Using Character N-gram Features and Multinomial Naïve Bayes for Sentiment Polarity Detection in Bengali Tweets

Abstract

Sentiment is meant by feelings-attitudes, emotions and opinions. Sentiment polarity detection is one of most popular sentiment analysis tasks. Now-a-days, celebrities as well as common people write a huge amount of blog posts, tweets and comments on the social media. Such social media texts are also written in Indian languages. The research on sentiment analysis in Indian language domain is also at the early stage. In this paper, we present a sentiment polarity detection approach that detects sentiment polarity of Bengali tweets using character ngram features and a supervised machine learning algorithm called Multinomial Na\{\ive Bayes. Our proposed approach has been tested on the SAIL 2015 dataset. The experimental results show that character n-gram features are more effective than the traditional word n-gram features. The overall performance of our proposed system is also significantly better than some existing sentiment polarity detection systems.