Método para representação de conceitos por meio de técnicas de análise de textos em sequencia temporal

• Ano de defesa: 2017
• Autor(a) principal: Figueiredo, Miguel Angelo Zaccur de
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia Civil; UFRJ
• Palavras-chave em Português: Engenharia civil; Sistemas computacionais; Séries numéricas; CNPQ::ENGENHARIAS::ENGENHARIA CIVIL
• Link de acesso: http://hdl.handle.net/11422/9772

This thesis consists of a method for representing concepts based on techniques of text mining in temporal sequences of documents. The presented method results in a computational model that identifies certain key expressions in a text, that somehow, are related to a concept. The main objective of elaborating a method for representing concepts of any nature, from a set of selected linguistic terms, has been reached. Twenty-five experiments were carried out on three distinct areas. After analyzing the results, in 80% of cases where there is an original reference series, the series generated followed the behavior of the original; in 13% of the cases the series generated anticipated the behavior and in 7% the series generated reflected the behavior later. In the experiments where there was no original reference series, in 90% of cases there was indication of success of the method, and in 66% of these cases this indication can be verified.