What Tweets Containing Emoticons Can Reveal About Collective Feelings?

• Pollyanna Gonçalves Federal University of Ouro Preto
• Fabrício Benevenuto Federal University of Minas Gerais
• Virgílio Almeida Federal University of Minas Gerais

Abstract

Twitter has become an important social communication mean where users post messages about everything. Several social networking's users use emoticons, a pictorial representation of a facial expression, to try to represent the sentiment expressed in their messages. Potentially, when analyzed in large scale, messages containing emoticons could be used to measure public mood fluctuations at the level of populations. Such measurement could have several applications, such as to predict collective mood trends of populations related to specific topics as well as to understand the effects of social phenomena in people's feelings. In this paper, we give a first step on this direction by providing a deep understanding of how emoticons are used by Twitter's users, analyzing the most popular emoticons and how they appear associated to global scale events. In order to validate our approach, we used a unique dataset containing 1.8 billion tweets to evaluate public mood fluctuations on a series of events that span topics related to tragedies, releases, politics, health, as well sport events. Our results suggest that emoticons can correctly capture expected sentiments of messages of this social networking.

Keywords: Sentiment Analysis, Collective Feelings, Twitter