Machine interpretation

This article,¹ discusses the problems of machine translation focusing on a special task, i.e. automatic dialogue interpreting. In the translation of spoken discourse, there is an increasing tendency to use pragmatic information which is commonly associated with speech or dialogue acts in order to handle the problems encountered in speech processing. After discussing some characteristics of spoken language as well as principal constraints operating on speech-to-speech translation, we will give a very brief overview of existing approaches, and then switch to the extended description of a sample solution within the VERBMOBIL-Project where transfer is assisted by a dialogue-act component. The last section describes how, with limited use of available information, transfer based on dialogue acts might result in new solutions for existing problems.