Automatically detecting members and instrumentation of music bands via web content mining

Abstract

In this paper, we present an approach to automatically detecting music band members and instrumentation using web content mining techniques. To this end, we combine a named entity detection method with a rule-based linguistic text analysis approach extended by a rule filtering step. We report on the results of different evaluation experiments carried out on two test collections of bands covering a wide range of popularities. The performance of the proposed approach is evaluated using precision and recall measures. We further investigate the influence of different query schemes for the web page retrieval, of a critical parameter used in the rule filtering step, and of different string matching functions which are applied to deal with inconsistent spelling of band members. © 2008 Springer-Verlag Berlin Heidelberg.