Construction d'ontologies à partir de textes : une approche basée sur les transformations de modèles

Abstract : Since its emergence in the early 1990s, the notion of ontology has been quickly distributed in many areas of research. Given the promise of this concept, many studies focus on the use of ontologies in many areas like information retrieval, electronic commerce, semantic Web, data integration, etc.. The effectiveness of all this work is based on the assumption of the existence of a domain ontology that is already built an that can be used. However, the design of such ontology is particularly difficult if you want it to be built in a consensual way. If there are tools for editing ontologies that are supposed to be already designed, and if there are also several platforms for natural language processing able to automatically analyze corpus of texts and annotate them syntactically and statistically, it is difficult to find a globally accepted procedure useful to develop a domain ontology in a progressive, explicit and traceable manner using a set of information resources within this area. The goal of ANR DaFOE4App (Differential and Formal Ontology Editor for Application) project, within which our work belongs to, was to promote the emergence of such a set of tools. Unlike other tools for ontologies development, the platform DaFOE presented in this thesis does not propose a methodology based on a fixed number of steps with a fixed representation of theses steps. Indeed, in this thesis we generalize the process of ontologies development for any number of steps. The interest of such a generalization is, for example, to offer the possibility to refine the development process by inserting or modifying steps. We may also wish to remove some steps in order to simplify the development process. The aim of this generalization is for instance, for the overall process of ontologies development, to minimize the impact of adding, deleting, or modifying a step while maintaining the overall consistency of the development process. To achieve this, our approach is to use Model Driven Engineering to characterize each step through a model and then reduce the problem of switching from one step to another to a problem of models transformation. Established mappings between models are then used to semi-automate the process of ontologies development. As all this process is stored in a database, we propose in this thesis, for Model Based Database (MBDB) because they can store both data and models describing these data, an extension for handling mappings. We also propose the query language named MQL (Mapping Query Language) in order to hide the complexity of the MBDB structure. The originality of the MQL language lies in its ability, through queries syntactically compact, to explore the graph of mappings using the transitivity property of mappings when retrieving informations.
Document type :
Theses
Complete list of metadatas

Cited literature [75 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00786260
Contributor : Bénédicte Carouge <>
Submitted on : Friday, February 8, 2013 - 11:20:39 AM
Last modification on : Friday, May 17, 2019 - 1:03:34 AM
Long-term archiving on : Thursday, May 9, 2013 - 3:55:15 AM

Identifiers

  • HAL Id : tel-00786260, version 1

Collections

Citation

Henry Teguiak. Construction d'ontologies à partir de textes : une approche basée sur les transformations de modèles. Autre. ISAE-ENSMA Ecole Nationale Supérieure de Mécanique et d'Aérotechique - Poitiers, 2012. Français. ⟨NNT : 2012ESMA0027⟩. ⟨tel-00786260⟩

Share

Metrics

Record views

452

Files downloads

5082