Translation By Abduction

Abstract

We propose an approach to machine translation using abduction. This approach overcomes the common machine translation bottleneck by allowing mapping of information from the source to the target language at a variety of levels from the most superficial to levels requiring deep interpretation and access to knowledge about the world, the context, and the speech act situation. We illustrate the approach with a translation of an English sentence into Japanese, and show how we might handle outstanding hard problems in a translation between this language pair.