Using fuzzy logic to handle the semantic descriptions of music in a content-based retrieval system

This paper explores the potential use of fuzzy logic for semantic music recommendation. We show that a set of affective/emotive, structural and kinaesthetic descriptors can be used to formulate a query which allows the retrieval of intended music. A semantic music recommendation system was built, based on an elaborate study of potential users and an analysis of the semantic descriptors that best characterize the user’s understanding of music. Significant relationships between expressive and structural semantic descriptions of music were found. Fuzzy logic was then applied to handle the quality ratings associated with the semantic descriptions. A working semantic music recommendation system was tested and evaluated. Real-world testing revealed high user satisfaction