Skip to main content Skip to main navigation

Publication

Contextual Constraints for MT

Walter Kasper; Kurt Eberle; Christian Rohrer
In: Proceedings of the 4th International Conference on Theoretical and Methodological Issues in Machine Translation. International Conference on Theoretical and Methodological Issues in Machine Translation (TMI-92), 4th, June 25-27, Montreal, QC, Canada, Pages 213-224, 1992.

Abstract

We outline an experimental machine translation system currently under development at the Institute for Computational Linguistics which aims at flexibility with respect to the level of transfer. The core of the system is a bidirectional transfer system using LFG-grammars for source and target languages. If translation is not possible on the basis of the linguistic information provided by the core system, i.e. in the case of translation mismatches, the system makes selective use of contextual and background knowledge. To this end we developed the concept of contextual constraints which are the interface between the core system and a module that we call the contextual resolver which deals among other things with the contextual information provided by the preceding text using Discourse Representation Theory. We concentrate on problems such as the translation of tense and anaphoric expressions between German and French.