Text Analysis Of Outbound Travel Comments Based On Text-CNN And LSTM

With the rapid improvement of people's living standards,the behavior of tourism has become an indispensable way to relax in life.More and more people like to travel abroad and are not satisfied with domestic tourism.At the same time,with the rapid development of e-commerce,more and more consumption is being carried out online.On the one hand,from the perspective of potential consumers,people can have a general understanding of the products they want to buy through the comments on the Internet,and can also make recommendations based on the evaluations of other users who have already purchased the products.decision making.On the other hand,from the perspective of the merchant,it is possible to further understand whether the group is satisfied with the purchased product or the experience during the purchase process through the commentary information left by the consumer,and the merchant can make corresponding adjustment according to the above feedback.To enhance word of mouth or income.This paper focuses on the theoretical basis of text analysis,and elaborates on topic mining and sentiment classification.It performs pre-processing and vectorization on the outbound travel comment text of mafengwo online travel website,and analyzes the three methods of text representation.The topics extracted from the theme of outbound travel commentary are analyzed and discussed accordingly,and the machine learning models such as Naive Bayes,Support Vector Machine,Logistic Regression,and two deep learning models of Text-CNN and LSTM are used to classify emotions.Empirical analysis,the final model comparison and the relevant conclusions and recommendations for the above analysis.This paper consists of five parts: The first part mainly elaborates the research background and significance,domestic and foreign literature review,research content,etc.,providing a basic idea for the full text.The second part introduces the data source of the text,cleans the text,and introduces the basic theory of text analysis,including Chinese word segmentation,LDA topic mining and word vectorization,and the topic feature extraction results for analysis and cleaned text.Pre-training was carried out.The third part uses TF-IDF and Word2 vec text vectorization techniques to extract the features of the text after segmentation,and then introduces three classic models for text classification,such as naive Bayes,support vector machine and logistic regression.The outbound travel comment text compares the classification effects ofthe three models.The fourth part introduces the convolutional neural network and the circulating neural network commonly used in the image field and the speech field,and applies the improved models Text-CNN and LSTM to the text field.The input layer uses Word2 vec and Glove training word vectors respectively.The sentiment classification research on the outbound travel comment text is compared with the third part of the machine learning model,and finally the two deep neural network structure models Text-CNN and LSTM are significantly better than the traditional shallow machine learning.Model,support vector machine with the highest accuracy rate of 82.87%,LSTM highest accuracy rate of 84.49%.The fifth part is the conclusions and suggestions,summarizing the relevant conclusions of the text analysis of the outbound travel comments and proposing targeted suggestions.