Placing Music Artists and Songs in Time Using Editorial Metadata and Web Mining Techniques

Abstract

This paper investigates the novel task of situating music artists and songs in time, thereby adding contextual information that typically correlates with an artist’s similarities, collaborations and influences. The proposed method makes use of editorial metadata in conjunction with web mining techniques, aiming to infer an artist’s productivity over time and estimate the original year of release of a song. Experimental evaluation over a set of Dutch and American music confirms the practicality and reliability of the proposed methods. As a consequence, large-scale correlational analyses between artist productivity and other musical characteristics (e.g. versatility, eminence) become possible.