摘要：Objective: To present a framework for the mining of opinions that can be applied in the discovery of knowledge of the customers about to their experiences, based on unstructured data extracted from social networks, and that is applicable to the reality of small and medium enterprises. 3) Methodology: This experimental research accessed data from the opinions of customers of four restaurants published in the social network TripAdvisor Brazil. The framework was based on the proposals formulated by Aranha (2007) and Feldman and Sanger (2007), techniques for Sentiment Analysis by Liu (2012) and Pang and Lee (2008) and Topic Modeling by Blei et al. (2012). 4) Originality: The relevance consists in proposing a solution that is both accessible to SMEs and capable of processing opinions in Portuguese, something not very common in literature. Almost all similar applications in literature are dedicated to the English language. 5) Main results: We highlight the generation of summaries and graphic visualizations that contribute to evidence knowledge about the relations between several expressions and terms that were not obvious. These allowed finding latent relationships between terms cited by different customers. 6) Theoretical contributions: The methodological solution uses efficient and state-of-the-art techniques and methods to extract, process, and analyze customer opinions on the Internet quickly, efficiently, and economically. 7) Social contributions: the framework developed presents an efficient, fast and economical way to mine data, presenting the results of the discovery of customer knowledge through the use of Sentiment Analysis and Topic Modeling techniques.
