Knowledge Acquisition from Music Digital Libraries

TitleKnowledge Acquisition from Music Digital Libraries
Publication TypeConference Paper
Year of Publication2015
Conference NameInternational Association of Music Libraries and International Musicological Society IAML/IMS Conference
AuthorsOramas, S., & Sordo M.
Conference Start Date21/06/2015
Conference LocationNew York
KeywordsDigital Libraries, information extraction, Knowledge Acquisition, Music, semantic web
AbstractA vast amount of musical knowledge has been gathered for centuries by musicologists and music enthusiasts. Most of this knowledge is implicitly expressed in artist biographies, reviews, facsimile editions, etc. Music Digital Libraries make this information available and searchable. Documents are indexed and keyword-based search is generally provided. However, implicit knowledge present in text is not understood by machines, so complex queries cannot be answered.

As a first step towards the approximation of machine understanding to the accumulated musicological knowledge, documents stored in Digital Libraries must be semantically annotated. Current descriptive metadata and markup annotations provide some structured information. Nevertheless, it is insignificant compared with the epistemic potential of the source text. Once documents are properly annotated, complex structures and meaningful relations between pieces of information may emerge. This supports a paradigm shift, from keyword-based systems to knowledge-based systems, hence enabling musicologists to formulate more complex queries. As a consequence of that, Digital Libraries will turn into real knowledge environments, instead of mere searchable repositories.

Manual annotation of documents is very expensive, and sometimes unwieldy. Thus, the use of reliable, automatic processes is crucial to build knowledge environments. In the last few years several studies and approaches on the use of Semantic Web technologies in Digital Libraries have been proposed. The result of these intersection has been coined as Semantic Digital Libraries. Most of the related work is focused on the acquisition of Semantic Web methodologies for knowledge representation, which, among other advantages, facilitates information exchange between multiple knowledge bases. In the case of Music Digital Libraries though, tools and methodologies developed around the Semantic Web for automatic knowledge acquisition have received less attention.

In our work, an extensive survey is provided about the applicability and performance of state-of-the-art semantic technologies for knowledge acquisition from Music Digital Libraries. These technologies are adapted to fulfill the requirements and specificity of the music domain. An evaluation of analyzed tools is performed over artist biographies and Flamenco Music articles gathered from different sources (e.g. The Grove Dictionary, Wikipedia). An exhaustive overview of the possibilities that a knowledge layer may offer to Music Digital Libraries is exposed. Finally, some guidelines for future work in this research direction are also provided.