| At present,online travel reviews are characterized by large quantity,mixed content and uneven quality,and it is often very difficult for users to find the information they are interested in among the mass text.To meet users’ travel information service needs,it is urgent to explore reviewers’ subjective emotions from the perspective of sentiment analysis,and find and properly solve the problems of poor attraction recommendation quality and review helpfulness caused by information overload.In this paper,we propose an alternative scenic spot recommendation algorithm that integrates sentiment analysis and intuitionistic fuzzy features,and constructs a data-driven helpfulness evaluation model,considering that intuitionistic fuzzy sets can accurately represent the uncertainty and fuzzy features of reviews,and deep learning technologies such as neural networks and attention mechanisms have strong feature extraction capabilities,and use these two technologies to fully explore online tourism review information.The working details are as follows:(1)A recommendation algorithm that integrates sentiment analysis and intuitionistic fuzzy features is proposed.In order to more accurately identify the sentiment information embedded in the comment text,a sentiment classification model based on BERT pretraining and attention mechanism is established.Based on the idea of feature fusion,the topic features identified by LDA are fused with the text features identified by BERT pretraining,and a collaborative topic and text analysis model is established to identify the sentiment tendency of different aspects of the review text.Considering that the sentiment information is fuzzy,it is transformed into intuitionistic fuzzy numbers according to the definition of intuitionistic fuzzy sets,and the decision matrix of alternatives is constructed,and the alternatives are ranked by applying the intuitionistic fuzzy TOPSIS method.Finally,taking Ctrip.com scenic spot reviews as the research object,the ranking results of alternative scenic spots are given by applying the method of this paper and the VIKOR method in multicriteria decision making,which verifies the effectiveness and feasibility of the method.(2)To solve the helpfulness voyage problem and help travelers automatically identify high-quality online reviews,a helpfulness evaluation model based on LSTM is proposed.The influencing factors of review utility are screened and quantified from aspects such as text features and other features of online reviews,and the importance of each influencing factor on review utility is identified by correlation analysis.Based on this,a helpfulness evaluation model was constructed using LSTM neural network to identify and study the review utility.Using accuracy,precision,recall and F1 value as the measurement criteria,the overall performance of the LSTM model was found to be better than other neural network models in terms of helpfulness recognition through comprehensive analysis of online reviews,review time and user ratings of Ctrip.com tourist attractions.In summary,the method proposed in this paper can effectively improve the efficiency and quality of decision making,which can provide important reference for users to make personalized travel decisions,and has positive significance to solve the problems of information overload and helpfulness disorientation. |
PDF Full Text Request