Finding the hidden gems: recommending untagged music

摘要: We have developed a novel hybrid representation for Music Information Retrieval. Our is built by incorporating audio content into the tag space in tag-track matrix, and then learning concepts using latent semantic analysis. apply this to task of music recommendation, similarity-based retrieval from query track. also develop new approach evaluating recommender systems, which based upon relationship users liking tracks. are interested measuring recommendation quality, rate at cold-start tracks recommended. able outperform tag-only representation, terms both quality that included as recommendations.