Cdric Pruski

1
On the Use of Ontologies for an Optimal
Representation and Exploration of the Web

• Cédric Pruski
• LASSY Seminar
• Luxembourg, February 22nd

2
Motivations and Problems
3
Motivations and Problems
A priori knowledge is needed Query language is
important
4
Agenda

• Related Work
• Our approach
• Data structures and query languages for the Web
• Ontologies
• WPGraph, W3Graph
• ASK
• Logical formalization
• O3 an optimal algorithm principles and
  implementation
• Conclusion

5
Related Work Web Search
6
Related Work Ontologies

• Philosophy
• Ontologies in computer science
• Artificial Intelligence
• Software Engineering
• Ontologies for the Semantic Web
• Metadata
• Ontology languages (DAMLOIL, OWL)
• Problems with ontologies
• Formalization
• Suitability
• Evolution

7
Our Approach
8
Data Structures for the Web Ontologies

• Oriented, connex graph where vertices are
  concepts, and arcs are labelled with the type of
  relation binding concepts
• Definition
• set of concepts.
• set
  of arcs.
• 
  

9
Ontologies for representing Web Data

• Useful to measure semantic distance between
  concepts
• Jiang-Conrath MetricJiang97
• Lin Metric Lin98
• Resnik Metric Resnik95
• Hirst-St-Onge Metric Hirst97
• Useful for concepts disambiguation

10
Ontologies for exploring the Web

• Ontologies can be used for
• Annotating Web pages
• Query expansion
• Advantages
• Reducing the size of the search domain
• Facilitating the construction of good queries
• No need to enrich existing Web pages in metadata

11
Data Structures for the Web WPGraph

• Definition
• set of vertices
• set of edges
• types of objects
• vertices labelling function
• weight function for vertices
• weight function for edges (based
  on an ontology)

12
Examples
13
Data Structures for the Web W3Graph

• Definition
• set of
  vertices
• set of edges
• weight function for edges (based
  on an ontology)

14
ASK A Language for querying Web data structures

• Developed for extracting knowledge contained in
  the WPGraphs and W3Graphs
• Improves languages of Web search engines
• Integration of new boolean operators
• Integration of document type
• Integration of concept importance

15
ASK Syntax and Semantic

• Syntax
• Example
• (mot1mot2)
• (mot1mot2)1 img

16
Ontology-based Query Expansion

• Reduce the size of the initial query
• Optimize the search results
• Example

17
Logical Formalization for Query Interpretation

• Definition of logical structures for WPGraph,
  W3Graph and ontologies
• Method for the transformation of ASK queries into
  logical formulae
• Method for interpretation of logical formulae on
  the logical structures

18
O3, an Optimal Algorithm for Web Search
Principles
19
O3, an Optimal Algorithm for Web Search
Implementation
20
Conclusion

• Results
• New concepts for Web data representation and
  exploration
• Improvement of Web information retrieval
• Limits
• Graphs construction needs to much time but can be
  used on the top of a search engine
• Ontologies are not easy to build
• Perspectives
• Improvement of ontologies (mainly the semantics)
• Development of new metrics

21
Questions ?