Now that MusicXML 0.5 can handle the basics of interchange between notation and performance applications, there are two main efforts planned to meet the goals of growing the downloadable sheet music market:
Music information retrieval is complex: the queries are often “fuzzy” (as in the ultimate goal of query by humming or singing), and the data relationships are complicated. We have made some initial attempts to use the June 2001 working draft of XQuery for extremely simple queries of musical melodies, but the results have been discouraging. XQuery’s current capabilities for handling queries based on complex, ordered relationships between XML document elements does not seem as strong as its capabilities for Structured Query Language (SQL)-like queries. SQL techniques unfortunately do not get us very far in music information retrieval.
We believe that MusicXML provides the structure that is needed for music information retrieval, but if XQuery will not handle this domain, specialized query tools may need to be developed. This would be unfortunate for music information retrieval. Music software is a small business compared to other software application areas. Much of XML’s attraction for music comes from its ability to leverage the investment in XML tools made by larger software markets. If standard XML query tools cannot meet even the most basic needs of music information retrieval, the music community is not likely to be able to take advantage of the optimizations and features provided by new generations of XML database tools.
Music information retrieval likely has years of research ahead on algorithm development. We believe that XML tools and music information retrieval tools can co-evolve together to meet user needs. Music can serve as a useful application area to broaden the scope of standardized XML query tools. Meanwhile, a standard XML format used in XML databases can let music information retrieval researchers focus on the difficult questions of useful algorithms. The focus needs to move to low-level database representation only when that directly affects these algorithms.
If we really can get query-by-humming to work well for average customers, this could have large commercial implications. But we do not expect to need these breakthroughs to meet the goals of growing the downloadable sheet music market. For this, it should be sufficient to expand the reach of MusicXML to more music software available on people’s personal computers. If all you can do with downloadable music is play it and print it with one program, why would you buy it compared to paper? But if you can edit the music, use it as a smart accompaniment, look at the musical score together with the playback of a CD, move the music to an electronic music stand, and write new musical programs yourself, the value of the downloadable sheet music increases dramatically. Once the music you download can be used on most any music program on your PC, downloadable sheet music will start to have more value – or different but complementary value – than paper music. Pervasiveness was a major part of what made the MP3 audio format so popular. We need that type of pervasiveness to make any type of digital sheet music more popular.
As with MP3 and MIDI, the flip side of pervasiveness is security, or the lack of same. We are hopeful that XML digital signatures can contribute in this area. This is an area that will require careful development before many music publishers are likely to embrace MusicXML or similar technologies.
In the short term, our focus will be to extend MusicXML’s reach to many different music software applications, such as music education. Our early successes with MusicXML give us hope that we may at last have a standardized music interchange format, which in turn will enable the growth of the downloadable sheet music and music software markets.