This report considers three different types of metadata schemes. The first type are schemes which describe the musical score is a markup languages. The schemes considered here are Music XML and SMDL. Other similar schemes include kern from the Humdrum toolkit and various proprietary formats. The second type are metadata schemes which markup the musical text adding additional information to the text such as interpretative information, in many ways being similar to TEI but for musical texts. The schemes considered here are MML. MPEG7 adopts a similar module allowing the tagging of audio an video extracts within the file with human and computer readable tags. The final type are for describing a catalogue of music, here MusicCat is considered. There is of course overlap between these schemes. SMDL is equally applicable as one of the second type as it is as one of the first type. MML, whilst designed to describe a single text, contains information pertinent to a music catalogue as described by Music Cat.
A format not covered below but relevant is the Music Exchange Format (MEF) employed in the OMRAS project (http://www.omras.org). One of this projects aims is to develop algorithms for searching music by content using schemes of the first type mentioned above, i.e. those which describe the music score. MEF is an internal format consisting of just those elements of such schemes being used by the algorithms in development so that work is not distracted by additional complexities. It is not intended as a new format, and the final algorithms will work on other formats such as those listed here.
NIFF was established as standards to meet the need for a vendor independent format for transferring data between different score notation writing packages. SMDL was developed is an ISO standard to allow the linking between four different identified representational domains for musical scores – namely:
SMDL uses SGML (the father of XML) for the description within the logical domain and HyTime (the father of linking mechanisms within XML such a XPath) to indicate links between the four domains. The EU Cantate project began with the intention of using SMDL as a basis for developing online delivery of scores with value added features such as interpretations, linking between audio and online score etc. but find itself spending most of its effort defining the standard. It adopted NIFF as the preferred (but not mandatory) format for the visual domain. The gesteral domain was defined to be audio formats or MIDI. The Analytical domain was left undefined. An output from Cantate was a NIFF toolkit. The project MuTaTed picked up the work of Cantate and developed a compiler for translating SMDL to NIFF and NIFF to SMDL. MuTaTedII, which began in November 1999 aims to produce a music information retrieval system for encoded music based on these standards.
There is no platform independent, generic and established standard that can be used to describe, represent, and retrieve time-based data such as music. An application such as this, with the combined implementation of two standards, supports the representation of music in the SMDL format, and provides a standard for high-quality display using the NIFF format. There has been an expressed need for features such as moving markers in displayed music notation, automatic page-turning, and searching and query of musical content across music databases, but no one had developed products to fulfill these needs.
SMDL has been called the XML of music, and is based on HyTime, which is the multimedia extension of SGML. SMDL is described as an architecture for the representation of music information, either alone, or in conjunction with text, graphics, or other information needed for publishing or business purposes. (ISO/IEC Draft International Standard 10743).
NIFF is a binary music representation format designed to
allow the interchange of music notation data between and among music notation
editing and publishing programs and music scanning programs. It is intended to
preserve the significant amount of detail that is normally lost when
translating through MIDI files. Music notation software applications had to
each create their own music data structures to contain all of the information
needed to print music, and NIFF is a dedicated notation score representation
which is also an interchange file format.
One advantage is that the integration of the two existing standards would establish a standard for music metadata that could be used by the entire music community to store and display music notation. Links can be made from audio files or image files to the music score. Performers can examine the decisions (eg. bowings, fingerings, tempo changes) made by other performers, a very useful facility given the difficulties of gaining access to these decisions through the traditional media. Students can follow the work of masters as the work develops, rather than seeing it as a whole once it is finished. SDML has been an academic ideal which currently has not had much appeal to industry. A major disadvantage is the complexity of not only SMDL but also HyTime upon which it is based. There has only ever been one attempted implementation of a HyTime engine, and that is still regarded as in beta with only a small subset of the HyTime standard implemented. SGML/HyTime is likely to be overtaken by XML/XPath/Xquery and SMDL would have to follow suite.
The music industry has had its own problems with music representation languages. MIDI, while very successful in its way, is not expressive enough for modern requirements. Unfortunately, industry leaders have had a tendency to use proprietary formats and attempt to gain market share, rather than agree upon on a format convenient for all. Several companies that develop music software recognised the futility of this approach and banded together to develop the NIFF music representation language, and there are now a couple of editors (Lime and in the future, Encore) that import and export NIFF as well as MIDI. However, despite work from Cantate on NIFF format toolkits it is still not taken up widely from my music software developers and some key players have withdrawn from the NIFF standard.
Several applications might be founded on the concept of a combination of these two standards such as platform independent file interchange and dynamic access to music. SMDL in particular lends itself to adding value to electronic delivery of music, e.g, in teaching and learning and analysis applications.
The primary use of this metadata scheme is to allow the representation of notation music scores adopting the current symbolic representation in western music notation since the 17th century. As such it could be a interchange format between notational score packages.
As the moment there are few vendor independent schemes for representing notation and scores. Some vendors of score-writing packages have disclosed the formats used by their packages, but these tend to be proprietary and under the control of that vendor hence liable to change without warning. Whilst there are existing open standards for this application such as NIFF and SMDL, these are not widely adopted. In some ways it is similar to MIDI but more addressed to the presentation of a score than its performance.
The standard is based on XML and is making use of XML Schema as well as the more traditional DTD’s to provide greater control of the underlying structure. One problem with using XML in this context is that music is an inherently two dimension structure. XML however is a tree structure and thus not suited to this purpose. Most music notational schemes either separate into parts before separating into measures or vice versa. Music XML specifies DTD for both schemes plus an XSL transform to convert a files from one to the other.
The advantages are that it uses XML (as opposed to RIFF in the case of NIFF and SGML/HyTime in the case of SMDL) and it has an interesting solution to the problem of whether to subdivide into parts or measures first. However, it is not yet a widely adopted standard, although the author is addressing this by working on a number of convertors to and from various proprietary formats.
There are a number of uses – the most obvious being in any circumstance where having an electronic form fo the score would be useful, e.g.
Music Markup Language (MML): (http://www.mmlxml.org/)
Music Markup Language (MML) was developed to act as an intervening interface between machine-level code and the user interface. The primary concern was the complexity of MIDI, which is a complex code not easily usable to the average person. MML was not intended to replace MIDI but to complement it by adding “human language elements” that will make MIDI easier to understand. The creators also foresee a mapping capability by which MIDI functions can be translated into MML and vice versa.
Standard metadata languages were not designed with the specific qualities of music in mind, unlike MML, which concerns itself only with the coding of music for disbursement. MML describes the structure of various aspects of music, including music performance, music representation (notation systems), music instruments, music manipulation (effect units), and MIDI. MML marks the structural objects or elements of a piece of music.
MML proposes to mark up music in much the same way as HTML marks up a web document for delivery to various devices. MML will deliver music in the following formats: a. Text, presenting notes as, for example, 1C:4 (quarter note C in the first ocatave); b. Block bars or piano roll (such as found in computer sequencer software); c. Western music notation (either on screen or on paper); d. MIDI device, played on a synthesizer, or controlling music processors. MML is an SGML (ISO 8879) subset, following the approach to SGML by XML. The method used for creating a markup language for MML is a “bottom-up” approach, and its implication is as follows:
Other metadata schemes: NIFF focuses on printouts. SDML focuses on music document archival storage of music. It is very complex and the element names are not human friendly. It is for music theorists, whereas MML is intended for the popular markup of music. SDML distinguishes between four domains:
MML adds another distinction to this, on the meta-level and makes an artificial distinction between a note and its manipulation. This is a very important distinction from the point of view of music description.
The major advantage of MML is its human-readable format, making it much easier to use than MIDI and thus creating a platform for the user to translate the MIDI interface. A possible disadvantage is that MML is currently designed to handle only Western music notation. MML cannot handle nonwestern scales and tunings, or styles where small differences in pitch and volume are regarded as important. MML recognizes this disadvantage and ignores it for the time being until the markup language for Western notation can make progress since it is the most dominant form of music in the world today.
Some of the foreseeable applications of MML are:
MML should also be applicable to many different music-related environments:
MusiCat is a Document Type Definition that was originally developed/created for use with Thematic Catalogs.
The author states, “The DTDs that exhibit potential for markup of thematic catalogs (TEI and MARC) are unsatisfactory. TEI is designed primarily for representing texts, not data, while MARC uses a flat database approach that doesn’t accommodate markup of existing catalogs. In addition, neither DTD allows markup of musical data using recognized musical terminology.” (Roland, Perry. “The ‘Preposterous Muddle’ Revisited: an XML Thematic Catalog DTD,” paper presented at ACH-ALLC 1999.)
The metadata scheme proposed in MusiCat is based on XML.
Advantages include: a. Syntax and semantics are treated separately; b. Data is infinitely malleable and can be distributed in both print and electronic forms; c. meets the specialized needs of thematic catalog creators as well as users; d. it provides traditional access points (title, creator, imprint) as well as music-specific access points (tempo, meter, key; e. it is structured enough to be useable for database-oriented applications, and loose enough for presentation-oriented existing catalos; and f. header information at the descriptive level allows for inclusion of management data that can be applied to collaborative projects and data access restriction. Disadvantages include: as yet this is not in wide use as an accepted standard
The author lists four possible applications: