Ontology Learning from Text using Relational Concept Analysis

Abstract : We propose an approach for semi-automated construction of ontologies from text whose core component is a Relational Concept Analysis (RCA) framework which extends Formal Concept Analysis (FCA), a lattice-theory paradigm for discovering abstractions within objects x attributes tables, to the processing of several sorts of individuals described both by own properties and inter-individual links. As a pre-processing, text analysis is used to transform a document collection into a set of data tables, or contexts, and inter-context relations. RCA then turns these into a set of concept lattices with inter-related concepts. A core ontology is derived from the lattices in a semi-automated manner, by translating relevant lattice elements into ontological concepts and relations, i.e., either taxonomic or transversal ones. The ontology is further refined by abstracting new transversal relations from the initially identified ones using RCA. The results of an initial validation of the approach through an application to astronomy texts are reported.
Type de document :
Communication dans un congrès
International MCETECH Conference on e-Technologies - MCETECH 2008, Jan 2008, Montréal, Canada. 2008
Liste complète des métadonnées

https://hal.inria.fr/inria-00322007
Contributeur : Mohamed Rouane-Hacene <>
Soumis le : mardi 16 septembre 2008 - 14:15:38
Dernière modification le : jeudi 24 mai 2018 - 15:59:20

Identifiants

  • HAL Id : inria-00322007, version 1

Citation

Amine Mohamed Rouane Hacene, Amedeo Napoli, Petko Valtchev, Yannick Toussaint, Rokia Bendaoud. Ontology Learning from Text using Relational Concept Analysis. International MCETECH Conference on e-Technologies - MCETECH 2008, Jan 2008, Montréal, Canada. 2008. 〈inria-00322007〉

Partager

Métriques

Consultations de la notice

220