Towards Deeper MT - A Hybrid System for German

Eleftherios Avramidis, Aljoscha Burchardt, Maja Popovic, Hans Uszkoreit

In: Jan Hajič, António Branco (Hrsg.). Proceedings of the 1st Deep Machine Translation Workshop. Deep Machine Translation Workshop (DMTW-2015) September 3-4 Prague Czech Republic Seiten 12-19 ISBN 978-80-904571-7-1 Charles University in Prague, Faculty of Mathematics and Physics, Institute of Formal and Applied Linguistics 9/2015.


The idea to improve MT quality by using deep linguistic and knowledge-driven information has frequently been expressed. If the goal is to use deep information for building an MT system, there are two extreme options: (1) to start from a purely knowledge-driven approach (RBMT) and try to arrive at the same recall found in current SMT systems; (2) to start from an SMT system and try to arrive at higher precision by modifying it so that more knowledge drives the translation process. The system architecture we will describe in this paper starts in the middle of these extreme options. It is a hybrid architecture that we take as a starting point for future experiments and extensions to increase MT quality by more knowledge-driven processing.


deeper-mt-hybrid.pdf (pdf, 203 KB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence