Sentiment Classification of Online Reviews Based on LDA and Semantic Analysis of Sentimental Words
Yongxia Jing, Heping Gou, Fu Chuanyi, Wei Sun · 2019
Online reviews can reflect the sentimental tendency of the reviewers, and sentiment classification is one of the most important methods to realize sentimental tendency analysis. It is very important that how to select the features that reflect the sentimental information of the document and have good classification ability. In this paper, an online review sentiment classification method based on topic analysis of text document and semantic analysis of sentiment words is proposed. First of all, the topic of text documents and topic of words are got based on LDA method. Then, in order to get the vector representation of words in text documents and the words in sentiment dictionary, Word2Vec is used for word vector training. The semantic similarity between words and documents and semantic similarity between words in text document and words in sentiment dictionary are calculated. Finally, the words with maximum similarity are selected as key classification features of text document by using the two kinds of semantic similarity. Experiments indicate that the algorithm has a good performance and can improve the sentiment classification effect of online reviews.