWO w 1IW I-CL '-4 j 04-4 0
1) W 04'4 W F->-M 0 : M Ct MI 4C L)0- C U "0P _ X - >0 e 2:j
0L W T0= CW(, "WL L j 0'C0 W 2: :e, F- LU -J L- *W C OCC1-CXzCQ
CC )- wWZ w4owa wo 0:= CC CC 00 001- CC Z
0l.'ZAXCA=A/A.0'OAO!%Z.. <CZ.A. 4CO" . Z'C 'LU'.AA f ~0 0 A. U- .'-.A CL *' a _j C.A. C 4i 0..A A 0 A W A A A
E- 11
LUJ
0
LUz
:2 C
O 1-:-4 0 cc: LUZ i-- Z) 2
X4 w 111 0 0 En LC CIx x !P- En 2U2 0 !w- u .0 2: C 2: C -
C LUI LU F- 0 - m I
LU CNi- 0 W II uu WIXJ,-4 f- CE: -. 1 '-4 2: -.1W ODz0 10
M0 e2L z- CE 1.0 U.1O 2:0 0'U C cc 0 0 -C =.in0 ON IDS
b~Lj U 3 Li. LU U. LU.1 'JC NF-~- 0 174E L) I
L) LU -WU( 0 (N0 U)iT .DL C 0 3- - Go R -.1Of :.JCL. 16- LUI (:0 C*4Li 0m ON0 0x 2: C-i NZ D. N1Ld :>Lil i= -J D- 6 - W -4 oc)!I w1:4LU LU LUI LU. LU =4 01 16- X4 cc:
Ol C Uj C. -C - -4 4- M "a M.!2 0 04 ;4 Z2: I4 1444C -3 LU (1)zflZ02:C > C wU LUc C CO C -CL
CL. CL C CLL c0- 111 WLi CID 0 -. 1 2: M1 .J z>- (nWU >< LU LU LULU U)2Z 2LU >- x:~0 LU)( II - C >- LU>- C -. 1
WUZ - LU ZU :L LJ- F- >- i-- :e J L)-1!- e UZ
-ZLU-C - Z L 1-- a I--C~J C -j If LW W C M -CJ L1
C cJ - -':'2C4 w-L~0 4 >~ r-a)E)crLc w > 'C 0 LU C C-!;ZI Z *0UcJL M,~c u Z 02 7 )E- 1 jUIL OE , 10 Wi4L
(r, '--C3 a, Im --m-- M, i4 0- 4- C "I OC4 ItCE 4-' UC -LCE '1 - 04a. zZ4 Z 2:l H~>f 0 * i)Uf * n i0>fl * .0-:)
m L L Mw= tn .i-,UC *. rJE-4 .U) C0 W W CE02LU0 U *.. -0
E-12
zz00
0-
U- Ix
-'el 0
-4 0
N CL I xO0w 00'mOl 00'l
'-4 -. 1C) LZ -. t-0 C1
0 N 10 '-4 0, ix 0Cr -4 ol z 1 <1 1
M- i= lX IT(A LO)--1 z n Lo o. LOU)
Dn LU>- <LIZ m LU)- zn LUJ>- z:
1-4 LZJ1-< a -4 X4 0 C1-4 m-.1 If LU <LIm4Zj 11 I~iC .Ai-i It Z4
We0"rn x00 WX0F CUl aULl,4iI OI, C (D 11 U' 0x LU0)4I 0-ca)H0W O W 0 w OL<LO (b > 0 wc 0> C 0 W 4M (D >i
m - 9- 00 mi 0<L0 II.4- CZ -M I I!. C0 -1 -i) 1IX4-) C -
If r0 c..-If 0 C LO 11 re 0 c .1-4 M 0-4.)U) --- 41004-)U)C.'Mj (n - M:;C -):1-1-4 c C -)* *UEU-W+1M 1 -* =LOj,~ * J U)
E-13
Lii LLI-iJ
L) u2: z
z
1
U) LO
0 0
LLI
CL in L c)4T 0 u C-4(N
2: J2: <1 (N
V)i 0 1-Z 0 f- 0 4r0 3 <1 00 74 :3 0'l V
S N 1O4 Wl (N'1 iLi IT Z N' -4-C- IT Lt.) i - 4T2: 0 <C 1 0 C-4 I-' C i " 0<i 0' (04 N: X) ir Lb-i 4T0
m~ 1:C (N 0 W4i 17-1 x lj >-Of' U.1 W- I WL ce LI il- U. w WI
Wa 0 1- 1= 1 o 0 -o (n --
1 Wj CE > N W -i < :> NW O i 0W-.1 >- >- E ) Z. 2W OsU En j .J : O'U (nw U)
W >- -1J -1)mW >- 0I' m 1 W>= - 0 X-4 M W1--I- m U) Wi Li LL I- Y '- uW -C<
cr <2 i-,W+ UZ E- co iC W4-)
OWC W >. -- 4 0 w < bIf > 0- = <w it > 0 OwC<W I >WOOM)L C- I H1 w mU)~-'l W U)U)E 9---1 1 WU 3E C-- ., 0
IX- +C Z - 1;4 ixai+-c z- aas. c z - ;-4 )L
C' -0 DL -4 - L4 1 ZC -- c D E-- H in ) + e'~ -,Z~ -H cnL -- C WZL 4-rcE . DLO0 ;(J4)1 *U~ W ) U13) 0) *C+-'1 W4N0 U 3 : W 4-j) 11 4-) .U->)q I
E-14
0
I-
00
Lj -4 a
u L UIx
Z0 C Or 4 -"
La .I L.) )--j I LJ o w I
z- z 1- C *: z
w wr U) WW La C4 M m
C *4 Os La a: Mr Ci -Il -Cs La Z w C0 sL ~* I- *E wJ <r.- Osi I-0> -I. <r _j0>
l' k- *0 >u i *) " M0s >~ Ix* CI- I- _O U *JfW1-- Z *J~ L~ i- *
M _j11 ,- " -j11E - 0 < -15_jI -LI4
m~-' Ce -d -) I ' Y X W4) IiZ l-I4XW4)
co -0 M U
z LI) z 0 a z -40 -4 <: 0Ii-
4-
no M
L~Jo '0 1 lW LJ' OLJcz c0'- 00 Mzo -l
1-4 C L)-L
ZL iJ) Ml E Lu'U LLu : L C r WC -Lu)- 1>-4 Z- L- 0) Ulr" C) ~ C i I- i 0 -4-LO
Oru rW flu' 2, 0W-~ I . >o P' ,> r)~)~-i -J -P -. -j C flE L-- > L
Lli~~W C~ O <r wZ1V Z U 1 -
E-16
ADAO8, 326 OPERATING SYSTEMS INC WOODLAND HILLS CA F/6 9/4SATELLITE AND MISSILE DATA GENERATION FOR AIS. (U)DEC 79 6 M SILVA, C A MONTGOMERY F306G2-78-C-0274
LICLAS7SXFIED SI-R779-037 RADC -TR-79-314 NL
3- END6
~. . .....
wUi0
z
L)
o w
w 0F- l'
U) z
Z--
En Ix4 AIN~C ~ M O Z O
Z -4 QW C-I
1,-H 1w 0 0 C H 4~
<- C I- 1-4 iI z
-1 4 w C-. w LL. -.1 Ill2
U) -w - Z ..J "* w .( wE
W 1- fl cir F- -40 E-e ZHe-eLOU
o)-L T J1 oIww' m-j !I C - II HC 95
<r IJ94CCW -) If:: 1- - X C U 4) ! 1- --i-m CC U)L.. IU 2 1- z I--0U L C0 -C (P."frOU Cc 0 H.- c atOOU ) wiii i.~~ron
IX IT P 11 1) ) W a:- P-1,.~i Q tc -r L v .
AO)CJC - -I aO )> f .- *oH Ln *. x0 C+ : =M j l S-IM*CD )LI(COI D *ai ) U ) 0 J O n *b -() W F 1 oua =1 *aj P11 4)c -'w z i )- 4 )W Z C oUc .IIJw z i
E-17
MISSIONOf
Rome Air Development CenterRAP pt~nb and execute6 weakahwi, devetopnent, tt6. and&deuew~ oxauc ton plwgu -in A uppoA~t o6 Commad, Conut'wCowmun&,At.A owd Intet'gafte (C31) aUtiiti. TechnioAand eng~mteM 4uppoAt .oLtha £L&e6 o6 tehe~ou~a aeorpetenei6 prwuided to U9VP~o Oi~e (P06) end otkeAf ESPe.Lewmn. The *ptlne pa tehnAt oL im on aAea6 a~e
c~oum~niaton6, e teuwwqneti gwudAouie and conf.tOL 6ut-veLU~Ame oi g4ocund and aeuohpeae objec*6D inteUgexee datia~e.Uton and hand ng, .1niomWon ,s6tm teduotogy,
I. L~~~onompheAiU p'Lopagat~on, .4oJi4 6toate 4O~venA, nic.uveSand etectwovda Juljai(Zb. , .ulntanabI2.i4 and
![Redacted Releasable Docs -Set 1 [Page 1-698] 6.10.13](https://static.documents.pub/doc/80x56/577cdc201a28ab9e78a9efee/redacted-releasable-docs-set-1-page-1-698-61013.jpg)