| Resumo: | Playing a massive impact on how the general population interacts with each other, and how they access and share opinions on news, namely on themes of political nature, social media became part of the daily lives of society. Such influence encouraged the creation of a set of techniques to build profiles and identify patterns in opinion. Moved by this potential, the main purpose of this dissertation was to develop a system to perform Sentiment Analysis on posts from Twitter with political theme, categorizing opinions between Positive, Neutral or Negative. The collected information can than be grouped in different topics like organizations or people, using Named Entity Recognition methods, that enable filtering the information on relevant themes. The results were analysed on a dashboard that compiles the information, by displaying the data in various intuitive ways, allowing a more concrete comparison between different political themes. Identifying relevant and specific study cases on the stored data enabled a subjective review on the different Sentiment Analysis methods.
|
|---|