Web Semantic inference using graph theory: the case of an industrial knowledge engineering software

For about ten years, knowledge engineering has been increasingly used by industrial or administrative organisations. Knowledge is considered as a resource which may be drilled, refined, organised in order to become a sharable or saleable product, especially in the areas of semantic web. This evolution and the involved needs of this domain are currently studied by computer science community.

In particular, the question of knowledge description and interpretation is fundamental: how can human knowledge be expressed in order to be processed by computers ? How to be sure that someone’s interpretation of descriptions is semantically compatible to someone else’s interpretation? How can computers provide automatic reasoning to permit the design of knowledge engineering software?

The purpose of this Ph.D. is to provide reasoning capabilities to ITM, the knowledge base management system designed by Mondeca. One characteristic of this work is to build these reasoning mechanisms using graph theory operations. At the end, the reasoning system should permit:

(1) import and export of knowledge serialized using web semantic standards : Topics Maps (ISO standard), RDF (W3C standard), OWL (W3C ontology representation language), RuleML (W3C rule markup language).

(2) knowledge internal ITM description, storage and management using ITM labeled hypergraph formalism,

(3) the full processing of this knowledge using conceptual graphs formalism.

Conceptual graphs (CGs) are a formal knowledge representation model provided with a graph operation called projection which permit the design of reasoning mechanisms which are sound and complete with respect to deduction in first order logics (see [Chein & Mugnier, 1992]) for simple graphs, ([Chein et al., 1998]) and for nested graphs.

The CG model is also provided with a formal semantic in mathematical model theory. This is useful to design correct reasoning mechanisms on knowledge expressed with web semantic standard languages.

Since 1991, RCR team (Knowledge and Reasoning Representation team) from LIRMM (Computer Science, Robotic, Microelectronic Montpellier's Laboratory) has been studying CGs as a graphical knowledge representation model, i.e. a model that uses graph-theoretic notions in an essential and nontrivial way. The aim of knowledge graphical descriptions and reasoning mechanisms consists in providing an interesting alternative to the classical first order logic knowledge representation model. A graph-based reasoning model provides two main advantages:

- From a computational viewpoint, reasonings benefit from combinatorial algorithms (from graph theory).

- From a modeling viewpoint, reasonings can be visualized in a natural way and are simple to understand for an end-user. This property is particularly significant for knowledge acquisition.

References

[BM02] Jean-François Baget, Marie-Laure Mugnier. Extensions of Simple Conceptual Graphs: the Complexity of Rules and Constraints, Journal of Artificial Intelligence Research (JAIR), vol. 16, 2002, pages 425-465. Available on JAIR website. http://www.jair.org/abstracts/baget02a.html.

[BA04] JF. Baget, « Homomorphismes d’hypergraphes pour la subsomption en RDF/RDFS », actes de Langages et Modèles à Objets, 2004

[FLTO4] F. Fürst, M. Leclère et F. Trichet «  Operationalizing domain ontologies : a method and a tool », European Conference on Artificial Intelligence, ECAI, Valence, août 2004.

ITM : Intelligent Topic Manager for Information Intelligence Content Classification, Ontology, Taxonomy and Thesaurus Management, Mondeca, http://www.mondeca.com

Cogitant : Conceptual Graphs Software Developper Tools (GPL Licence) http://cogitant.sourceforge.net/

TooCom (a Tool to Operationalize an Ontology with the COnceptual graph Model. TooCoM) : graphical tool allowing to edit and test ontologies. This tool uses the graphical inference engine Cogitant to provide inference services on ontologies. (GPL Licence) http://sourceforge.net/projects/toocom/.