Pierre Pompidor

Abstract Supervised classification has been extensively addressed in the literature as it has many applications, especially for text categorization or web content mining where data are organized through a hierarchy. On the other hand, the automatic analysis of brand names ...

Résumé Ce papier se situe dans le cadre du projet Chimère dont objectif est de faciliter l'accès à plusieurs ser-veurs d'information spécifiques à un domaine donné et dont la particularité est de délivrer des informa-tions extraites de bases de données après que l'utili-sateur ait rempli des formulaires. Les pages de tels serveurs incluent ces formulaires qui se composent de zones de saisies et de données textuelles apparais-sant à proximité immédiate de ces zones. Ces types de pages sont très fréquents dans le domaine de la réservation de moyens de transport (ex. Air France, SNCF) et dans le commerce électronique. Un tel contexte pose deux problèmes à résoudre : i)l'extraction des informations de telles pages "for-mulaires" en combinant les données structurées et les "brèves" données textuelles, ii) l'intégration et l'exploitation des informations extraites de différents sites et répondant à un même service, sachant que les problèmes d&#3...

Within the framework of course-assisted creation, we are developing a prototype whose goal will be to enrich an ontology of teaching concepts. To do that, we automatically query search engines with key words extracted from pages that were previously analysed (beginning with an ontology which treats on a hierarchical basis the first keywords). This analysis (lexical, syntactic and semantic), targeted towards the extraction of definitions ("this concept-X is a ...etc.") and of specializations ("this concept-X comes in ...etc."), is based on the exploitation of a base of syntactic templates automatically learned from the analysis of definitions of several on-line dictionaries, and from a minimal dic- tionary (in particular to treat synonymies). The ontology thus created finally generates a course draft that must be "finalized" manually, in spite of the use of a few subroutines for knowledge synthesis (for the moment our work is mainly focused on the enrich...

This paper is dedicated to the ontologies manage- ment, and more particularly to a tool called Kaon. One of its objectives is to propose the integration of a such application in a peer-to-peer platform. In- deed, the tools which are provided by Kaon can be used for the management of the distributed re- sources sharing. The integration of Kaon to mod- elize and to organize the knowledge in the Edutella network, for example, would allow the realization of more effective search engines. The Kaon plat- form is already used for the annotations manage- ment in Edutella project. It would be judicious to spread it to all the resources which are shared within the Internet network. The use of Kaon would allow to increase the efficiency of searching services thanks to the modeling of the semantic links between the various learning objects which are shared through the Edutella network.

Currently most servers contain essentially pages of texts, but the great economic stake is to provide access at large databases in order to offer a service to a user. However the Web sites which offer accesses to databases contain textual information additional to structured data. We are interested in services offered by this type of site. The first objective is

The aim of our work is to provide researchers in experimental sciences means to review their knowledge in a given domain, while confronting it to experimentation data or to calculation tool results. To set up this environment and a methodology, we described the researcher’s knowledge as an oriented object framework and we make use of a format of exchange facilitating the comparison of the results generated by the different tools. Besides, our tool is able to provide researchers the useful information which is extracted from the domain databases. In this article, we present the architecture of our approach and its components. Finally, we will illustrate our framework by an acquisition/revision cycle argued by an example in immunogenetics.

The World Wide Web offers a great availability of heterogeneous educational resources. This suggests the idea that such materials can be re-used in compose courses. In this paper we address this issue by proposing an architecture for composing teaching courses using “the best parts“ of heterogeneous educational materials available on the Web. Course composition relies on a simple but effective

The aim of this article is to propose to a teacher a framework to evaluate and revise his courses by following the students during the learning phase. We assume that the teacher doesn't use an educational platform and have resources at his disposal. We model the course by using a global ontology (an overall view of the system) with limited

RÉSUMÉ . Confronté à la problématique de l'indexation de trè s grands corpus documentaires d'entreprises, nous avons mis au point une méthode simple mais efficace (en terme de temps de calcul et de volumétrie), permettant de filtrer par document les co-occurrences les plus représentatives de ceux-ci. Nous nous plaçons dans un contexte de co-occurrences pour deux raisons. D'une part les requêtes portant sur des co rpus spécialisés, et composées par des experts, s'appuient sur peu de termes précisément c hoisis dont nous indexons les associations, et d'autre part car cela facilitera la construction de cartes sémantiques de navigation dans les concepts du corpus. L'axe principal de ce travail e st la prise en compte de la structure des documents en validant les contenus des paragraphes par ceux de leurs titres. Notre méthode s'appuie sur des mesures tf.idf successives effectu ées dans le contexte d'un document et non d'un corpus, sur les con...