Skip to Main content Skip to Navigation
Theses

Un système multi-agent adaptatif pour la construction d'ontologies à partir de textes

Kévin Ottens 1
1 IRIT-SMAC - Systèmes Multi-Agents Coopératifs
IRIT - Institut de recherche en informatique de Toulouse
Abstract : Semantic Web refers to a set of technologies trying to make the World Wide Web resources content available and usable by programs and software agents. Then, it should ease users access to information. Its raising requires the availability of ontologies which are formalized knowledge models usable by a software systems for their communication. Unfortunately their construction is generally slow and costly, and their maintenance raises issues generally underestimated. The usage of texts for designing ontologies is considered as a possible solution to their cost, despite the fact that this task is still difficult.

Since ontologies must be maintained, and because they can be seen as complex systems of concepts, we propose to use adaptive multi-agent systems to semi-automate the process of building ontologies from texts. The stable state of those systems comes from the cooperative interactions between the constituting software agents. In our case, the agents use distributed algorithms of statistical analysis to find the most satisfying structure from a syntactical and distributional analysis of texts. Then, the user can validate, criticize or modify parts of the agent structure, which is the base of the ontology in progress, to make it fulfil his objectives and his own vision of the modeled domain. In return, the agents reorganize themselves to satisfy the newly introduced constraints. Ontologies which are generally fixed become here dynamic, they become a "living design". Those are the underlying principles used in our system named Dynamo.

The relevance of this approach has been tested thanks to experiments aiming at evaluating the performances of our system, and by its use in real operations. In this thesis, we present and analyse the obtained results.
Document type :
Theses
Complete list of metadatas

Cited literature [87 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00176883
Contributor : Kévin Ottens <>
Submitted on : Monday, October 8, 2007 - 5:17:00 PM
Last modification on : Friday, October 23, 2020 - 4:38:34 PM
Long-term archiving on: : Friday, November 25, 2016 - 6:59:41 PM

Identifiers

  • HAL Id : tel-00176883, version 2

Citation

Kévin Ottens. Un système multi-agent adaptatif pour la construction d'ontologies à partir de textes. domain_stic.gest. Université Paul Sabatier - Toulouse III, 2007. Français. ⟨tel-00176883v2⟩

Share

Metrics

Record views

873

Files downloads

2231