Research on music information retrieval algorithm based on deep learning

Abstract

With the exponential growth of various network resources, the use of search engine has become one of the most basic skills of everyone in today's society, and an efficient information retrieval model is also of more significance. The traditional text-based music information retrieval method can retrieve music data by inputting text information such as song name, composer, singer and album name. The content-based music information retrieval queries the target music through the input music melody information. In the actual music information retrieval scene, there is interaction between the user and the retrieval model. The user gives feedback on the retrieval results, and the retrieval model returns a new page of document according to this feedback. The existing ranking learning model regards ranking as a one-time process, ignores user feedback, and the ranking effect needs to be improved. With the increasing demand for digital music information and the continuous expansion of application fields based on massive music data sets, content-based music information retrieval method is attracting more and more researchers' attention.