Construção de estruturas ontológicas a partir de textos : um estudo baseado no método formal concept analysis e em papéis semânticos

Abstract

This work aims to study conceptual structures based on the Formal Concept Analysis method. We build these structures based on lexico-semantic information extracted from texts, among which we highlight the semantic roles. In our research, we propose ways to include semantic roles in concepts produced by this formal method. We analyze the contribution of semantic roles and verb classes in the composition of these concepts through structural measures. In these studies, we use the Penn Treebank Sample and SemLink 1.1 corpora, both in English. We test, also for English, the applicability of our proposal in the Finance and Tourism domains with text extracted from the Wikicorpus 1.0. This applicability was extrinsically analyzed based on the text categorization task, which was evaluated through functional measures traditionally used in this area. We also performed some preliminary studies for a corpus in Portuguese: PLN-BR CATEG. In our studies, we obtained satisfactory results which show that the proposed approach is promising.