Domain-specific English-Czech Neural Machine Translation

This project is part of the Master's thesis of Martin Wörgötter (IS MU thesis archive).


We present the state-of-the-art neural machine translation framework, explain the principles behind training neural machine translation systems and summarize the difficulties of achieving decent quality results. Next, we provide a brief survey on two state-of-the-art systems capable of English-Czech translation. And finally, we demonstrate our experiments of translating legal texts from the EUR-Lex corpus. We show that it is possible to achieve results on par with statistical machine translation systems on the translation of legal documents.


Access to computing and storage facilities owned by parties and projects contributing to the National Grid Infrastructure MetaCentrum provided under the programme "Projects of Large Research, Development, and Innovations Infrastructures" (CESNET LM2015042), is greatly appreciated.