Méthodes ascendantes pour l'ingénierie des connaissances

Abstract : The research reported in this document is part of Knowledge Engineering field. We hope that it contributes to better delimitate and define this scientific area. Knowledge Engineering aims at defining technical, methodological and organisational solutions to the problem of identifying, modelling and broadcasting knowledge through software applications. Modelling is considered here as an independent step prior to formalization in an operational system.

The approach that we promote underlines the importance of the knowledge related to the practices of the target system's users. Our goal is to design tools and methods that will support modelling, and more precisely the bottom-up process required for knowledge abstraction.

The studies that we carried out systematically rely on experiments, through enterprise applications and case-studies, and on cross-disciplinary collaborations. The contributions presented further on include knowledge elicitation and analysis techniques and tools, methods and formalisms for knowledge representation and modelling, as well as software work-benches that integrate these various aid tools.

These methods and tools successively contribute to three different ways of formulating the modelling problem, in keeping with the historical evolution of the research field. The first and large problem addressed here covered conceptual modelling from domain experts' knowledge and activity. To this end, we referred to theories and techniques from psychology and human factors. A second way to deal with modelling has been to focus on text analysis and to the gain brought by linguistic approaches and natural language processing tools. Then the goal was to build up specialized domain models like terminological knowledge bases and ontologies. From the various experiments carried out to evaluate these contributions, we can conclude that the way models will be used in applications should be taken into account as soon as possible when building them. As a consequence, the third paradigm relates to the study of how conceptual models, in particular ontologies, may be used in some specific application fields, so that engineering methods and tools could be adapted to these target uses. Most of the applications that we focused on refer to information search or the access to document content. This issue raises main questions about how models and texts form complementary knowledge supports.

This research offers rich perspectives along two different directions. On the one hand, if models make it easier to explore and manage document content, which could be the proper modalities for semantic document indexing? On the other hand, because document collections and users' needs are regularly changing, how can model (and ontology) maintenance be taken into account early during the building stage ? The originality of the approach that we propose is to deal with these two issues simultaneously, and to look for a coherent solution that would rely on methods and tools for engineering models from text analyses.
Document type :
Habilitation à diriger des recherches
Complete list of metadatas

Cited literature [7 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00089165
Contributor : Nathalie Aussenac-Gilles <>
Submitted on : Thursday, August 10, 2006 - 6:51:32 PM
Last modification on : Thursday, October 17, 2019 - 8:57:14 AM
Long-term archiving on : Friday, May 13, 2011 - 11:40:31 PM

Identifiers

  • HAL Id : tel-00089165, version 1

Citation

Nathalie Aussenac-Gilles. Méthodes ascendantes pour l'ingénierie des connaissances. Informatique. Université Paul Sabatier - Toulouse III, 2005. ⟨tel-00089165⟩

Share

Metrics

Record views

1046

Files downloads

3451