The Design And Implementation Of Personalized Book Recommendation Algorithm Based On Sentiment Analysis

With the improvement of living standards,the popularity of electronic products has made digital reading accessible,and people are paying more and more attention to digital reading platforms.However,most of the current recommendation algorithms for digital reading are based on scoring data,without considering comments with high degree of recognition.And the active scoring users are still in the minority,which will cause the sparsity of the scoring matrix.Aiming at the problem of ’information overload’ caused by the uneven quality of books,the wide variety of books and the excessive homogenization of books,this paper introduces the sentiment analysis of comments into the existing recommendation algorithm to obtain the evaluation of the whole book of users and the fine-grained preference of users,and improve the accuracy of recommendation.The main research contents are as follows:The corpus of the experimental data in this paper comes from ’Douban Reading’.In order to make the best use of the books and user-related information contained in the corpus,this paper proposes a personalized book recommendation algorithm based on sentiment analysis.This method includes two parts:positive book recommendation based on sentiment analysis and personalized book recommendation based on sentiment matching.In the praise book recommendation module based on sentiment analysis,A XL-BM deep learning model based on XLNet embedding layer,multi-head self-attention mechanism and bidirectional long short-term memory neural network is designed.The XL-BM deep learning model uses the XLNet Embedding layer to vectorize comments;capture bidirectional long-distance semantic dependencies through the BiLSTM layer;the multi-head self-attention layer calculates the attention of BiLSTM multiple times to further capture the long-distance semantic dependence and highlight the importance of contribution words to context semantics.Then,based on the predicted sentiment value of each comment,the review time and review utility factors are fused to recommend the user to the book.The feasibility of the book recommendation method designed by the performance index recall rate,accuracy rate,F1 value and coverage rate is evaluated.In the emotional matching book personalized recommendation module,a book personalized recommendation method based on LDA probabilistic topic model is designed.According to the comments of users on the book,the user is modeled to clarify its fine-grained preference,and the LDA probabilistic topic model is improved to generate corresponding labels for the praise books generated in the praise book recommendation module based on sentiment analysis.The correlation score between the fine-grained preferences of users and the book labels are calculated,and the preferences of users and the book labels are emotionally matched.Finally,the recommended books can not only meet the user’s fine-grained preference,but also ensure the quality(praise),so as to realize the user ’s book personalized recommendation.Finally,based on the ’Douban Reading’ dataset,the effectiveness of the proposed method is verified,which also provides a reference value for the digital reading platform to a certain extent.