Análise de sentimentos no Facebook utilizando o Senticnet

Abstract

Know the opinion of people on the Internet has been object of study in many recent researches, seeing that, more than ever, these opinions have gained large space in blogs, forums and social networks. Companies invest a lot of money to collect this important data about people in order to develop marketing strategies and targeted sales. Social networking has been having a significant growth in users. People are increasingly connected to these networks to express their opinions, tastes, preferences and relationships. The study of opinions, feelings and emotions expressed in documents is known as sentiment analysis. Among some tasks of the sentiment analysis, stands out the polarity classification into texts, in other words, just how positive or negative are the opinions expressed therein. This work aims the study, implementation and evaluation of a system of classification of posts, collected through surveys on Facebook, using the SenticNet. The system is able to collect messages and break them by words, by searching the database of SenticNet and obtaining the value and sense of the polarity of the message. To evaluate the proposed solution, a questionnaire was given to some people in order to make a comparison between the results collected in this questionnaire and submitted by the system developed and able to measure system efficiency. The results were positive, since the system managed to hit 70.4% of the messages and make adjustments as future work and get better this value to a value closer to the ideal.