Short Text Topic Mining Of Hotel Comments Based On Emotional Classification

With the rapid development of social networks in recent years,online users have increasingly published their own comments and opinions on the Internet.These comments and opinions have exploded,including users information such as emotions,which are important in public opinion,monitoring,product improvement and product recommendations.Text sentiment classification research conforms to this development trend and has become a research hotspot.With the development of social economy and the people's living standards,tourism customers pay more attention to the feelings of the spiritual level while paying attention to the service experience.After the customer completes the experience,the comments posted on the web become an important reference for future users to choose a hotel.The economical express hotel has become the first choice for people to travel because of its low price and wide distribution.This paper first obtains the comments of the Home Inn and Hanting Hotel in Hangzhou from the Ctrip website through the Octopus Reptile Tool.Secondly,according to the original corpus,through the steps of data cleaning pretreatment,etc.,18685 articles are reviewed at Home Inns,and 19690 articles are reviewed by Hanting Hotels.Then,based on the statistical dictionary,a custom hotel sentiment dictionary is constructed,the comment text is processed by word segmentation,and the word vector is trained using the word segmentation result.By the way,the word vector with the emotional label comment is used as the input variable to train the TextCNN,and the trained network model is used to emotionally classify the comments without the emotional label.There were 14547 positive comments from Home Inns and 4138 negative comments.The negative comments accounted for 22.15%;Hanting Hotel is positively commented on 14390 articles,negative comments on 5308 articles,and negative comments accounts for 26.95%.Finally,LDA theme mining is carried out on the texts of the positive comments and negative comments of the two hotels.The advantages and disadvantages of the two hotels are analyzed by LDAvis visualization,and some suggestions are put forward.The final analysis results show that from the perspective of research methods,the TextCNN performs very well in text classification,and the classification results are remarkable.The LDA theme model has unique advantages for analyzing the problems existing in the two hotels,such as Home Inn and Hanting.From the perspective of hotel development,the advantages of the two hotels,Home Inn and Hanting,outweigh the disadvantages.The feedback from Home Inns is better than that of Hanting Hotels.Positive comments from Home Inns show that Home Inns is fully equipped,has a wide distribution and has free room upgrades.Negative comments show that in addition to the common problems of budget hotels,Home Inns does not provide services such as breakfast and parking.The positive comments of the users of Hanting Hotel show that Hanting Hotel provides pick-up and drop-off service,the environment is comfortable,and the distribution is concentrated in the West Lake Scenic Area;while the negative comments point out more problems,except for the common problems of budget-based hotels,the facilities are relatively old,and insufficiency of toiletries and poor service attitudes have disappointed users.