Inferring similarity between music objects with application to playlist generation

Abstract

The growing libraries of multimedia objects have increased the need for applications that facilitate search, browsing, discovery, recommendation and playlist construction. Many of these applications in turn require some notion of distance between, or similarity of, such objects. The lack of a reliable proxy for similarity of entities is a serious obstacle in many multimedia applications.In this paper we describe a simple way to automatically infer similarities between objects based on their occurrences in an authored stream. The method works both for audio and video. This allows us to generate playlists by emulating a particular stream or combination of streams, recommend objects that are similar to a chosen seed, and derive measures of similarity between associated entities, such as artists.