20110518 Yoon kyoung-a
A Semantic Match Algorithm for Web Services Based on Improved Semantic
Distance
Gongzhen Wang, Donghong Xu, Yong Qi, Di HouSchool of Electronics and Information Engi-neering, Xi’an Jiaotong University, Xi’an 710049, China
UDDI Current web service discovery mechanism is mainly based on it Include WSDL Based on syntax Limit the precision ration and the recall ration of service discovery
Presented semantic Match algorithmBasic semantic Match algorithm Semantic Match algorithm based on semantic distance
also limit the precision ration and recall ration
Propose a semantic match algorithm based on improved semantic distance To eliminate defects Improve the recall ration and the precision
ration of service discovery
Introduction
OWL-S Describe the properties and capabilities of their web services Three essential type of knowledge about a services
Service Profile: What the services does Service Model: How the services workService Grounding: details of how to access a service
Service Profile Describe the function and interface of web services
Important role in semantic match Services are described in terms of IOPE(Input, Output, Preconditions
and Effects) Current semantic match algorithms mainly based on Input and
Output Advertisements and search queries:
are expressed in terms of OWL-S The process of service match :
extract Inputs and Outputs from the advertisement match them with Inputs and Outs of search queries
Ex) Input: date , region , Output: weather
Related Work
Four matching degrees exact > plugIn > subsumes > fail
Matching degree of the advertisement against the re-quest degreeOfMatch(outR, outA) , degreeOfMatch(inR, inA)
ProblemIf an advertisement claims to output a certain concept C, it will output
each subclass of CHowever, in the real world, it will usually output some subclasses of C,
not each subclass
Analyze of current sematic match algo-rithms (1 / 3)
Four matching degrees exact > plugIn > subsumes > fail
ProblemDoes not cover the binary relation
Advertisement: Ballpen, Ballpen has a property BallenLead Request: “BallenLead”
Does not cover the similar relation Advertisement: HireHonda Request: “HireBMW”
About matching degrees(only four matching degree) (Car , BMW) , (Vehicle , BMW)
Analyze of current sematic match algo-rithms (2 / 3)
considered Semantic dis-tance
considered binary rela-tion
Semantic match algorithms based on semantic distance Represents the similarity degree of two concepts A B & B A : equivalent
ProblemThere is no direction
Ex) Concept A is a subclass of concept B
- A B B A There are some false positives
C and E are not catchable at all
Analyze of current sematic match algo-rithms (3 / 3)
Specialization If concept C1 is a subclass of concept C2, C1 is a specialization of C2. If
C1 is an immediate subclass of C2, in weighted ontology map, there is a direction edge representing the specialization from C2 to C1.
Generalization If concept C1 is a superclass of concept C2, C1 is a generalization of C2.
If C1 is an immediate superclass of C2, in weighted ontology map, there is a direction edge representing the generalization from C2 to C1.
The binary relation If concept C2 is a part of concept C1, the relation from C1 to C2 is a bi-
nary relation. If C2 is a immediate part of C1, in weight ontology map, there is a direction edge representing a binary relation from C1 to C2.
The similar relation If concept C1 and concept C2 have a same superclass, there is a similar
relation from C1 to C2.
Four kinds of relations in Improved algo-rithms
C2
C1
C1
C2
C1
C2
C1
C2
Calculate semantic distances linked just by generalizations or just by specializations
Improved algorithm
Performance comparison
Al 2
Current Al
Proposed Al
Specialization, Generalization 1Binary relation 2
Specialization 1Generalization 1Binary relation 2
1. precision of the matching degree2. consideration of the binary relation3. consideration of the similar relation4. consideration of the direction 5. False positives
Differences of these three algorithms
Proposed a semantic match algorithm based on improved semantic distance Compared to the algorithm 2
It considers the binary relation and similar relation Compared to current semantic match algorithm based on sematic
distanceIt removes the false positives It considers the direction
Improved algorithm improves the recall ration and the precision ration of service discovery
Conclusion