We present a middleware system that facilitates Internet users? access to web-based digital music libraries and allows them to manipulate audio meta-information taking into consideration content and semantic information of music data. Useful relations in the data are automatically extracted through semantic networks (constructed and maintained in the library). Our system is complemented with a query-by-example retrieval subsystem, user relevance feedback facilities, and a new approach for musical genre classification based on the features extracted from signals that correspond to distinct musical instrument sources, as these sources have been identified by a source separation process. The system operation is illustrated in detail.