LRE 62-080:
The COMPASS Project
COMPASS
Adapting Bilingual Dictionaries for On-Line
Comp
rehension
Ass
istance
The COMPASS crew at the Final
Review Meeting at RXRC, Grenoble, in April 1996.
From left
to right:
Elisabeth Breidt, Helmut Feldweg, Sue Atkins,
Lauri Karttunen Annie Zaenen, Nicholas Ostler, Irene Maxwell,
Hélène Lewis, Nadine Zaouche, Giuseppe Valetto, Romeo Peter
Wetzel
The COMPASS project aims at the adaption and integration of bilingual
print dictionaries into an intelligent, context-sensitive dictionary
look-up system. The project is funded by the Commission of the
European Union under the Linguistic Research and Engineering
Programme. It runs from April 1994 until March 1996.
The following documentation on the project is currently available online:
-
A short description
of the project,
including its objectives, approach, and methodology.
-
A more up-to-date description of theproject
(also available in
German
). Including a description of the
components of the COMPASS system and a screen dump from the
prototype of the COMPASS system.
-
A more comprehensive description of the COMPASS system, its
various modules and the chosen approach to dictionary conversion (
English
or German
-
A description of the implementation ofa Hidden Markov Model for German
that serves as a part-of-speech
disambiguator in the COMPASS system.
-
A description of the finite-stateapproach
used in the project to recognize if the word to be looked-up
by the COMPASS system is part of a multi-word lexeme.
The main focus of the Tübingen group of the project is the analysis,
preparation, and enrichment of the Collins German-English dictionary.
We have developed a dictionary lookup tool
for
ready access to the tagged version of this dictionary. Note that the
access to the tool is restricted to in-house use only due to license
restrictions.
The Tübingen group is also involved in the development of a German
language model for Xerox' LocoLex system which serves as the lookup
engine of COMPASS. One module for this German model is a
part-of-speech
disambiguator
based on a Hidden Markov Model for German.
For the language pair German-English, the COMPASS system has been
tested with a number of selected articles from German
newspapers.
The following people were involved in the Tübingen team of the COMPASS
project:
-
Elisabeth Breidt(breidt@sfs.nphil.uni-tuebingen.de)
-
Sabine Engelke
-
Helmut Feldweg
-
Michael Hipp
-
Christine Thielen