Abstract

In 2006, the rockanango system was developed for music annotation by music experts. The system allows these experts to create new musical parameters within a flat data structure [1]. Rockanango is deployed in a commercial environment of hotels, restaurants and cafés. One of the main concerns is the time it takes to manually annotate the music and to introduce new parameters. In this paper, we investigate the possibilities to assist the experts by means of automatic metadata generation. Two case studies are described. One focuses on the use of association rules, in combination with lower level metadata like mode and key. The other case study concerns the generation of a topic or subject marker for songs through harvested lyrics and a

Citations