Research On Music Recommendation System Based On Users' Reviews

With the rapid development of the Internet,there are more and more music products on all kinds of music websites.Facing massive music products,people often find it difficult to find music that they are interested in.Recommending systems can help users to make a choice.Therefore,how to make a more accurate and personalized recommendation based on the personal preferences of the user has become an important research content.At present,all kinds of music website platform contains a lot of information on user comments,however,most of the music product recommendations are based on the traditional collaborative filtering algorithm or recommendation algorithm based on contents,ignoring the importance of information user comments,and the two algorithms have been faced with problems such as data sparse,recommend the simplification,leading to poor performance of personalized recommendation.In view of the above problems,firstly,this article designed a music recommendation system framework based on user reviews by embarking from the users' comments information mining.Secondly,this article introduced in detail the use of user comment text to improve the process of the original recommendation algorithm,and respectively put forward the integration of user reviews matrix decomposition song recommendation algorithm and the clustering song recommendation algorithm based on user reviews.The matrix decomposition song recommendation algorithm,which combines user reviews,extracts user interest and song features from the user's comment text,and integrates the results of their nearest neighbors into the matrix decomposition model as a regularization item,thus improving the traditional matrix decomposition algorithm and improving the accuracy of the song recommendation.The clustering song recommendation algorithm based on user reviews calculats the text similarity between the text of user comments and the text of the song to fill the original song score matrix.On this basis,the clustering of the users with similar preference solves the problem of data sparsity further,improves the traditional collaborative filtering algorithm,and provides personalized song list recommendation for the target users.Finally,the improved song and the song list recommendation algorithm were analyzed respectively in this paper.The experimental results show that the improved algorithm significantly improves the recommendation effect and improves the recommendation quality of the music recommendation system.