Research On Personalized Book Theme Recommendation Based On Deep Learning

With the rapid development of artificial intelligence and big data technology,the information age is showing a trend of intelligence.The traditional publishing and distribution industry needs to keep pace with the times and speed up the intelligent process of the industry.Nowadays,when readers are faced with a large number of publications,it is difficult to find something that meets their needs.Therefore,how to improve the recommendation technology and accurately recommend the books of interest for readers has become a key issue restricting the development of the publishing and distribution industry,and is of great significance to increase sales and reduce inventory.However,the amount of behavioral data of readers on major e-commerce platforms and book review forums is very rare relative to the number of published books,and the sparsity of reader-book data seriously affects the efficiency and accuracy of the recommendation algorithm.At the same time,after the publication of the new book,the effect of recommendation by calculating the word similarity between the content introduction of the new book and the introduction of the read book is not ideal,and there is a problem of cold start of the new book,so it is necessary to introduce factors such as readers’ interest preference into the recommendation process to improve the accuracy of recommendation.The field of personalized book recommendation is mainly faced with the problems of sparse reader behavior data and cold start of new books.therefore,this paper designs a similarity recommendation algorithm based on convolution neural network and a cold start recommendation algorithm based on long-term and short-term memory network.Similarity recommendation algorithm based on convolution neural network.In order to solve the problem that the accuracy of the existing personalized recommendation technology is reduced by sparse matrix,this paper focuses on how to use the relevant principles and techniques of deep learning to reduce the negative impact of data sparsity in the recommendation system.A similarity recommendation algorithm based on convolution neural network is designed and constructed,which iteratively adjusts the initial reader-book score matrix through the adjustment layer to locally characterize the user’s interest preference,and then uses the convolution neural network to predict the missing score,so as to implement personalized recommendation.The experimental results show that the average relative error of the model is lower than that of the existing recommended methods under different data sparsity,which verifies the effectiveness of the algorithm.Cold start recommendation algorithm based on long-term and short-term memory network.Aiming at the cold start problem of new book recommendation,irstly,the feature information extraction method of reader-book keyword is designed,and the improved Text Rank is used to obtain the characteristic keyword of new book content information and the comment keyword of read book,and the feature of new book keyword is obtained by weighted comprehensive evaluation.Then,the long-term and short-term memory network is used to model to obtain the knowledge sequence that reflects the user’s preference experience,based on which the reader’s preference degree to the new book is calculated.Finally,the reader preference degree is sorted to recommend the new book for the reader.The experimental results show that this algorithm is better than the existing algorithms in terms of recall and accuracy,and alleviates the negative impact of the cold start problem of new book recommendation to a certain extent.