Text Sentiment Analysis Based on Similarity and Cloud Model

Text sentiment analysis is an important research topic in natural language processing field. There are many problems in existing research methods due to semantic incoherence and independent topics. To address these problems, a comprehensive rating model is proposed in this paper for text sentiment analysis. It integrates topic similarity with cloud model theories. Firstly, LDA model is adopted to extract topic features. A sentiment revision algorithm based on cloud model is proposed to optimize the word sentiment polarities. Secondly, cosine similarity theory is introduced to measure the potential correlation of topics. Then, aspect segmentation algorithm is performed to obtain feature matrix and feature weight of reviews. Finally, feature parameters are optimized and the feature ratings of the reviews are concluded on the basis of above results.