Towards Semantic Music Information Extraction from the Web Using Rule Patterns and Supervised Learning

We present first steps towards automatic Music Information Extraction, i.e., methods to automatically extract semantic information and relations about musical entities from arbitrary textual sources. The corresponding approaches allow us to derive structured meta-data from unstructured or semi-structured sources and can be used to build advanced recommendation systems and browsing interfaces. In this paper, several approaches to identify and extract two specific semantic relations from related Web documents are presented and evaluated. The addressed relations are members of a music band (band−members) and artists’ discographies (artist − albums,EPs, singles). In addition, the proposed methods are shown to be useful to relate (Web-)documents to musical artists. For all purposes, supervised learning approaches and rule-based methods are systematically evaluated on two different sets of Web documents.