Mineração de opinião: análise das emoções em mensagens de reclamação dos consumidores

• Ano de defesa: 2020
• Autor(a) principal: Toledo, Laylla Pereira de
• Orientador(a): Balaniuk, Remis
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Católica de Brasília
• Programa de Pós-Graduação: Programa Stricto Sensu em Governança, Tecnologia e Inovação
• Departamento: Escola de Educação, Tecnologia e Comunicação
• País: Brasil
• Palavras-chave em Português: Mineração de texto; Mineração de opinião; Atendimento ao consumidor; Processamento de linguagem natural; Emoções
• Palavras-chave em Inglês: EmoLex; Mining opinion; Analysis of emotion; Natural language processing; Customer service
• Área do conhecimento CNPq: CNPQ::CIENCIAS SOCIAIS APLICADAS::CIENCIA DA INFORMACAO
• Link de acesso: https://bdtd.ucb.br:8443/jspui/handle/tede/3073

The objective of the research was to know the emotional behavior of consumers in the Complain Here platform complaints. The theoretical foundation was based on the literature review on opinion mining and emotion analysis. The case study analyzed free website data where consumers post their complaints about companies and their products / services, awaiting resolution. This system allows other users to track the reputation of the cited entities. Messages are public and only message authors have their identity preserved. The effort of this work was to extract the messages and to classify the emotional aspects of the opinions. The sample focused on complaints directed to banks and their attendance. From a quantitative approach, the research identified in the complaints the positive or negative orientation in the user's text and classified the emotional state of the corpus according to the EmoLex lexicon. The work was inspired by the research by Gonçalves (2016) that analyzed the feelings of consumers on the same website. The hypothesis was based on the possible variations of emotion in two different situations: when users opened the complaint call and when evaluating the service, after the online interaction. Finally, the hypotheses raised were considered viable from the results achieved, revealing post-care emotional variations. The work leaves a discrete contribution in the area of natural language processing, as it adopts the analysis of emotional aspects in Portuguese language texts.