Construction Of Intelligent Tourism Platform With Aspect Sentiment Analysis For Product Reviews

With the integration of Internet technology and tourism,tourists often browse the reviews of the products to decide whether to buy them or not when purchasing travel products.These authentic reviews often contain information about tourists’ shopping experience,product quality and after-sales service.Aspect sentiment analysis technology obtains aspect sentiment polarity in comment sentences,helps tourists quickly understand the advantages and disadvantages of products,and assists merchants to improve services.Accurately identifying aspects in sentences and deeply mining the dependencies between words in sentences are the key steps of aspect sentiment analysis.In this paper,the deep learning model and attention mechanism are combined to study the two sub-tasks under aspect sentiment analysis,and the intelligent tourism platform is constructed and the product review and analysis function is realized.The details are as follows:(1)Proposed a method of aspect category detection based on attention mechanism.First,the BERT pre-training model is used to accurately encode the Chinese comment text containing semantic relations into word vectors,then the deep learning model Bi LSTM is used to deeply extract hidden features,and finally the attention mechanism is used to use the complete semantic relations to prevent the loss of important information.In Sem Eval The experimental model accuracy and F1 value on the 2016 dataset reached 86.89% and85.96%,respectively.(2)Propose an aspect category sentiment analysis method based on interactive attention mechanism.First,use the pre-trained model Ro BERTa to convert the aspect category and context information into vector representations,then use the Bi GRU model to extract the aspect category and context to extract the hidden text features,and finally use the interactive attention mechanism to calculate the impact of the aspect category on the comment text.After connecting with the influence of the comment text on the aspect category,it is input to the fully connected layer for sentiment classification.The experimental accuracy and F1 value reached 90.65% and 89.71% on the Sem Eval 2016 dataset.(3)An intelligent tourism is constructed and the product review and analysis function is realized.The intelligent tourism platform includes modules of scenic spots,optimization,delicacies,routes and so on.The proposed algorithm is used to conduct aspect category sentiment analysis for commodity reviews in the optimization module to further improve tourist experience and verify the effectiveness of the model.To sum up,this paper mainly conducts researches on aspect category detection and aspect category sentiment analysis for commodity reviews,constructs an intelligent tourism platform and realizes the function of commodity review analysis.The affective polarity of the aspect category is judged by the product reviews in the platform,which further verifies the performance of the model.