The transcription system EXMARaLDA: an application of the annotation graph
formalism as the basis of a database of multilingual spoken discourse

Thomas Schmidt
University of Hamburg

The SFB "Mehrsprachigkeit" (Research Center on Multilingualism) at the
University of Hamburg brings together linguists doing research on
multilingualism from a variety of theoretical perspectives. The majority of
the projects work with spoken language data, and the formats, tools and
platforms these data come in are as diverse as the backgrounds of the
research groups  there are data in thirteen different languages (German,
English, French, Italian, Spanish, Portuguese, Basque, Luganda, Turkish,
Japanese, Norwegian, Swedish, Danish), created and processed on different
platforms (Windows, Macintosh, Linux) and with different transcription
conventions (Verbmobil, HIAT and others) and transcription tools
(syncWriter, HIAT-DOS, LAPSUS (dBase), conventional text editors etc.). The
systems in use are for the most part outdated and technically as well as
conceptionally incompatible with one another. The urgent need for an
exchange format between them is therefore obvious.

The project "Database Multilingualism" not only aims at providing such an
exchange format. It also takes care of the implementation of appropriate
tools for input and output of the data for different purposes and on
different platforms.

Its main activity until now has been the development of the discourse
transcription system EXMARaLDA (EXtensible MARkup Language for Discourse
Annotation). Based on the annotation graph formalism, EXMARaLDA defines a
way of coding discourse transcriptions in XML in a way that is independent
of both linguistic theory and concrete graphical representation and that is
able to comprise, and mediate between, the various formats in use at the
SFB. At the same time, EXMARaLDA provides mechanisms and tools for
user-friendly input and output of the data that reflect the needs of the
individual projects. In particular, tools for input and output of discourse
transcriptions in musical score notation (German: Partiturformat) have been
designed and implemented.

As a next step, the existing data at the SFB will all be converted into the
EXMARaLDA format. The result of this conversion process will be a number of
XML-coded corpora that can be easily exchanged between researchers and
adapted to their respective needs. The ultimate goal of the project
"Database Multilingualism" is the implementation of a database that brings
together all the multilingual data at the SFB and make them available for
an integrated search and analysis.