Descoberta de relacionamentos semânticos não taxonômicos entre termos ontológicos

• Ano de defesa: 2017
• Autor(a) principal: Andrade, Arthur Morais de
• Orientador(a): Santos, Marilde Terezinha Prado
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação - PPGCC
• Palavras-chave em Português: Ontologia; Extração de informação; Processamento da linguagem natural
• Palavras-chave em Inglês: Ontology; Information extraction; Natural language processing
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/8946

Ontologies have become an important tool to structure knowledge. However, the construction of an ontology involves a careful process of defining representative terms of the domain and its relationships, which requires a lot of time from ontology engineers and domain experts. These relationships can be taxonomic (hyponymy and meronymy), representing a taxonomy of concepts, and non-taxonomic, referring to the other relationships that occur between the nodes of this taxonomy. The main difficulties of constructing an ontology are related to the time spent by domain specialists and the necessity of guaranteeing the quality and reliability of the ontologies create. In this way, we are welcome the efforts to elaborate approaches that aim to reduce the amount of time dedicated by specialists without reducing the quality of the ontology created. In this master's project, an approach was developed for the discovery of semantic relationships between non-taxonomic ontological terms from semi-structured documents written with informal vocabularies of the Brazilian Portuguese language. Thus, it aids ontology engineers and domain experts in the arduous task of discovering the relationships between ontological terms. After the discovery of semantic relationships, the relationships were converted into a conceptual structure, generated by the Formal Concept Analysis (FCA) method. This approach was validated in two experiments, with the help of domain experts in special education. The first experiment consisted of a comparison between manually extracted relationships and automatic extraction, presenting a good value of precision, coverage and measurement F, respectively, 92%, 95% and 93%. The second experiment evaluated the relationships extracted, automatically, in the structure generated by the FCA, it gets average accuracy 86,5%.These results prove the effectiveness of the semantic relationship discovery approach.