Contact: Goals of MEI
Notated music forms the basic source material for musicology. It is the basis of most musical performance, especially in classical Western art music. It is also often used to record and describe performed music not based on a notated text. Notated music has been produced in the West for more than one thousand years. A vast quantity of notated scores, in both printed and manuscript forms, is stored in the world's libraries. However, only a small portion of this music is available in digital form, and usually as image files. An even smaller portion is available digitally in a machine-readable form that represents the structural and semantic information contained in written or printed scores and that would facilitate computer-assisted research.
A commonly accepted digital, symbolic representation of music is necessary in order to move musicology into the modern era; that is, to make it possible to carry out the same kinds of operations that are commonly performed on electronic textual sources, for example, compiling musical corpora, data interchange, and comparative analysis. Also, because music research routinely combines the study of manuscript sources, printed music editions and time-based media, an encoding mechanism that facilitates the creation and management of relationships between components of digital versions of these materials will lead to improvements in research efficiency.
Several scholars have defined requirements of a suitable representation. Byrd and Isaacson [local copy] have enumerated many of the requirements for CMN, while Loos et al. [unpublished typescript] and the Music Encoding Study Group [local copy]have approached the problem more generally. Also, conferences such as the annual International Symposium on Music Information Retrieval (ISMIR) conferences and the conference on music editing held in Paderborn in 2007, indicate that interest in finding a solution to the challenges of representing music to support research is increasing. Despite these efforts, there are still many unresolved questions.
Assuming the need for encoded musical data, the current problem is not that there is no method for encoding notated music's structural and semantic information, but rather that there are too many ways, at least for the common cases. There is no single, accepted, non-proprietary standard for electronically representing and sharing this important cultural data. Over the last 40 years, dozens of music encoding formats have been created. However, much of the work on other digital representations of symbolic music up to this point is unsuitable for the scholarly editing and analysis of music.
Almost all existing music representations narrowly focus on supporting one or two functional objectives; that is, printing or automated performance. Few encodings or representations exist that facilitate analysis. Most codes are limited in scope to Common Music Notation (CMN), the most commonly recognized form of music notation, in use principally between 1700 and 1935, or to a single repertoire or notational style. Many representations are useful as input codes, but they have limited use as complete representations. Even the ubiquitous MIDI (Musical Instrument Digital Interface) files found in great quantities on the World Wide Web fit this paradigm. MIDI narrowly focuses on the storage and transmission of performed music. Therefore, it is completely unconcerned with score notation. MIDI does not allow the capture of the structural and semantic information necessary for applications in musicology, for example, for harmonic or structural analysis.
On the other hand, rather than focusing too narrowly, other representations, such as the Standardized Music Description Language (SMDL), have attempted to represent music too broadly. SMDL was unable to attract a large user group in part because it was difficult for potential users and tool developers to see how SMDL might apply to their particular situation.
Many existing codes are also proprietary. After expending a great deal of effort to create them, their owners are reluctant to divulge their inner workings. Therefore, their use for information exchange is severely limited. In addition, commercial forces often function as stumbling blocks in the creation of new standards. For example, work on NIFF (Notation Interchange File Format) was abruptly halted when the commercial parties involved in its development lost interest in it. Even the de facto MIDI standard is produced and controlled by the MIDI Manufacturers Association.
Furthermore, formats for music analysis generally do not deal with the graphical aspects of notation while those that work well for graphical applications do not have the semantic content necessary for other uses. Finally, few existing representations have even minimal support for the ambiguities and uncertainties encountered in manuscript sources.
The Music Encoding Initiative (MEI) data model attempts to address many of these issues and has been successful up to this point. Michael Kay, internationally-known for his work on Extensible Stylesheet Language Transformations (XSLT), an XML-based language used for the transformation of XML documents into other XML or "human-readable" documents, calls MEI a "serious contender." MEI has also been covered by both the popular (Stewart in Electronic Musician) and academic press (Williams and Webster).
The development of a standard digital representation of music by the scholarly community has the potential to transform musicology. Anyone who creates musical editions for publication or study, even ephemeral ones such as those for classroom use, will likely find the standard and tools based on it useful. Faculty teaching classes in the theory and history of music, students, and researchers conducting contentbased analysis or historical research would all benefit enormously. Music publishers, universities and research institutes will be able to utilize the data model and associated tools to create and distribute music in print or on-line. Finally, since the schema developed in this project will be freely available, open, and extensible, other software developers can build open-source tools upon it. For example, applications may be developed for navigation within a score, for synchronized display of the score and a machinegenerated or pre-recorded performance, for navigation among textual variants within the score, etc.