Intelligent Retrieval Method of Massive Music Information Resources Based on Deep Learning

With the increase of data storage capacity and the development of transmission technology, the number of music shows an unprecedented growth. However, this explosive growth makes it more and more difficult to find interesting music clips in such a huge library of music information resources, so it is of great significance to study the intelligent retrieval methods of massive music information resources. Use the word segmentation tool to segment and label the music information resources. According to the word segmentation and part of speech tagging, build a feature extractor based on deep learning, obtain the features of massive music information resources and label the resources using approximate matching method. Combined with the result of resource feature annotation, the size of resource buffer is determined to build an intelligent hierarchical index. For different hierarchical indexes, sparse reconstruction technology is used to preprocess information resources, and pseudo-correlation feedback is used to expand the cache query. Based on the cache query expansion, match the query content entered by the user with the extension words with the similar meaning of the keyword to obtain the matching cost. According to this cost, the result of matching cost less than the set threshold is fed back to the user through feedforward neural network to obtain the retrieval results of massive music information resources. The experimental results show that the maximum expansion coefficient of this method can reach 0.99, the retrieval speed is fast, and the maximum error between the retrieval recall and the actual data is 0.30 kB.