Mineração de mídias sociais como ferramenta para a análise da tríade da persona virtual

• Ano de defesa: 2016
• Autor(a) principal: Lima, Ana Carolina Espírito Santo
• Orientador(a): Silva, Leandro Nunes de Castro
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Presbiteriana Mackenzie
• Palavras-chave em Português: mineração de dados; mineração de mídias sociais; comportamento social; persona virtual; sentimento; personalidade; temperamento
• Área do conhecimento CNPq: CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://dspace.mackenzie.br/handle/10899/24278

Understanding the human being is a continuous work of perception and inference about how he/she interacts and responds to various environmental stimuli in which he/she is inserted. Each person behaves based on how he/she sees and reacts to the world. However, the Internet has created an environment in which behaviors are also expressed and social media constantly stimulate the creation of social ties and the sharing of information. Identifying and mapping patterns from this data opens up opportunities to understand what is the persona that one wants to express within such environment, leading to what was named here the virtual persona. Therefore, data mining techniques are powerful tools for data exploration and analysis, making it possible to develop a computing infrastructure that allows to infer psychological aspects of the virtual persona. All this technical and computational framework for social media data analysis makes up the social media mining field of research. Motivated by this perspective, this thesis proposes computational frameworks to analyze three aspects of the virtual persona: sentiment; temperament; and personality. As a result, it was developed the Virtual Persona Triad, composed of frameworks for sentiment analysis, temperament prediction and personality identification. We assessed various techniques to represent text and classification algorithms, analyzed via social media mining methods capable of providing the desired inferences about the Triad. The results show a higher predictive ability for the category dictionaries combined with ensembles of classifiers.