Exploring human emotion via Twitter

Sentiment analysis or opinion mining on twitter data is an emerging topic in research. In this paper, we have described a system for emotion analysis of tweets using only the core text. Tweets are usually short, more ambiguous and contains a huge amount of noisy data, sometimes it is difficult to understand the user's opinion. The main challenge is to feature extraction for the purpose of classification and feature extraction depends on the perfection of preprocessing of a tweet. The preprocessing is the most difficult task, since it can be done in various ways and the methods or steps applied in preprocessing are not distinct. Most of the researches in this topic, have been focused on binary (positive and negative) and 3-way (positive, negative and neutral) classifications. In this paper, we have focused on emotion classification of tweets as multi-class classification. We have chosen basic human emotions (happiness, sadness, surprise, disgust) and neutral as our emotion classes. According to the experimental results, our approach improved the performance of multi-class classification of twitter data.