UNIVERSIDADE ESTADUAL PAULISTA
"JÚLIO DE MESQUITA FILHO"

Faculdade de Ciências e Letras - Campus de Araraquara
  Agenda Pós-Graduação - Lingüística e Língua Portuguesa

Aluno(a) Isa Mara da Rosa Alves
Titulo POLYSET: Modelo Linguístico-Computacional para a Estruturação de Redes de Polissemia de Nominais
Orientador(a)
Data 01/10/2009
Resumo Esta pesquisa visa a representar a polissemia de nominais de maneira útil a sistemas computacionais; mais especificamente, o tema deste trabalho é a inclusão da especificação de relações de polissemia em bases wordnets, em especial, na WordNet.Br (DIAS-DA-SILVA, 1996, 1998, 2003). A metodologia do trabalho está baseada em Dias-da-Silva (1996, 1998, 2003, 2006), abrangendo três domínios mutuamente complementares: o linguístico, o linguístico-computacional e o computacional. O domínio linguístico-computacional forneceu o tema para esta pesquisa e articulou a relação entre os domínios linguístico e computacional. Das investigações realizadas no cenário linguístico-computacional, destacamos a relevância da introdução de níveis distintos de generalidade entre os sentidos em uma base de dados de modo a reduzir a quantidade de processamento lexical a ser realizada pelo sistema. Ao mesmo tempo, essa representação múltipla fornece as informações necessárias a um sistema que necessite grau maior de detalhamento do sentido. Esse tipo de tarefa é ainda um desafio para as wordnets. Do cenário linguístico, destacamos que a Semântica Lexical Cognitiva se mostrou a teoria mais adequada aos propósitos desta tese. Olhar para o fenômeno do significado múltiplo sob o viés cognitivo possibilitou descrever os sentidos como uma entidade complexa, estruturada em termos de redes. As redes de polissemia sincrônicas, em sua configuração livre e multidimensional, conforme propõem Blank (2003) e Geeraerts (2006), demonstraram ser a estratégia descritiva mais adequada à representação da flexibilidade do sentido para os propósitos desta tese. Respondendo à fase aplicada dos domínios linguístico e linguístico-computacional, propomos um modelo de representação denominado polyset. Os polysets são constructos estruturados em termos de redes de polissemia, de maneira que possibilitam representar diferentes níveis de generalidade entre os sentidos, diferentes graus de saliência e diferentes tipos de relações de polissemia. Para dar conta do domínio computacional, propomos, seguindo a metodologia proposta por Ganguemi (2004), uma estratégia para a codificação dos polysets em OWL com auxílio do editor de ontologias Protégé-OWL. O modelo polyset representa uma tentativa de contribuir para a representação da polissemia de nominais no paradigma wordnet, em especial, para a WN.Br.

Palavras-chave: polissemia; redes de polissemia; relação de similaridade; automeronímia; autohiponímia; relações de sentido.


ABSTRACT
This research aims at representing noun polysemy so that it can be useful to computational systems; more specifically, the subject of this work is the inclusion of specification of polysemy relations in wordnet bases, particularly in WordNet.Br (DIAS-DA-SILVA, 1996, 1998, 2003). The methodology has been based on Dias-da-Silva (1996, 1998, 2003, 2006), comprehending three mutually complementary domains: linguistic, computational-linguistic, and computational ones. The computational-linguistic domain has both provided the subject for this research and articulated the relationship between the linguistic domain and the computational domain. From the investigations carried out in the computational-linguistic scene, we have highlighted the relevance of the introduction of distinct levels of generality among meanings in a database, so as to reduce the amount of lexical processing to be carried out by the system. At the same time, that multiple representation provides the necessary information for a system that needs a higher degree of meaning detailing. This kind of task is still a challenge to wordnets. From the linguistic scene, we have highlighted that Cognitive Lexical Semantics has shown to be the most suitable theory for the purposes of this thesis. Regarding the phenomenon of the multiple meaning from the cognitive perspective has allowed for describing meanings as a complex entity, structured in terms of nets. The nets of synchronic polysemy, in their free, multidimensional configuration, as Blank (2003) and Geeraerts (2006) have proposed, have shown to be the most suitable descriptive strategy for the representation of the meaning flexibility for the purposes of this thesis. Answering to the applied phase of both the linguistic and computational-linguistic domains, we have proposed a representation model called polyset. Polysets are constructs structured in terms of polysemy nets, allowing for representing different levels of generality among meanings, different degrees of salience, and different types of polysemy relationships. Following the methodology pointed out by Ganguemi (2004), in order to consider the computational domain, we have proposed a strategy for the codification of polysets in OWL with the help of the Protégé-OWL ontology editor. The polyset model represents an attempt at contributing towards the representation of noun polysemy in the wordnet paradigm, especially WN.Br.

Keywords: polysemy; polysemy network; similarity relation; automeronymy; autohyponymy; sense relation.
Tipo Defesa-Doutorado
Texto Completo

APG 2.0
Copyright 2014 (c) UNESP - Faculdade de Ciências e Letras do Campus de Araraquara