Research On Topic Extraction And Sentiment Analysis Of Restaurant Reviwes

With the development of the Internet industry,human society's production and life style are deeply influnced in many aspects,Text data from different areas are produced explosively.The total amount of these texts is growing very large which contain lots of important information.Therefore,how to extract useful information from these text data accurately and quickly has important practical value.Topic extraction and sentiment analysis are commonly used methods to extract text information,and recommendation system model is an important field of text information application.However,because of the diversity of network text structure,writing styles,and the complexity of Chinese language,the Chinese text information processing is a big challenge.Therefore,in the current recommendation system model,the use of Chinese text information is also very limited.To solve the above problems,this paper focus on the personalized recommendation model based on the restaurant user review text topic extraction and sentiment analysis algorithms.Most of the traditional personalized recommendation systems build user's preference model based on historical interaction data between users and products,and then recommend information based on the user's interest model.With the improvement of the Internet industry,the huge online text review data can intuitively reflect the user's behavioral preferences.The use of such information has very important practical value for the personalized recommendation model.The extraction of web-based text topics can understand the product features that the user is concerned with,and the sentiment analysis can reflect the user's degree of likes and dislikes to the product features.Compared with the user preference analysis from the historical behavior,it is more intuitive and accurate to obtain the user's preferences from the user reviews.This paper consists three parts:first,the original reviews are processed to construct the clustering semi-supervised expansion model to get the topics that the user pays attention to.Then,a special sentiment dictionary is constructed by integrating existing universal sentiment lexicons and comment texts.The special sentiment dictionary is used to analysis the sentiment of comment texts which represent the user's preference information.Finally,the user's preference information is used to personalized recommendation.Experiments using the public restaurant reviews text data from comment website shows that the clustered semi-supervised expansion model proposed in this study has good result on topic extraction in the subdivided domain(reviews of restaurants in this paper);the construction of a special sentiment dictionary improves the accuracy of sentiment analysis.The infrastructure and algorithms of obtaining user's preference from the user's review text data for personalized recommendation has very important practical value in the field of personalized recommendation.