Tracking Sentiment Evolution on User-Generated Content: A Case Study on the Brazilian Political Scene

Opinion mining, which aims the automatic processing of subjective information, has become a key field. In the political context, the awareness of people’s sentiment towards their representatives, public organizations, political parties or politicians, can support political decisions, campaign moves, government policies, marketing strategies, etc. This paper describes a case study over opinions expressed about politicians as a reaction to news. Our ultimate goal was to detect whether user comments on an on-line newspapers reflect external indicators of public acceptance (e.g. vote intention). The paper outlines the approach used to identify and classify sentiment in news comments written in Portuguese language and to correlate it to external indicators, and discusses the main results for this case study.