The Sentiment Analysis Of The Comments Of The E-commerce Goods

With the rapid development of the Internet, more and more users start to buy goods on the Internet. They always post many comments about the goods. These data can provide some useful information about the goods. However, the large number of reviews is often difficult for customers to extract useful information, so the key issue is that the sentiment analysis about the comments efficient.Sentiment analysis model of the paper based on the web site product reviews, it can accurately identify the attitude of the users, provides some information for the users, and also to provide feedback to the business. It also can be as one of the useful aspect in the website personalized recommendation.In the paper, first the author introduces the research background, research status, significance and related technologies. Then, the author proposes an unsupervised sentiment analysis model. The model introduces the emotional factor variables in the model, based on the original topic model, combined with hierarchical clustering results for secondary clustering. The paper mainly completed the following work:(1) Data de-duplication and cleaning.(2)Build the word vector model; According to word to vector model to extend the representation of the documents to solve sparse problems;(3) Modeling and model training, then model the data with the sentiment analysis model, and the author uses Gibbs sampling algorithm to solve the model to obtain comments-emotion-theme-the distribution of the word, and then uses the hierarchical clustering algorithm to combine topics.(4) The emotional polarity value judgments and emotional tendencies analysis. Based on the result of the model, the author extracts the emotion words with emotion word dictionary, then get the polar and polar value of a word with four rules, after the procedure, the author can calculate the emotional scores of the reviews. And then get the score of each goods.(5) The results of the analysis and experimental verification, through the design of experiments and comparative experiments, the author uses the existing data sets for testing, and analysis the experimental results. The results show that the model can accurately determine the emotion of the goods.