Discovering Patterns in Sentimental Analysis

• Flavio Carvalho CEFET/RJ
• Leonardo F. dos Santos CEFET/RJ - UFF
• Heraldo Borges CEFET/RJ
• Eduardo Ogasawara CEFET/RJ
• Gustavo Paiva Guedes CEFET/RJ

Resumo

Users are continually using text-based social networks and messaging applications to express their sentiments throughout the day. Each post can be interpreted as an event with a message and a timestamp associated with it. With these considerations, interesting information can be extracted with pattern mining techniques. A question that arises from such techniques is: How strong are the correlations between the time of day and emotions? This work aims to analyze sentiment intensity variation in social network users. The objective is to evaluate if pattern mining techniques applied over events enriched with sentiment analysis metadata can be used to address such a question. The results show patterns with higher measures of quality between negative sentiments and the afternoon and late morning hours. These results also lead to the possibility of future work applied to health care and social market interventions.