The Sentiment Analysis Of Comments From Catering Enterprises On Dianping.com

With the rapid development of the Internet and 020 industry,020 enterprises have stored large amount of user data,and how to make use of these data has become the focus of research.Because of the development of the Internet Plus in Catering industry which is one of the 020 industry,more consumers begin to get used to getting access to service of catering enterprise through online reviews and then make decisions whether to consume in it.Therefore,it becomes the key point that how to make consumers quickly find out the their concerns from the massive online comments,and how to make the owner of catering enterprise realize the weaknesses of their own business scope to improve the business situationThis paper applies the the mature methods of text analysis to the comments in the catering field.Firstly,it constructs a new feature of VSM which can obviously reduce the dimension of the input data,and this new feature based on the traditional TF-IDF feature,LogRatio feature and the word embedding of word2vec Furthermore,it achieves a stable classification model with the common machine learning methods to predict the emotional polarity of the text.Secondly,based on the understanding of the catering business,this paper constructs a dimension rule based on syntactic analysis to mine the comment data,for deeply understanding the commenters' preference on different dimensions of the catering enterprise.Finally,this paper defines the formula based on consumers' preference on catering enterprises and displays the portraits of catering enterprises in the form of radar charts.These charts not only play an important role on helping consumers to choose the right catering enterprises,but also help owners of catering enterprises to improve their own operations in a timely manner.The experimental results in this paper show that the LogRatio feature performs better on text classification than the TF-IDF feature in the catering review data,while the classifier based on the LogRatio-weigh ted word vector feature and the support vector machine perform well on sentiment data analysis,of which the recall rate and precision rate of the model on the testing dataset have achieved 93%and 92%,respectively,and the effect of the model is stable.