A Sentiment Analysis Method Based on Emoticons and Sentiment Words

Weibo, a Twitter-like online social network in China. The statistical analysis indicates that, most weibo users will use the emoticons when they release tweets. And it is more intuitive to observe the user's sentiment attitude through these emoticons. The purpose of this paper is dividing the weibo text into four categories: happy, sad, angry and disgust, according to the expression of emotion and attitude. This paper uses the Naive Bayes classification method, in the division on the training set, collect 165 commonly used emoticons and 200 sentiment words.According to the emotion intensity that the emoticons and emotion words expressed, give various weights. By computing the emoticons and sentiment words' weighted average value, determine the classification. At the same time, use the category balance method to keep the proportion of four categories balance in the training set, to prevent classifier deviation. Through the experiments, the classification accuracy can reach 78.89%. Indicate that this method can get a better result in weibo text sentiment classification.