Research On Aspect Sentiment Analysis Technology Of Tourists’ Catering Reviews And Implementation Of Intelligent Tourism Platform

As one of the six elements of tourism,food plays an important role in tourism activities.The rapid development of information technology has brought new development opportunities for the tourism industry.Online catering platforms have become an important way for tourists to obtain information and express their opinions,and a large number of comment texts have been generated in the process.It is very difficult for users to directly obtain effective information from massive data.The sentiment analysis technology can automatically extract the opinions contained in the comment text.Different aspects of emotional expressions can be obtained through aspect mining and sentiment analysis of review texts,providing rich decision-making reference information for tourists in unfamiliar tourism environments.Based on the tourist catering reviews texts,this paper conducts research in three aspects:aspect word extraction,aspect sentiment analysis,and practical application,as follows:(1)An aspect word extraction method based on word embedding is proposed.First,the comment text is labeled by BIOES through the sequence labeling method,and then the RoBERTa model is used to encode the comment text to obtain a vector with contextdependent information.Finally,the relationship between labels is learned by the transfer score matrix in the CRF layer,reduces the number of invalid labels,and selects all possible labels.The most plausible predicted sequence in the sequence path.The experimental results show that the word embedding method can solve the problem of aspect word extraction.(2)A cross-attention-based aspect sentiment classification method is proposed.First,use RoBERTa as the embedding layer to convert the review text and aspect words into word vectors,respectively,then obtain the deep semantic features of the review text and aspect words through BiLSTM,and finally use cross attention to jointly learn the representation of the aspect words and the review to obtain the aspect words and The interactive attention of the review text captures the relationship between the review text and the aspect words,and the updated review text and aspect word vector representations are multiplied with the output of the BiLSTM model,and input to the fully connected layer to obtain the final classification prediction.The pipeline method divides the sentiment analysis task into two independent parts,ignoring the potential connection between the tasks,and proposes aspect extraction and aspect sentiment classification based on the RoBERTa joint model,which effectively alleviates the problem of error propagation and context loss in the pipeline.(3)Build a intelligent tourism platform and implement the text analysis function of tourist catering reviews.The platform integrates functional modules such as scenic tours,hotels,food,travel,routes and parking services.On this basis,the model proposed in this paper is applied to the food module of the platform to perform aspect word extraction and sentiment analysis on tourist catering reviews review texts to verify this paper.The practicality and effectiveness of the proposed model.