musicSpace : Improving Access to Musicological Data

1 musicSpace: Improving Access to Musicological Data mc schraefel, David Bretherton, Daniel Alexander Smith, and Joe Lambert Problem Not unlike the sciences, musicological data is widely distributed and exists in numerous formats and in many databases. Efforts over the past decade to digitize bibliographies, artists’ works lists, recordings, program guides and related ephemera should mean that data that was once physically distributed (requiring a researcher to visit various sites to consult items) is all readily accessible from the comfort of one’s desktop. But this has not entirely been the case: the geographical dispersal of material in the physical world has been replaced by the dispersal of data into a plethora of discrete and disparate online databases (“data silos”) in the virtual world, according to, for example, media type (text, image, audio, video), date of publication, subject, language, and/or copyright holder. Because musicological research typically cuts across such artificial divisions, musicologists thus routinely expend valuable research time consulting a bewildering number of heterogeneous data repositories. A related problem is that the quality of the metadata often of course determines the quality of the search. In this regard, note that even some of musicology’s leading data providers use legacy or ad hoc metadata specifications that are unsuited to modern demands. For instance, a source that provides lists of the works that a composer has composed has no consistent format for the production of these lists, while another is inconsistent with terminology – a composer in one record is an author in another. So, while the data has been digitized, there is no easy way to ask questions like “which scribes have created manuscripts of a composer’s works, and which other composers’ works have they inscribed?”, and “which poets have had their poems set as songs by Schubert, and which other song composers have also set them?” The combined effect of these shortcomings is that such real-world musicological research questions are effectively intractable, not because the data needed to answer them is not in a data source somewhere, but because there is insufficient metadata or metadata granularity, and a lack of data source integration (meaning that metadata from one source cannot be used as the basis for a query of another source).