Emotional Analysis Of Online Comments On Yan’an Tourist Spots Based On Deep Learning

With the continuous development of the Internet,online ticket purchase has become the mainstream for purchasing ticket for tourist spots,so a large number of tourist spot review data will be generated.The emotion behind rich comments has significant reference value for improving the quality of tourist spots and making decisions for potential tourists.It takes a lot of manpower,time,and cost to summarize the information contained in massive comment data only by manual work,and there are also problems that everyone has different subjective opinions and can’t get the only definite emotional tendency,and artificial intelligence technology which has fast speed and accurate detection results can effectively solve these problems.This dissertation finds the following problems by studying the online review data of tourist spots in Yan ’an: The emotional tendency contained in some review data is inconsistent with the star level;the emotional tendency of comments is inconsistent with that of additional comments;At present,most emotional analysis models of tourist spots mainly train the models with star level as emotional tendency label,without considering the above problems.Therefore,it is worthy of research to construct the emotional dictionary of the comments for tourist spots in Yan ’an and re-label the emotional tendency of the comment data.ERNIE-Bi LSTM-DPCNN emotion analysis model is constructed,which improves the performance in the face of emotion analysis tasks.The main work of this dissertation is as follows:1.The data set of comments for tourist spots in Yan ’an and the emotional dictionary for tourist spots in Yan ’an have not yet been made public at present.This paper will use the TF-IDF and SO-PMI algorithms to construct an emotional dictionary for Yan’an tourist spots comments.The constructed emotional dictionary is used to mark the emotional tendency of the comment text,and manually revise the emotional tendency that is not consistent so that the high-quality data set of Yan’an tourist spots comments will be obtained.2.By studying the current emotion analysis algorithm,this dissertation uses ERNIE word vector expression to replace the word vector expression of DPCNN model and combines DPCNN model and Bi-LSTM model for feature extraction,combined with their respective advantages,designs an emotion analysis model based on network structure.3.The constructed data set of comments for tourist spots in Yan ’an is used as the experimental data set,and it is continuously trained and tested on the ERNIE-Bi LSTMDPCNN model,and compared with other models,the accuracy,recall,and F1 value of ERNIE-Bi LSTM-DPCNN model are improved,all of which are above 90%,thus verifies the effectiveness and feasibility of ERNIE-Bi LSTM-DPCNN model.