Twitter sentiment analysis using multi-class SVM

Abstract

Sentiment Analysis or Opinion Mining is a field which is used to study the user's opinions towards a product or an organization or a person. Twitter, a micro-blogging service allows the user to share their opinions and aspects of life. Sentiment analysis on twitter is a big challenge because it has diverse topics. The classifier that is trained on a specific topic will not perform well on another topic. So, the topic adaptive training method is proposed to address this problem. In this method, non-text features are also extracted from tweets for training the algorithm. The algorithm classifies the tweets of different topics as positive, negative, neutral. The proposed method is evaluated across different topics and it outperforms in terms of recall, precision and F-score.