The aim of the proposed project is to work towards the automatic semantic description of digital music. The project intends to contribute to bridging the current semantic gap in music information and apply the results to content-based music processing. Specifically, in this project we will work on (1) the analysis and manipulation of low-level audio descriptors by spectral modelling analysis and synthesis techniques, (2) the extraction of high-level musical attributes from these low-level audio descriptors, (3) the study and development of pattern recognition and machine learning techniques applied to sequential data for building semantic models of different aspects of music and, based on these semantic models, (4) the development of prototypes of music mining, personalization, and postproduction next generation systems. As a tool for carrying out (1) and (2) we will implement a distributed real-time concurrent programming language. This project is a natural continuation of the UPF-coordinated projects TABASCO (TIC2000-1094-C02-01/02) and PROMUSIC (TIC 2003-07776-C02-01).