Research On Song Recommendation Method Based On Lyrics And Comments

The Internet has developed rapidly in recent years and has been integrated into all walks of life and all aspects of our work and life.A large number of various forms of data have also been generated on the Internet.Selecting the information that everyone is interested in among the many information has become a hot topic in our research today.From this,the recommendation system was born,and with the development of Internet technology,especially big data and artificial intelligence technology in recent years,the recommendation system has also been applied to all walks of life.Recommender systems are also widely used in the field of music.Music is an indispensable part of modern people’s life.With the sharp increase in the number of various styles of music on the Internet,a powerful recommendation system is needed for personalized recommendation.Some recommended methods appeared early on.These traditional methods have indeed achieved good results,but with the increase of user information consumption,music comment information should be paid enough attention,and many wonderful comments can also reflect the information of the song itself;and in the field of natural language processing,more recent Many powerful text matching models based on deep learning have been developed in 2010.If such technologies are applied to the field of music recommendation,they will have better results than traditional recommendation methods.Based on the study of a large number of previous studies on music recommendation systems,this paper summarizes the commonly used music recommendation methods,and designs a method for music recommendation based on lyrics and comment texts combined with a deep text matching model.In this paper,data acquisition is performed first,that is,the song information is crawled,including the lyrics and comment text information of each song,as well as the similar song information marked by the music platform;then different models are used to match the song text.In terms of model selection,This paper selects a traditional text matching model and two deep text matching models,compares the deep text matching model with the traditional text matching model,and integrates the song comment text on the basis of the lyrics text to compare the changes in the recommendation effect before and after.The results show that the music recommendation method based on the lyrics and comment text using the deep text matching model has high recommendation accuracy.In addition,in order to explore the influence of different weights of review texts on the recommendation results,this paper improves the recommendation effect by determining the best review weights.