Approches hybrides pour la recherche sémantique de l'information : intégration des bases de connaissances et des ressources semi-structurées

Abstract : Semantic information retrieval has known a rapid development with the new Semantic Web technologies. With these technologies, software can exchange and use data that are written according to domain ontologies describing explicit semantics. This ``semantic'' information access requires the availability of knowledge bases describing both domain ontologies and their instances. The most often, these knowledge bases are constructed automatically by annotating document corpora. However, while these knowledge bases are getting bigger, they still contain much less information when comparing them with the HTML documents available on the surface Web.Thus, semantic information retrieval reaches some limits with respect to ``classic'' information retrieval which exploits these documents at a bigger scale. In practice, these limits consist in the lack of concept and relation instances in the knowledge bases constructed from the same Web documents. In this thesis, we study two research directions in order to answer semantic queries in such cases. The first direction consists in reformulating semantic user queries in order to reach relevant document parts instead of the required (and missing) facts. The second direction that we study is the automatic enrichment of knowledge bases with relation instances.We propose two novel solutions for each of these research directions by exploiting semi-structured documents annotated with concept instances. A key point of these solutions is that they don't require lexico-syntactic or structure regularities in the documents. We position these approaches with respect to the state of the art and experiment them on several real corpora extracted from the Web. The results obtained from bibliographic citations, call-for-papers and geographic corpora show that these solutions allow to retrieve new answers/relation instances from heterogeneous documents and rank them efficiently according to their precision.
Document type :
Theses
Liste complète des métadonnées

Cited literature [42 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00737282
Contributor : Abes Star <>
Submitted on : Monday, October 1, 2012 - 2:37:10 PM
Last modification on : Friday, April 19, 2019 - 11:53:31 AM
Document(s) archivé(s) le : Wednesday, January 2, 2013 - 6:45:18 AM

File

VA2_MRABET_YASSIN_12072012.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-00737282, version 1

Collections

Citation

Yassine Mrabet. Approches hybrides pour la recherche sémantique de l'information : intégration des bases de connaissances et des ressources semi-structurées. Autre [cs.OH]. Université Paris Sud - Paris XI, 2012. Français. ⟨NNT : 2012PA112135⟩. ⟨tel-00737282⟩

Share

Metrics

Record views

960

Files downloads

751