English Version

Workshop: Cost and Benefit of Machine Translation

A Workshop by the Special Interest Group on Machine Translation In conjunction with the GSCL-Conference in Fall 2009

Potsdam, September 30th - October 2nd, 2009

The workshop takes place at Wednesday, 30th of September. Scheduled time is from 10:45 AM to 1:15 PM.

Conference Venue: University of Potsdam (Informations)

Registration is only necesarry for the GSCL conference, but not for the workshop itself.

Current Research and Development of Machine Translation is often dedicated to quickly create cost-effective Translationsystems. This is done by including knowledge from bilingual Corpora. There are several questions:

Which methods are best suited to combine a high level of automation in the production with the highest possible quality?

Are there any differences in the methods of optimization when the quantity of the currently available corpora and / or linguistic characteristics of the considered languages are different?

What is the User's opinion? How much and which kind of quality is needed, for what purpose, in which language pairs, and how useful are the current systems?

What is the relationship between academic research and industrial research and development? Is there still a need for academic research if companies like Google are acting on the market?

Approaches to these questions are different depending on which perspective is taken.

The following lectures are scheduled:

1. Translation between 23 languages: Architectural considerations in Euro Matrix + (Andreas Eisele, DFKI)
2. Beyond phrase-based statistical machine translation (Alex Fraser, IMS Stuttgart, previously Language Weaver)
3. Translation quality from the user's perspective (Uta Seewald-Heeg, TH Koethen)
4. Academic Research in Machine Translation in times of Google and Microsoft (Jonas Kuhn, University of Potsdam)
5. Multiwordprocessing in the translation system translate: techniques, costs, achievable quality (Kurt Eberle, lingenio, Heidelberg)

The question of costs and benefits should not only base on general considerations of the state-of-the-art and evaluation studies on various methods. But it should also refer to a number of linguistic phenomena we are going to analyse. The context here is the specific question regarding the translation of functional verb phrases and multi-word units. In this respect the unique language of study is not seperable from the general problem of translation. Therefore, this part of the workshop will be hosted in cooperation with the Special Interest Group on Computational Lexicography. The last lecture by Kurt Eberle builds a transition to the second part of the Workshop discussing a commercial translation systems treatment of multiwords.


The first part of the workshop with invited lectures will take place in the morning (09/30) between 10:45 to 13:15.

The second part of the workshop with lectures concerning multi-word units and their application in machine translation is part of the workshop "Multi-Word Units - Extraction, Representation and Automatic Translation"

Contact and Organizers

Kurt Eberle
Uta Seewald-Heeg
Daniel Stein
Lothar Lemnitzer

Contact address for submission of contributions: Lothar Lemnitzer, lemnitzer(at)bbaw.de