Back to the Future: The Case for Direct English-Swedish MT

In this presentation we will report work on a data-driven direct machine translation system that has been under development for some years in the Linköping part of the VINNOVA-sponsored project “Corpus-based Machine Translation”. An important aim for the system is that it should support quick development of domain-restricted machine translation. Thus, the system relies on a development environment with alignment and analysis tools that can produce the data required for a given translation task from a parallel corpus.

We currently call the system T4F, which is an abbreviation for tokenization, tagging, transfer, transposition and filtering, all of which are main phases/modules of the system. Other modules such as ranking of alternative translations are used in the system, but have not (yet) been deemed to be important enough to enter into the acronym.