Prof. John Hutchins
University of East Anglia

Current developments in machine translation: an overview of progress and future prospects

This talk will review current developments in machine translation, both activity in machine translation (MT) research and trends in the applications and usage of MT systems.

In research, the principal developments have been the use of corpus-based methods (rather than older rule-based approaches, using grammars and dictionaries), in particular statistical machine translation and example-based machine translation . I shall describe the basic methods. With these developments have also come improved methods of alignment, shallow parsing and tagging, automatic evaluation, translation memories, and new kinds of translation tools.

I shall next survey how MT systems and computer-based translation tools are being used (in large organisations, by government, and for personal use, including online services), and new applications of MT (including speech translation and the embedding of MT in other NLP systems)

The final parts will consider various methods for improving the output quality of existing systems (the use of controlled languages and restriction to specific domains), the evaluation of systems, the integration of different methods and techniques to take advantage of their various strengths, the development of systems for languages with scarce resources and for (commercially) minor languages, ending with a summary of future prospects in general.