Skip to Main content Skip to Navigation

Recherche d'Information Monolingue et Translinguistique : de la Désambiguïsation vers l'Expansion Sémantique de Requêtes

Abstract : This Ph.D. thesis proposes possibilistic models for word sense disambiguation (WSD) and query expansion (QE) applied in monolingual and cross-language information retrieval (IR). In order to solve the problem of semantic ambiguity in an uncertain and imprecise IR context, the possibility theory is appropriated for this kind of application. Indeed, we treat the uncertainty, caused by polysemy, using possibilistic networks. These networks offer a framework for modeling the dependencies between the ambiguous terms on the one hand and the semantically related words on the other hand. We propose in this work a «Semantic Dictionary of Contexts» to set up the semantic relations between words. Secondly, we focused on the analysis and experimentation of the effect combining the semantic disambiguation with query expansion on IR. A co-occurrence graph representation was used to compute the similarity between query terms (in the case of expansion) or between terms and meanings (in the case of disambiguation). The proposed approach is based on possibilistic networks for query disambiguation and expansion by considering, for modeling the co-occurrence graph, that two nodes are related if they exist in the same sentence. The edges are weighted by the normalized frequency of co-occurrence of the related terms. On the other hand, ambiguous words are related with their appropriate meanings in the dictionary. The last part of the thesis focus on CLIR in which the query is represented in a source language and the collection of documents is represented in another target language. Thus, we extend the framework for studying the possibilistic query disambiguation and expansion applied on monolingual RI towards a CLIR framework. As a technical contribution, we proposed an architecture and implementation of a Possibilistic System for Query Expansion and Disambiguation (SPEEDSER) dedicated to WSD, QE and query disambiguation in monolingual IR and CLIR. A set of graphical user interfaces are included to assist the user in the reformulation task by exploiting the navigation module in the Hierarchical Small-World (HSW) network dictionary graph for monolingual IR. The system also offers a concordance analysis of terms as well as a search for translation candidates. These functionalities aim to assist the user in bilingual text analysis in both French and English language.
Complete list of metadata

Cited literature [198 references]  Display  Hide  Download
Contributor : Oussama Ben Khiroun Connect in order to contact the contributor
Submitted on : Wednesday, January 16, 2019 - 2:39:00 AM
Last modification on : Wednesday, October 28, 2020 - 12:48:03 PM


Files produced by the author(s)


  • HAL Id : tel-01982805, version 1


Oussama Ben Khiroun. Recherche d'Information Monolingue et Translinguistique : de la Désambiguïsation vers l'Expansion Sémantique de Requêtes. Informatique [cs]. Ecole Nationale des Sciences de l'Informatique (ENSI), Université de la Manouba, 2018. Français. ⟨tel-01982805⟩



Record views


Files downloads