Restaurant Business Evaluation System Based On Text Mining

The popularity of the Internet has led to the transformation of the traditional catering industry into the e-commerce industry.More and more consumers choose to order food on the Internet,and the resulting online reviews are also increasing.The analysis of consumer customer reviews can solve the problem of asymmetric information between catering merchants and consumers,so that merchants can understand consumers more and consumers can personalize choices.There are not many platforms for comment processing on the market at present,and some of the evaluations are not very detailed.For example,Meituan has counted the topics mentioned in the comments,and sorts the text according to the tags marked by the user.Based on the analysis of structured data and unstructured data of catering businesses on a platform,a text-based mining catering business evaluation system was developed.The research contributions include the following four aspects.Firstly,the business evaluation data preprocessing for the purpose of speed up.Using the business reviews on the catering website as research data,the data is classified and stored according to the characteristics of the network comment data and the data dimension,and the theme model according to the city and the food category is designed.Compared with the original storage,the storage method has greatly improved the speed of word segmentation and sentiment analysis,and the data query efficiency of the system function has been improved.Secondly,topic extraction with the goal of getting accurate,comprehensive business topics.Through the feature reduction processing data,for the inconsistency of different categories of keyword words,LDA topic extraction is carried out with different categories of business comment data,and then the first-level keyword of the restaurant is screened according to word frequency and co-occurrence degree;in order to avoid Chinese synonyms The comprehensiveness of expressions and subject words,using Word2 vec similar words to extract secondary subject words.This method of extracting themes includes a common theme for all meals,ensuring the comprehensiveness of the keywords.Thirdly,the sentiment analysis designed for the purpose of accurately evaluating the business.For the identification of multiple topics in a sentence,the system uses the dependency syntax analysis to analyze the comment text and extract the topic-emotion relationship pairs in the sentence.Combined with the HowNet sentiment dictionary to calculate the emotions corresponding to the theme,the calculation method of each topic score and the comprehensive score of the merchant is designed.The calculated score is compared with the score on Meituan website,and the calculation results are consistent,which proves the accuracy of the merchant evaluation.Finally,the system uses the B/S architecture and uses Web-related technologies to develop the catering business evaluation system.According to the score results,the functions of business inquiry,business comparison and satisfaction analysis are designed to meet the needs of current users and businesses.