Um arcabouço para pesquisas de opinião em redes sociais

• Ano de defesa: 2014
• Autor(a) principal: Renato Miranda Filho
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Caracterização de usuários; Pesquisas de opinião; Redes sociais; Twitter; Opinião pública Pesquisa; Estudo de usuários; Computação; Redes sociais on-line
• Link de acesso: http://hdl.handle.net/1843/ESBF-9HXNZY

Extrapolating the initial intention of serving as means of interpersonal communication, social networks have evolved to be sophisticated sources of information that allow monitoring the opinion and the daily lives of millions of individuals. In this context, the main goal of this work is to use public data available from these networks to conduct opinion analysis in the same way the NIELSEN'' does. In order to do that, we take into account inherent caracteristics of virtual environments, including sentiment analysis of messages and the detection of spammers and journalistic content. This first analysis was combined with the generation of a representative sample of the target population of study. For the formation of this sample, the following characteristics of individuals were taken into account: (i) identification of unique users; (ii) gender; (iii) age; (iv) social class; and (v) geographical location. The experimental evaluation was performed with data from various contexts which results were validated by real agencies: intention to vote in municipal elections of 2012, and poll of Big Brother Brasil. The results show that more elaborate analyzes are able to improve the numbers achieved by methods used in other studies, mostly based on counting techniquesComparing the results achieved with traditional research methods (polls conducted by major sites and research conducted by renowned institudos) also achieved competitive results, showing that the technique developed can be considered as a good alternative.