A Data Model for Digital Musicology and its Current State - The Music Encoding Initiative

During the last 10 years, XML has gained
general acceptance as a data model in the
Digital Humanities. Actually, it even leveraged
the success of digital projects in the humanities.
Meanwhile TEI is the unchallenged standard for
all projects in the fields of literature studies,
epigraphy, linguistics, history sciences and so
on. Many thoughts were invested to bring TEI
and other related formats like EpiDoc to a level
that suffices general scholarly needs.
At first sight, things went differently in the
field of music encoding. Around the year 2000
a couple of XML-based encoding schemes for
music notation emerged, and within just a
few years MusicXML became the best-known
and most widespread music encoding format.
It was intended to serve as an interchange
format between different music applications,
and even today it is virtually indispensable for
this very important task. At the same time,
this orientation of MusicXML requires a certain
"simplicity" that facilitates implementations in
various applications.
The Music Encoding Initiative (MEI) went a
different way. Not aiming at application support
in the first place, an encoding model for
scholarly
purposes was developed over years.
Strongly influenced by the concepts of the
TEI, Perry Roland as initiator of the format
tried to transfer these concepts to the field
of music encoding. For large parts, this is
quite easy: music notation is a kind of text,
and many unspecific modules of TEI can be
reused for music encoding with only small
changes. But then again, music notation itself
offers a much higher complexity than other
texts. It is multi-dimensional not only because
of its layout of multiple vertically aligned
staves, but also because of its simultaneity of
harmonic and melodic progression. In music
notation, the text itself consists of overlapping
hierarchies and therefore demands a quite
sophisticated data model. Most often, it is
virtually impossible to preserve all possible
meanings (or better: interpretations) of a
musical text with reasonable effort. The reason
is that the written text is only a part of the
complete information. Every notation serves a
certain purpose, and each composer or copyist
uses only as many symbols as he needs to be
explicit to his contemporaries. Besides this, the
"rules" of music notation changed significantly
over time, even though these developments
often seem to be very subtle.
All this leads to the problem that there is no
absolutely fixed terminology in music notation.
Some phenomena are still not completely
understood or even defined, such as the problem
of dots, strokes and hooks in scores from the
classical period. The lack of a complete and
well-defined terminology even for restricted
repertoires makes the encoding of music
notation on a scholarly level highly demanding,
and, at the same time, the implementation and
usage of such an encoding scheme is anything
but trivial.
The Music Encoding Initiative has chosen this
way, and currently it stands on an important
turning point: In a one-year project funded
by the NEH and DFG the original model was
revised and has proven to meet all essential
scholarly requirements for such a format. In
the next years, it needs to be disseminated
in the fields of musicology, music information
retrieval, music philology and digital humanities
in general. A first step in this direction is the
TEI's Special Interest Group on music encoding,
whose members were actively involved in the
recent developments on MEI, and who seek to
find ways to bring MEI and TEI closer together.
Due to the complexity of music notation
– and thus music notation encoding too –
application support for MEI is crucial to
ensure its dissemination: Almost no traditional
musicologist would be willing to work with
a XML-editor like Oxygen. There are several
projects currently working on such applications
for MEI: The DiMusEd-Project, situated in
Tübingen (Germany), uses SVG to render
encodings of multiple sources of music notated
with medieval neumes. Although this repertoire

2
uses a limited set of symbols, this project already
shows the benefits of a dynamic rendering from
an encoding instead of engraved scores. The
Edirom project, (Detmold, Germany) aims to
establish workflows for digital scholarly editions
of music. In the application for preparing
such editions it is already using MEI to store
all structural information about the musical
text as well as the containing documents.
For moving from basically facsimile-based
editions to completely digital editions it is
planned to offer complete encodings of all
relevant sources including the rendering-
facilities already demonstrated by the DiMusEd-
project. In order to achieve this goal Edirom
closely collaborates with the most ambitious
of all ongoing MEI-related projects: TextGrid.
A sub-project of this major German initiative,
which is also located in Detmold, seeks to
develop a limited scorewriter for MEI offering
a graphical user interface for musicologists. In
this case „limited“ means that the project neither
intends to support MEI completely nor tries to
keep up with the engraving quality of already
existing scorewriters: the unambiguity of the
output is more important than its beauty.
All these German projects collaborate closely
with the ongoing efforts in the US to further
improve the format itself and to provide
interchange to other relevant formats such
as Humdrum and MusicXML. Depending on
further funding by NEH and DFG respectively
it is intended to provide reasonable collections
of MEI encodings to facilitate further usage of
the format. Although MEI will not find the wide
acceptance MusicXML already has, all these
components will help to disseminate MEI in
the academic world, to promote interchange of
high-quality data and to explore new methods
for digital representations of written music.
The talk will provide a short introduction to the
current state of MEI – both the format itself and
the projects and applications already working on
and with it.