- 7h -
INTERNATIONAL SOCIETY FOR PHOTOGRAMMETRY AND REMOTE SENSING
Commission VI
Symposium held in Mainz, FR Germany, 22 - 25 September 1982
SUSSY-SO'FTWARE FOR INFORMATIVE AUTOMATIC
TRANSLATION OF ABSTRACTS
Heinz-Dirk Luckhardt
SFB 100, University of Saarbrücken, FR Germany
ABSTRACT
The "Saarbrücker Übersetzungssystem" SUSY has been developed by the "Sonderfor-
schungsbereich Elektronische Sprachforscnung" at the University of the Saar
in Saarbrücken and does not aim only at the automatic translation (or machine
translation - MT) of abstracts, but of all kinds of texts (except literary
texts): word lists, titles of articles, abstracts with complete sentences or
just noun phrases, and running text. In addition to this the language analysis
part of SUSY can be used for document analysis and information gathering which
has been proved by the project JUDO of Prof. Dr. Zimmermann. SUSY has reached
a stage, where it can be put to the test, i.e., it is not yet ready for pro-
duction. This paper deals with the question what the system can do and what
is left to be done, before it can enter production.
1. Dictionaries
The main problem for the more advanced MT systems is the treatment of "unknown
words" during the analysis of the scource language and the actual translation
phase. SUSY can handle a large amount of them in analysis by the module for
the segmention of compound or derivated words like Fassaden/pläne, Staellit*en/
bild/interpreation, Orthoprojection, EARTHNET/-/programm. A prerequisite for
this is the completeness of the dictionary for analysis, i.e., "ortho" and
"Projection" must be there. Most of the segmented words can be translated by
addition of the translations of the components: ortho'projection, facade wall/
plan, staellite / picture / interpreation. Again, this can only be done, if
the translations of the components are in the translation dictionary. All this
demonstrates that relatively complete dictionaries are indispensable for a
successful production system. The completion of the dictionaries surely cannot
be accomplished by a research project like the SFB 100, but remains the task
ofa potential user or a pilot project.
2. Program system
The development of SUSY started about 1972, and these ten years have produced
a relatively stable system ith varying degrees of linguistic description for
the different languages. As no literary translation is aimed at and a certain
percentage of errors is expected to remain in the system, it is a question
of "user-acceptance", whether SUSY can enter the production phase after a period
of testing and subsequently improving the program. "User-acceptance" means
that a user is prepared to regard the produced translation as "informative",
i.e., he accepts it, if it gives him the desired information, and ignores certain
stylistic deficiencies.
Bibliographic quotation :
Luckhardt, H.D. : SUSY - Software for informative automatic translation of abstracts.
In: Int. Archive of Photogrammetry, 24 - VI, pp 74 - 77, Mainz 1982