Uma proposta para descoberta automática de relações não-taxonômicas a partir de corpus em língua portuguesa

• Ano de defesa: 2012
• Autor(a) principal: Ferreira, Vinicius Hartmann
• Orientador(a): Vieira, Renata
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação
• Departamento: Faculdade de Informáca
• País: BR
• Palavras-chave em Português: INFORMÁTICA; ONTOLOGIA; PROCESSAMENTO DA LINGUAGEM NATURAL
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://tede2.pucrs.br/tede2/handle/tede/5207

The construction of ontologies is a complex process that includes steps such as extraction of domain concepts, as well as the extraction of taxonomic and non-taxonomic relations between these concepts. The step of extracting non-taxonomic relations is the most neglected, specially for texts in portuguese. This dissertation presents a proposal for extracting non-taxonomic relations from texts in portuguese (corpora). These texts are represented by a list of domain concepts and contextual informations extracted by the tool ExATOlp. An application of the proposed process was performed with corpora of five domains and analysis on the relevance of the concepts, the specificity of relations and relations extracted application was made. Through this analysis, the proposed process seemed to be relevant and is considered the main contribution of this dissertation. Additionally, a tool for visualizing the extracted non-taxonomic relations, useful for various linguistic applications, is also proposed.