An Improved Text Sentiment Analysis Algorithm based on TF-Gini

With the development of social media, more and more people prefer to express their opinions on the Internet. Therefore, developing a way to mine people’s emotional attitudes has become an important area of research. Text sentiment analysis is a method to mine people’s emotional attitudes through texts and an effective tool to grasp Internet users’ emotional tendencies. Naïve Bayes is a reliable text classification algorithm that has been approved by many researchers. Feature weighting is the most important problem for Naïve Bayes. Hence, this paper proposes an improved feature weighting algorithm, entitled TF-Gini, to enhance the performance of Naïve Bayes. The experimental results demonstrate the effectiveness of the improved algorithm.