Social Media Text Sentiment Analysis Method Based on Comment Information Mining

Abstract

Emotion analysis can clarify the semantic orientation of emotions in social media texts. In order to improve the ability of emotion analysis of social media texts, a method of emotion analysis of social media texts based on comment information mining is proposed. Firstly, preprocess social media comment information to better explore the emotional information within it. Then, based on the linguistic features at different levels of social media texts, including auxiliary features, word level linguistic features, phrase level linguistic features, and sentence level linguistic features, rich linguistic features are extracted. In order to calculate the emotional polarity score of social media texts, the emotional polarity score of characters was considered, and the emotional polarity score of words was combined. At the same time, a sentence level sentiment score influencing factor has been introduced to more accurately calculate the sentiment score of social media texts. Through these calculations, emotional scores of social media texts can be obtained for emotional analysis. In order to better represent the emotional polarity of social media texts, a representation model integrating semantic knowledge was adopted. By integrating semantic knowledge with social media texts, the probability value of emotional polarity in social media texts can be calculated, achieving accurate analysis of social media texts’ emotions. The experimental results show that the method in this paper has a higher accuracy rate for social media text emotion classification.