english-unl-analysis-hyderabad-10jun08

7/27/2019 english-unl-analysis-hyderabad-10jun08

1/108

Semantic Parsing

Pushpak Bhattacharyya,Computer Science and Engineering

Department,IIT Bombay

[email protected] contributions from Rajat Mohanty, S. Krishna, Sandeep Limaye


2/108

Motivation

Semantics Extraction has many applications MT IR IE

Does not come free Resource intensive

Properties of words Conditions of relation establishment between words

Disambiguation at many levels Current Computational Parsing less than

satisfactory for deep semantic analysis


3/108

Roadmap

Current important parsers Experimental observations Handling of difficult language phenomena

Brief Introduction to the adopted Semantic

Representation: Universal Networking Language (UNL) Two stage process to UNL generation:

approach-1 Use of better parser: approach-2

Consolidating statement of resources Observations on treatment of verb Conclusions and future work


4/108

Current parsers


5/108

Categorization of parsers

Output

MethodConstituency Dependency

Rule Based

Earley Chart(1970), CYK(1965-70), LFG(1970), HPSG

(1985)

Link (1991),Minipar (1993)

Probabilistic

Charniack(2000), Collins(1999),

Stanford

Stanford(2006), MST(2005), MALT

(2007)


6/108

Observations on some well-known Probabilistic Constituency

Parsers


7/108

Parsers investigated

Charniak: Probabilistic Lexicalized Bottom-UpChart Parser

Collins: Head-driven statistical Beam Search

Parser Stanford: Probabilistic A* Parser

RASP: Probabilistic GLR Parser


8/108

Investigations based on

Robustness to Ungrammaticality

Ranking in case of multiple parses

Handling of embeddings

Handling of multiple POS

Words repeated with multiple POS

Complexity


9/108

Handling ungrammatical

sentences


10/108

Charniak

S

NP VP

NNP

Joe

AUX

has

VP

VBG

reading

NP

DT NN

the book

Joe has reading the book

haslabelledas aux


11/108

Collins

hasshouldhavebeen

AUX


12/108

Stanford

has is treated asVBZ and not AUX.


13/108

RASP

Confuses as acase of sentenceembedding


14/108

Ranking in case of multiple

parses


15/108

Charniak

S

NP VP SBAR

NNP VBD S

John said NP VP

VB

Marry

VBD NP PP

sang DT NN IN NP

NNPwithsongthe

MaX

John said Marry sang the song with Max

semanticallycorrect onechosen fromamong

possiblemultipleparse trees


16/108

Collins

Wrongattachment


17/108

Stanford

Same asCharniak


18/108

RASP Different

POS Tags,but parsetrees arecomparable


19/108

Time complexity


20/108

Time taken

54 instances of the sentence This is jus t tocheck the t ime is used to check the time Time taken

Collins : 40s Stanford : 14s Charniak : 8s RASP : 5s

Reported complexity Charniack: O(n5) Collins: O(n5) Stanford: O(n3) RASP: not known


21/108

Embedding Handling


22/108

Charniak

S

NP

NP SBAR

DT NN WHNP S

The cat WDT VP

that VBD NP

killed NP SBAR

NNDT WHNP S

the rat WDT VP

that VBD NP

stole NP SBAR

DT WHNPNN S

WDTthat VP

NP

NPNP

escaped

VBD

VP

A

A

VBD PP

SBARNP

VP

INspilled

SINNNDTon

was

thatfloorthe

AUX ADJP

slippery

The cat that killed the rat that stole the milk that spilled on thefloor that was slippery escaped.


23/108

Collins


24/108

Stanford


25/108

RASP


26/108

Handling words with multiple

POS tags


27/108

Charniack

S

NP VP

NNP VBZ PP

Time flies IN NP

like DT NN

an arrow

Time flies like an arrow


28/108

Collins


29/108

Stanford


30/108

RASP

Flies tagged asnoun!


31/108

Repeated Word handling


32/108

Charniak

S

NP VP

NNP VBZ SBAR

Buffalo buffaloes S

NP VP

NNP VBZ SBAR

Buffalo buffaloes S

NP VP

NN NNP NNP VBZ

buffalo buffaloBuffalo buffaloes

Buffalo buffaloes Buffalo buffaloes buffalo

buffalo Buffalo buffaloes


33/108

Collins


34/108

Stanford


35/108

RASP

Tags all words as nouns!


36/108

Sentence Length


37/108

Sentence with 394 words

One day, Sam left his small, yellow home to head towards the meat-packing plant where he worked, a task whichwas never completed, as on his way, he tripped, fell, and went careening off of a cliff, landing on and destroyingMax, who, incidentally, was also heading to his job at the meat-packing plant, though not the same plant at whichSam worked, which he would be heading to, if he had been aware that that the plant he was currently headingtowards had been destroyed just this morning by a mysterious figure clad in black, who hailed from the small,remote country of France, and who took every opportunity he could to destroy small meat-packing plants, due tothe fact that as a child, he was tormented, and frightened, and beaten savagely by a family of meat-packing plantswho lived next door, and scarred his little mind to the point where he became a twisted and sadistic creature,capable of anything, but specifically capable of destroying meat-packing plants, which he did, and did quite often,much to the chagrin of the people who worked there, such as Max, who was not feeling quite so much chagrin asmost others would feel at this point, because he was dead as a result of an individual named Sam, who worked ata competing meat-packing plant, which was no longer a competing plant, because the plant that it would becompeting against was, as has already been mentioned, destroyed in, as has not quite yet been mentioned, amassive, mushroom cloud of an explosion, resulting from a heretofore unmentioned horse manure bombmanufactured from manure harvested from the farm of one farmer J. P. Harvenkirk, and more specificallyharvested from a large, ungainly, incontinent horse named Seabiscuit, who really wasn't named Seabiscuit, butwas actually named Harold, and it completely baffled him why anyone, particularly the author of a very longsentence, would call him Seabiscuit; actually, it didn't baffle him, as he was just a stupid, manure-making horse,who was incapable of cognitive thought for a variety of reasons, one of which was that he was a horse, and theother of which was that he was just knocked unconscious by a flying chunk of a meat-packing plant, which hadbeen blown to pieces just a few moments ago by a shifty character from France.


38/108

Partial RASP Parse

(|One_MC1| |day_NNT1| |,_,| |Sam_NP1| |leave+ed_VVD| |his_APP$| |small_JJ| |,_,| |yellow_JJ| |home_NN1| |to_TO| |head_VV0| |towards_II| |the_AT| |meat-packing_JJ| |plant_NN1| |where_RRQ| |he_PPHS1| |work+ed_VVD| |,_,| |a_AT1| |task_NN1| |which_DDQ| |be+ed_VBDZ| |never_RR| |complete+ed_VVN| |,_,| |as_CSA||on_II| |his_APP$| |way_NN1| |,_,| |he_PPHS1| |trip+ed_VVD| |,_,| |fall+ed_VVD| |,_,| |and_CC| |go+ed_VVD| |careen+ing_VVG| |off_RP| |of_IO| |a_AT1| |cliff_NN1| |,_,||land+ing_VVG| |on_RP| |and_CC| |destroy+ing_VVG| |Max_NP1| |,_,| |who_PNQS| |,_,| |incidentally_RR| |,_,| |be+ed_VBDZ| |also_RR| |head+ing_VVG| |to_II||his_APP$| |job_NN1| |at_II| |the_AT| |meat-packing_JB| |plant_NN1| |,_,| |though_CS| |not+_XX| |the_AT| |same_DA| |plant_NN1| |at_II| |which_DDQ| |Sam_NP1||work+ed_VVD| |,_,| |which_DDQ| |he_PPHS1| |would_VM| |be_VB0| |head+ing_VVG| |to_II| |,_,| |if_CS| |he_PPHS1| |have+ed_VHD| |be+en_VBN| |aware_JJ||that_CST| |that_CST| |the_AT| |plant_NN1| |he_PPHS1| |be+ed_VBDZ| |currently_RR| |head+ing_VVG| |towards_II| |have+ed_VHD| |be+en_VBN| |destroy+ed_VVN||just_RR| |this_DD1| |morning_NNT1| |by_II| |a_AT1| |mysterious_JJ| |figure_NN1| |clothe+ed_VVN| |in_II| |black_JJ| |,_,| |who_PNQS| |hail+ed_VVD| |from_II| |the_AT||small_JJ| |,_,| |remote_JJ| |country_NN1| |of_IO| |France_NP1| |,_,| |and_CC| |who_PNQS| |take+ed_VVD| |every_AT1| |opportunity_NN1| |he_PPHS1| |could_VM||to_TO| |destroy_VV0| |small_JJ| |meat-packing_NN1| |plant+s_NN2| |,_,| |due_JJ| |to_II| |the_AT| |fact_NN1| |that_CST| |as_CSA| |a_AT1| |child_NN1| |,_,| |he_PPHS1||be+ed_VBDZ| |torment+ed_VVN| |,_,| |and_CC| |frighten+ed_VVD| |,_,| |and_CC| |beat+en_VVN| |savagely_RR| |by_II| |a_AT1| |family_NN1| |of_IO| |meat-packing_JJ||plant+s_NN2| |who_PNQS| |live+ed_VVD| |next_MD| |door_NN1| |,_,| |and_CC| |scar+ed_VVD| |his_APP$| |little_DD1| |mind_NN1| |to_II| |the_AT| |point_NNL1||where_RRQ| |he_PPHS1| |become+ed_VVD| |a_AT1| |twist+ed_VVN| |and_CC| |sadistic_JJ| |creature_NN1| |,_,| |capable_JJ| |of_IO| |anything_PN1| |,_,| |but_CCB||specifically_RR| |capable_JJ| |of_IO| |destroy+ing_VVG| |meat-packing_JJ| |plant+s_NN2| |,_,| |which_DDQ| |he_PPHS1| |do+ed_VDD| |,_,| |and_CC| |do+ed_VDD||quite_RG| |often_RR| |,_,| |much_DA1| |to_II| |the_AT| |chagrin_NN1| |of_IO| |the_AT| |people_NN| |who_PNQS| |work+ed_VVD| |there_RL| |,_,| |such_DA| |as_CSA||Max_NP1| |,_,| |who_PNQS| |be+ed_VBDZ| |not+_XX| |feel+ing_VVG| |quite_RG| |so_RG| |much_DA1| |chagrin_NN1| |as_CSA| |most_DAT| |other+s_NN2| |would_VM|

|feel_VV0| |at_II| |this_DD1| |point_NNL1| |,_,| |because_CS| |he_PPHS1| |be+ed_VBDZ| |dead_JJ| |as_CSA| |a_AT1| |result_NN1| |of_IO| |an_AT1| |individual_NN1||name+ed_VVN| |Sam_NP1| |,_,| |who_PNQS| |work+ed_VVD| |at_II| |a_AT1| |compete+ing_VVG| |meat-packing_JJ| |plant_NN1| |,_,| |which_DDQ| |be+ed_VBDZ||no_AT| |longer_RRR| |a_AT1| |compete+ing_VVG| |plant_NN1| |,_,| |because_CS| |the_AT| |plant_NN1| |that_CST| |it_PPH1| |would_VM| |be_VB0| |compete+ing_VVG||against_II| |be+ed_VBDZ| |,_,| |as_CSA| |have+s_VHZ| |already_RR| |be+en_VBN| |mention+ed_VVN| |,_,| |destroy+ed_VVN| |in_RP| |,_,| |as_CSA| |have+s_VHZ||not+_XX| |quite_RG| |yet_RR| |be+en_VBN| |mention+ed_VVN| |,_,| |a_AT1| |massive_JJ| |,_,| |mushroom_NN1| |cloud_NN1| |of_IO| |an_AT1| |explosion_NN1| |,_,||result+ing_VVG| |from_II| |a_AT1| |heretofore_RR| |unmentioned_JJ| |horse_NN1| |manure_NN1| |bomb_NN1| |manufacture+ed_VVN| |from_II| |manure_NN1||harvest+ed_VVN| |from_II| |the_AT| |farm_NN1| |of_IO| |one_MC1| |farmer_NN1| J._NP1 P._NP1 |Harvenkirk_NP1| |,_,| |and_CC| |more_DAR| |specifically_RR||harvest+ed_VVN| |from_II| |a_AT1| |large_JJ| |,_,| |ungainly_JJ| |,_,| |incontinent_NN1| |horse_NN1| |name+ed_VVN| |Seabiscuit_NP1| |,_,| |who_PNQS| |really_RR||be+ed_VBDZ| |not+_XX| |name+ed_VVN| |Seabiscuit_NP1| |,_,| |but_CCB| |be+ed_VBDZ| |actually_RR| |name+ed_VVN| |Harold_NP1| |,_,| |and_CC| |it_PPH1||completely_RR| |baffle+ed_VVD| |he+_PPHO1| |why_RRQ| |anyone_PN1| |,_,| |particularly_RR| |the_AT| |author_NN1| |of_IO| |a_AT1| |very_RG| |long_JJ||sentence_NN1| |,_,| |would_VM| |call_VV0| |he+_PPHO1| |Seabiscuit_NP1| |;_;| |actually_RR| |,_,| |it_PPH1| |do+ed_VDD| |not+_XX| |baffle_VV0| |he+_PPHO1| |,_,||as_CSA| |he_PPHS1| |be+ed_VBDZ| |just_RR| |a_AT1| |stupid_JJ| |,_,| |manure-making_NN1| |horse_NN1| |,_,| |who_PNQS| |be+ed_VBDZ| |incapable_JJ| |of_IO||cognitive_JJ| |thought_NN1| |for_IF| |a_AT1| |variety_NN1| |of_IO| |reason+s_NN2| |,_,| |one_MC1| |of_IO| |which_DDQ| |be+ed_VBDZ| |that_CST| |he_PPHS1||be+ed_VBDZ| |a_AT1| |horse_NN1| |,_,| |and_CC| |the_AT| |other_JB| |of_IO| |which_DDQ| |be+ed_VBDZ| |that_CST| |he_PPHS1| |be+ed_VBDZ| |just_RR||knock+ed_VVN| |unconscious_JJ| |by_II| |a_AT1| |flying_NN1| |chunk_NN1| |of_IO| |a_AT1| |meat-packing_JJ| |plant_NN1| |,_,| |which_DDQ| |have+ed_VHD||be+en_VBN| |blow+en_VVN| |to_II| |piece+s_NN2| |just_RR| |a_AT1| |few_DA2| |moment+s_NNT2| |ago_RA| |by_II| |a_AT1| |shifty_JJ| |character_NN1| |from_II||France_NP1| ._.) -1 ; ()


39/108

What do we learn?

All parsers have problems dealing withlong sentences

Complex language phenomena causethem to falter

Good as starting points for structure

detection But need output correction very often


40/108

Needs of high accuracy parsing(difficult language phenomena)

Systems

LanguagePhenomena

Link Parser Charniak

Parser

Stanford

Parser

Machinese

Syntax

MiniPar Collins

Parser

Our System

Empty-PRO Detection No No No No Yes No Yes

Empty-PROResolution

No No No No Yes No Yes

WH-TraceDetection

Yes No No No Yes No Yes

Relative Pronoun Resolution No No No No Yes No Yes

PP Attachment

Resolution

No No Yes No No No Yes

Clausal Attachment Resolution Yes No No No Yes No Yes

Distinguishing Arguments

from Adjuncts

No No No No No No Yes

Small Clause Detection No Yes Yes No Yes No Yes


41/108

Context of our work: Universal

Networking Language (UNL)


42/108

A vehicle for machine translation

Much more demanding than transfer approach or direct approach

Interlingua(UNL)

English

French

Hindi

Chinese

generation

Analysis


43/108

A United Nations project

Started in 1996 10 year program

15 research groups across continents

First goal: generators

Next goal: analysers (needs solving various ambiguityproblems)

Current active groups: UNL-Spanish, UNL-Russian,UNL-French, UNL-Hindi

IIT Bombay concentrating on UNL-Hindi and UNL-English

Dave, Parikh and Bhattacharyya, Journal of Machine Translation, 2002


44/108

UNL represents knowledge:John eats rice with a spoon

Semantic relations

attributes

Universal words


45/108

Sentence EmbeddingsMary claimed that she had composed a poem

compose(icl>do)

Mary(iof>person)

claim(icl>do)

she poem(icl>art)

:01

agt

obj

objagt

@entry.@past

@entry.@past@complete


46/108

Relation repository

Number 39 Groups:

Agent-object-instrument: agt, obj, ins, met

Time: tim, tmf, tmt Place:plc, plf, plt

Restriction: mod, aoj

Prepositions taking object: go, frm

Ontological: icl, iof, equ

Etc. etc.


47/108

Semantically Relatable

Sequences (SRS)Mohanty, Dutta and

Bhattacharyya, Machine

Translation Summit, 2005

S ti ll R l t bl S


48/108

Semantically Relatable Sequences(SRS)

Defini t ion: A semantically relatablesequence (SRS)of a sentence is a groupof unordered words in the sentence (not

necessarily consecutive) that appear in thesemantic graph of the sentence as linkednodes or nodes with speech act labels

E l t ill t t SRS


49/108

Example to illustrate SRS

The man bought a

new car in Junein: modifier

a: indefinite

the: definite

man

past tense

agent

bought

object

time

car

new

June

modifier

SRS f th b ht


50/108

SRSsfrom the man bought a newcar in June

a. {man, bought}b. {bought, car}c. {bought, in, June}d. {new, car}e. {the, man}f. {a, car}


51/108

Basic questions

What are the SRSs of a given sentence? What semantic relations can link the words

in an SRS?


52/108

Postulate

A sentence needs to be broken into sets ofat most three forms

{CW, CW}

{CW, FW, CW} {FW, CW}

where CW refers to content word or a clauseand FW to function word


53/108

Language Phenomena and

SRS

Clausal constructs


54/108

Clausal constructs

Sentences: The boy said that he was reading a

novela. {the boy}

b. {boy, said}c. {said, that, SCOPE}d. SCOPE:{he, reading}

e. SCOPE:{reading, novel}

f. SCOPE:{a, novel}

g. SCOPE:{was, reading}

scope: umbrella for clauses or compounds

P iti Ph (PP)


55/108

Preposition Phrase (PP)Attachment

John published the article in June{John, published}: {CW,CW}

{published, article}: {CW,CW}

{published, in, June}: {CW,FW,CW}

{the, article}: {FW,CW}

Contrast with

The article in June was published by John

{The, article}: {FW,CW}{article, in, June}: {CW,FW,CW}

{article, was, published}: {CW,CW}

{published, by, John}: {CW,CW}

T I fi iti l


56/108

To-Infinitival

PRO element co-indexed with the object him

I forced Johni[PRO]ito throw a party

PRO element co-indexed with the subject I

Iipromised John [PRO]ito throw a party

SRSs are {I, forced}: {CW,CW}

{forced, John}: {CW,CW}

{forced, SCOPE}: {CW,CW}

SCOPE:{John, to, throw}: {CW,FW,CW} SCOPE:{throw, party}: {CW,CW}

SCOPE:{a, party}: {FW,CW}

replacedwith Iin the 2nd

sentence

go deeperthan surfacephenomena

C l iti f th t


57/108

Complexities of that

Embedded clausal constructs as opposed torelative clauses need to be resolved Mary claimed that she had composed a poem

The poem that Mary composed was beautiful

Dangling that I told the child that I know that he played well


58/108

Two possibilities

told

I the child

that hePlayed

well

that Iknow

told

I the child that I knowthat that he

Playedwell


59/108

SRS Implementation

Syntactic constituents to Semantic


60/108

Syntactic constituents to Semanticconstituents

Used a probabilistic parser (Charniak,04)

Output of Charniack parser: tags give

indications of CW and FW NP, VP, ADJP and ADVP

CW

PP (prepositional phrase), IN(preposition) and DT (determiner)FW

Observation:


61/108

Observation:Headwords of sibling nodes form SRSs

John hasbought

a car.

SRS:

{has, bought},{a, car},

{bought, car}a

(C) VP bought

(F) AUX has (C) VP bought

(C) VBD bought (C) NPcar

(F) DT a (C) NN car

bought

car

has


62/108

Work needed on the parse

tree

Correction of wrong PP attachment


63/108

Correction of wrong PP attachment

John has publishedan article onlinguistics

Use PP attachmentheuristics

Get

{article, on, linguistics}

on linguistics

(C)VP published

(F) PP on(C)VBD published (C)NP article

published

(F)DT an

an

(C)NNarticle

(F)IN on

article

(C)NNSlinguistics

(C)NPlinguistics

To Infinitival


64/108

To-Infinitival

Clause boundary is the VP

node, labeled with SCOPE.

Tag is modified to TO, a FW

tag, indicating that it heads

a to-infinitival clause.

The duplication and insertion

of the NP node with head h im

(depicted by shaded nodes) as

a sibling of the VBD node with

head forcedis done to bring out

the existence of a semantic

relation between fo rceand h im.

(C)VPwatch

(C)VBD forced (C)NP him(C) S SCOPE

(F)TOtoto

(C)VP forced

to

forced

(C)VP

(C)PRP him

him

(C)NPhim

him

(C)PRP him

Linking of clauses:


65/108

Linking of clauses:John said that he was reading a novel

Head of S nodemarked as ScopeSRS:

{said, that, SCOPE}

Adverbial clauseshave similar parsetree structures

except that thesubordinatingconjunctions aredifferent from that

(C)VBD said (F) SBAR that

(C) VP said

(F) IN that(C) S SCOPE

said that

Implementation


66/108

Implementation Block Diagram of the system

Parse Tree

Charniak Parser

Scope Handler

Attachment Resolver

WordNet 2.0

Sub-categorization Database

Input Sentence

Parse Tree modification and

augmentation with head and scope

information

Augmented

Parse Tree

Semantically Related Sets

Noun classification

Semantically Relatable Sets

Generator

THAT clause as Subcat property

Preposition as Subcat property

Time and Place

features

Evaluation


67/108

Evaluation

Used the Penn Treebank (LDC, 1995)asthe test bed

The un-annotated sentences, actually fromthe WSJ corpus (Charniak et. al.1987),

were passed through the SRS generator Results were compared with the

Treebanks annotated sentences

Results on SRS generation


68/108

Results on SRS generation

0 20 40 60 80 100

Total SRSs

(FW,CW)

(CW,FW,CW)

(CW,CW)

Parametersmatched

Recall/Precision

Recall

Precision

Results on sentence constructs


69/108

Results on sentence constructs

0 20 40 60 80 100

To-infinitival clause resolution

Complement-clause resolution

Clause linkings

PP Resolution

Parameter

Recall/Precision

Recall

Precision


70/108

SRS to UNL


71/108

Features of the system

High accuracy resolution of different kinds of attachment Precise and fine grained semantic relations between

sentence constituents

Empty-pronominal detection and resolution

Exhaustive knowledge bases of sub-categorizationframes, verb knowledge bases and rule templates forestablishing semantic relations and speech act likeattributes using Oxford Advanced Learners Dictionary (Hornby, 2001)

VerbNet (Schuler, 2005)

WordNet 2.1 (Miller, 2005)

Penn Tree Bank (LDC, 1995) and

XTAG lexicon (XTAG, 2001)

Side effect: high accuracy parsing


72/108

Side effect: high accuracy parsing(comparison with other parsers)

Systems

LanguagePhenomena

Link Parser Charniak

Parser

Stanford

Parser

Machinese

Syntax

MiniPar Collins

Parser

Our System

Empty-PRO Detection No No No No Yes No Yes

Empty-PROResolution

No No No No Yes No Yes

WH-Trace

Detection

Yes No No No Yes No Yes

Relative Pronoun Resolution No No No No Yes No Yes

PP Attachment

Resolution

No No Yes No No No Yes

Clausal Attachment Resolution Yes No No No Yes No Yes

Distinguishing Arguments

from Adjuncts

No No No No No No Yes

Small Clause Detection No Yes Yes No Yes No Yes

Rules for generating


73/108

CW1 FW CW2 REL(UW1,UW2)

Syntactic

Feature

Semantic

Feature

Syntactic

Feature

Semantic Feature Syntactic

Feature

Semantic

Feature

SynCat POS SemCat Lex SynCat POS SemCat Lex SynCat POS SemCat Lex Rel UW1 UW2- - V020 - - - - into N - - - gol 1 3

V - - - - - - within N - TIME - dur 1 3

Rules for generating

Semantic Relations

e.g., turn water into steam

e.g., finish within a week

R l f ti tt ib t


74/108

Rules for generating attributes

String of FWs CW UNL attribute list

generated

has_been VBG @present

@complete

@progress

has_been VBN @present

@complete

@passive

S t hit t


75/108

System architecture

E l ti h


76/108

Evaluation: scheme

E l ti l


77/108

Evaluation: example

Input: He worded the statement carefully.[unlGenerated:76]

agt(word.@entry, he)

obj(word.@entry, statement.@def)

man(word.@entry, carefully)[\unl]

[unlGold:76]

agt(word.@entry.@past, he)obj(word.@entry.@past, statement.@def)

man(word.@entry.@past, carefully)

[\unl]

F1-Score = 0.945

Not heavily

punishedsince attributes

are not crucial

to the meaning!!


78/108

Approach 2: switch to rule

based parsing: LFGlink


79/108

Using Functional Structure from an LFGParser

Sentence

John eats a pastry

SUBJ (eat, John)OBJ (eat, pastry)VTYPE (eat, main)

agt (eat, Ram)obj (eat, mango)

Functional Structure

(Transfer Facts)UNL

Lexical Functional Grammar


80/108

Lexical Functional Grammar

Considers two aspects

Lexical: considers lexical structures and relations Functional: considers grammatical functions of different

constituents, like SUBJECT, OBJECT

Two structures: C-structure(Constituent-structure) F-structure(Functional-structure)

Languages vary in C-structure (word order, phrasalstructure) but have the same functional structure

(SUBJECT, OBJECT, etc.)

LFG St t l


81/108

LFG Structuresexample

Sentence: He gave her a kiss.

C-structure F-structure

XLE P


82/108

XLE Parser

Developed by Xerox Corporation

Gives C-structures, F-structures andmorphology of the sentence constituents

Supports packed rewriting systemconverting F-structure to transfer facts,used by our system

Works on Solaris, Linux and MacOSX

N ti f T f F t


83/108

Notion of Transfer Facts

Serialized representation of the Functional structure

Particularly useful for transfer-based MT systems

We use it as the starting point for UNL generation

Example transfer facts

Transfer Facts - Example


84/108

Transfer Facts - Example

Sentence:

The boy ate the apples hastily.

Transfer facts (selected):ADJUNCT,eat:2,hastily:6

ADV-TYPE,hastily:6,vpadv

DET,apple:5,the:4

DET,boy:1,the:0

DET-TYPE,the:0,def

DET-TYPE,the:4,def

NUM,apple:5,pl

NUM,boy:1,sg

OBJ,eat:2,apple:5

PASSIVE,eat:2,-

PERF,eat:2,-_PROG,eat:2,-_

SUBJ,eat:2,boy:1

TENSE,eat:2,past

VTYPE,eat:2,main

_SUBCAT-FRAME,eat:2,V-SUBJ-OBJ

Workflow in detail


85/108

Ph 1 S t t t f f t


86/108

Phase 1: Sentence to transfer facts

Input: SentenceThe boy ate the apples hastily.

Output: Transfer facts (selected are shown here)ADJUNCT,eat:2,hastily:6

ADV-TYPE,hastily:6,vpadv

DET,apple:5,the:4

DET,boy:1,the:0

DET-TYPE,the:0,def

DET-TYPE,the:4,def

NUM,apple:5,pl

NUM,boy:1,sg

OBJ,eat:2,apple:5PASSIVE,eat:2,-

PERF,eat:2,-_

PROG,eat:2,-_

SUBJ,eat:2,boy:1

TENSE,eat:2,past

VTYPE,eat:2,main

_SUBCAT-FRAME,eat:2,V-SUBJ-OBJ

Phase 2: Transfer facts to word entry


87/108

Phase 2: Transfer facts to word entrycollection

Input: transfer facts as in the previous example Output: word entry collection

Word entry eat:2, lex item: eat

(PERF:-_ PASSIVE:- _SUBCAT-FRAME:V-SUBJ-OBJ VTYPE:mainSUBJ:boy:1 OBJ:apple:5 ADJUNCT:hastily:6 CLAUSE-TYPE:declTENSE:past PROG:-_ MOOD:indicative )

Word entry boy:1, lex item: boy

(CASE:nom _LEX-SOURCE:countnoun-lex COMMON:count DET:the:0NSYN:common PERS:3 NUM:sg )

Word entry apple:5, lex item: apple

(CASE:obl _LEX-SOURCE:morphology COMMON:count DET:the:4NSYN:common PERS:3 NUM:pl )

Word entry hastily:6, lex item: hastily

(DEGREE:positive _LEX-SOURCE:morphology ADV-TYPE:vpadv )

Word entry the:0, lex item: the

(DET-TYPE:def )

Word entry the:4, lex item: the

(DET-TYPE:def )

Phase 3 (1): UW and Attribute


88/108

ase 3 ( ) U a d tt butegeneration

Input: word entry collection Output: Universal Words with (some) attributes generated

In our example:

UW (eat:2.@entry.@past) UW (hastily:6)

UW (boy:1) UW (the:0)

UW (apple:5.@pl) UW (the:4)

Example transfer facts and their mapping to UNL attributes

Digression: Subcat Frames, Arguments and


89/108

Digression: Subcat Frames, Arguments andAdjuncts

Subcat frames and arguments A predicatesubcategorizes for its arguments, or

arguments aregoverned by the predicate.

Example: predicate eat subcategorizes for a

SUBJECTargument and an OBJECTargument.The correspondingsubcat frameis

V-SUBJ-OBJ.

Arguments are mandatory for a predicate.

Adjuncts Give additional information about the predicate

Not mandatory

Example: hastily in The boy ate the apples hastily.

Ph 3(1) H dli f S b t F


90/108

Phase 3(1): Handling of Subcat Frames

Input: Word entry collection

Mapping of subcat frames to transfer facts

Mapping of transfer facts to relations orattributes

Output: relations and / or attributes

Example: for our sentence, agt(eat,boy),obj(eat,apple)relations are generated inthis phase.

R l b f S b t h dli l (1)


91/108

Rule bases for Subcat handlingexamples (1)

Mapping Subcat frames to transfer facts

R l b f S b t h dli l (2)


92/108

Rule bases for Subcat handlingexamples (2)

Mapping Subcat frames, transfer facts to relations / attributesSome simplified rules

Phase 3(2): Handling of adjuncts


93/108

Phase 3(2): Handling of adjuncts

Input: Word entry collection

List of transfer facts to be considered for adjuncthandling

Rules for relation generation based on transfer factsand word properties

Output: relations and / or attributes

Example: for our sentence, man(eat,hastily)relation; and @defattributes forboy, appleare generated in this phase.

Rule bases for adjunct handling


94/108

j gexamples (1)

Mapping adjunct transfer facts to relations / attributesSome simplified rules

Rule bases for adjunct handling examples (2)


95/108

Rule bases for adjunct handlingexamples (2)

Mapping adjuncts to relations / attributes based on prepositions- some example rules

Final UNL Expression


96/108

Final UNL Expression

Sentence:The boy ate the apples hastily.

UNL Expression:

[unl:1]

agt(eat:2.@entry.@past,boy:1.@def)

man(eat:2.@entry.@past,hastily:6)

obj(eat:2.@entry.@past,apple:5.@pl.@def)[\unl]

Design of Relation Generation Rules


97/108

gan example

Subject

Verb Type Verb Type

agt aoj aoj aoj aoj obj

ANIMATE INANIMATE

dobe

occur dobe

occur


98/108

Summary of Resources

Mohanty and Bhatacharyya, LREC 2008

Lexical Resources


99/108

Verb

Knowledgebase

Semantic

Argument

Frame

Syntactic and Semantic

Argument mapping

Verb

Senses

Functional

Elements with

Grammatical

attributes

Lexical

Knowledgebase

with

Semantic

attributes

N V AdvA

UNL Expression Generation

Auxiliary verbs Determiners

Tense-Aspect morphemes

Syntactic Argument

database

PPs as

syntactic

arguments

N V Adv

Clause as

syntactic

arguments

A

SRS Generation

Use of a number of lexical data


100/108

We have created these resources over a long periodof time from

Oxford Advanced Learners Dictionary (OALD) (Hornby, 2001)

VerbNet (Schuler, 2005)

Princeton WordNet 2.1 (Miller, 2005)

LCS database (Dorr, 1993)

Penn Tree Bank (LDC, 1995), and

XTAG lexicon (XTAG Research Group, 2001)

Use of a number of lexical data


101/108

Verb Knowledge Base (VKB) Structure

VKB statistics


102/108

VKB statistics

4115 unique verbs 22000 rows (different senses)

189verb groups

Verb categorization in UNL and its relationship totraditional verb categorization


103/108

traditional verb categorization

Traditional(sytactic)

UNL(semantic)

Transitive(has direct

object)

Intransitive

Do(action)

Ram pulls therope

Ram goes home

(ergative

languages)

Be(state)

Ram knows

mathematics Ram sleeps

Occur(event)

Ram forgotmathematics

Earth cracks

Unergative(syntactic subject=semantic agent)

Unaccusative(syntactic subjectsemantic agent)

Accuracy on various phenomena


104/108

and corpora


105/108

Applications

MT and IR


106/108

MT and IR

Smriti Singh, Mrugank Dalal, Vishal Vachani,Pushpak Bhattacharyya and Om Damani, HindiGeneration from Interlingua, Machine TranslationSummit (MTS 07), Copenhagen, September, 2007.

Sanjeet Khaitan, Kamaljeet Verma and Pushpak

Bhattacharyya, Exploi t ing Semant ic Proxim i ty forInformation Retrieval, IJCAI 2007, Workshop onCross Lingual Information Access, Hyderabad, India,Jan, 2007.

Kamaljeet Verma and Pushpak Bhattacharyya,Context-Sens i t ive Semant ic Smoothing us ingSemant ical ly Relatable Sequences, submitted

Conclusions and future work


107/108

Conclusions and future work

Presented two approaches to UNLgeneration

Demonstrated the need for Resources

Working on handling difficult languagephenomena

WSD for correct UW word

URLs


108/108

URLs

For resourceswww.cfilt.iitb.ac.in

For publications

www.cse.iitb.ac.in/~pb
http://www.cfilt.iitb.ac.in/http://www.cse.iitb.ac.in/~pbhttp://www.cse.iitb.ac.in/~pbhttp://www.cfilt.iitb.ac.in/

Date post:	14-Apr-2018
Category:	Documents
Upload:	arteepu4
View:	216 times
Download:	0 times

english-unl-analysis-hyderabad-10jun08

Documents