Similarity Measurement about Ontologybased Semantic Web Services

1
Similarity Measurement about Ontology-based
Semantic Web Services

• Xia Wang Yi Zhao
• Digital Enterprise Research Institute, Galway
  Ireland
• FernUniversitaet, Hagen Germany

2
Outline

• Motivation
• Scenario of Semantic Services
• Related Work
• Our Work
• Conclusion

3
Motivation

• Similarity of Ontologies is key challenge in
  service discovery, composition, etc
• Specific compound name feature of Ontology in SWS
  application domain
• To calculate the Ontology similarity is also the
  challenge in traditional knowledge management
  fields

4
Scenario of Semantic Services

• Two cases of Ontology terms in service
  descriptions
• Formal terms , e.g., zip and Code
• Compound terms, e.g., findZipCodeDistance and
  CalcDistTwoZipKm

5
Related Work

• Ontology Mapping work
• Hierarchical structure
• Edge-based the short path from one node to the
  other
• Node-based shared information content
• Some distance measures
• Bernstein et al 2003
• Ontology distance shortest path through a
  common ancestor in a directed acyclic graph
• Vector Space Approach compute cosine or
  Euclidean distances
• Full-text retrieval Method (TF-IDF) to compare
  documents in information retrieval
• Dong et al 2004
• Clustering similar terms
• Computing the similarity of operation based on
  TF-IDF
• Hau et al 2005
• Information-theoretic based
• RDF statements inferencibility of OWL-Lite
  constructs
• Overall similarity computed as an amalgamation
  function

6
Our Work
7
Name features of Ontology concepts in Service
description

• Abbreviations e.g. CalcDistTwoZipsKm
• Associated words with capitalization or
  delimiters e.g. LogIn, ArrivalAirport_In
• Words with suffix and prefix e.g. hasFlavour,
  locatedIn
• Variations or misspelling e.g. Booking,
  madeFromGrape
• Free inventions e.g. Code_1

8
Ontology similarity

• dis w1Diss w2 Disi w3 Disc,
• 
  w1 w2 w3 1
• Diss concept structure
• fuzzy weighted associative network
• Disi common contents shared by concepts
• information-theoretical approach
• Disc - concept clustering for compound terms

9
Distance based on fuzzy weighted associative
network, Diss

• Assume four kinds of binary, weighted
  relationships
• (g)eneralization - superclass
• (s)pecilization - subclass
• (n)egative association - disjoined
• (p)ositive association - equivalent
• Fuzzy-weighted edge and distance calculation
• Distancethe strength of the shortest path among
  the two nodes
• Combining the strength of each relation in a path
  is done by using the triangular norms for fuzzy
  set intersections

Following Table1-3, ?2(?2(0.9,0.9),0.9)0.729,
then State p,0.729 Zip, finally Diss(state,
Zip)0.729
10
Information-theoretical distance, Disi

• Relates to the concepts commonality and
  difference
• Ontology terms C and D in service description,
• CnD - number of common elements
• C/D - elements in C but not in D
• ?,d- weight values defining the relative
  importance of their non-common characteristics.

11
Concept Clustering for Compound Terms, Disc

• Association rule
• support
• confidence

12
Concept Clustering for Compound Terms, Disc
(cont.)

• Algorithm
• 1. Extract Ontology terms from a mass of service
  descriptions to get Tt1,t2,,tn
• 2. Compute s and c, cluster T and get clusters
  Xk
• 3. Get the optimal clustering by

13
Conclusion

• Investigate Ontology terms for the semantic Web
  services context
• Review three similarity measure methods
• Present a combined approach to calculate the
  semantic distance of service Ontologies
• The future work is to improve our WSMO mediator
  tool and evaluate our work

14
Thank you!