Factual and cultural metadata for music service
In addition to factual metadata, subjective, culturally determined information at the level of the whole track is often used to retrieve tracks. Common classes of such metadata are mood, emotion, genre, style, and so forth. Most current music services use a combination of factual and cultural metadata. There has also been much interest in automatic methods for assigning cultural, and factual, metadata to music. Some services collect user preference
data, such as the number of times particular tracks have been played, and use the information to make new music recommendations to users based on the user community.
For example, Whitman and Rifkin used music descriptions generated from community metadata; they achieved Internet-wide description by using data mining and information retrieval techniques. Their extracted data was time aware, reflecting changes both in the artists’
style and in the public’s perception of the artists. The data was collected weekly, and language analysis was performed to associate noun and verb phrases with musical features
extracted from audio of each artist.
The textual approach to MIR is a very promising new direction in the field: a comprehensive review of the methods and results of such research is beyond the scope
of the current paper. For all its utility, metadata cannot solve the entirety of
MIR due to the complexities outlined above. Commercial systems currently rely heavily on metadata but are not able to easily provide their users with search capabilities for
finding music they do not already know about, or do not know how to search for. This gap is one of the opportunities for content-based methods, which hold the promise of being able to complement metadata-based methods and give users access to new music via processes of self-
directed discovery and musical search that scales to the totality of available music tracks. For the remainder of this paper, we focus primarily on content-based music
description rather than factual or culturally determined parameters.
However, content-based methods are considered not replacements but enhancements for metadata based methods.
