Design of a hybrid high quality machine translation system
Bogdan Babych, Kurt Eberle, Johanna Geiß, Mireia Ginestí-Rosell, Anthony F. Hartley, Reinhard Rapp, Serge Sharoff, Martin Thomas · 2012
This paper gives an overview of the ongoing FP7 project HyghTra (2010--2014). The HyghTra project is conducted in a partnership between academia and industry involving the University of Leeds and Lingenio GmbH (company). It adopts a hybrid and bootstrapping approach to the enhancement of MT quality by applying rule-based analysis and statistical evaluation techniques to both parallel and comparable corpora in order to extract linguistic information and enrich the lexical and syntactic resources of the underlying (rule-based) MT system that is used for analysing the corpora. The project places special emphasis on the extension of systems to new language pairs and corresponding rapid, automated creation of high quality resources. The techniques are fielded and evaluated within an existing commercial MT environment.