E.g., 06/17/2018
E.g., 06/17/2018
Free for GALA Members
Free for Non-Members

Details

QT21, a research project on Machine Translation (MT) funded by the European Commission, has come to a successful end. This webinar will explore the many ways QT21 impacted the MT landscape. We will review the paradigm shift toward the use of Neural Network based machine translation (NMT), which allowed QT21 to win more than 80% of all news translation tasks at WMT (the “Olympic Games” for MT) in 2016 and 2017. We will present the highlights of NMT for morphologically rich languages and show what can be done when not enough training data is available to profit from NMT. In addition, looking into automatic continuous learning from errors, we will illustrate how online Automatic Post Editing can iteratively adapt to the specifics of a translation task. Finally, we will exemplify the differences between former phrase-based statistical MT systems and the new state-of-the-art NMT, and explore what impact these differences have on post-editing activities during the localization process.

This webinar is part of the Research and Innovation Action "Quality Translation 21 (QT21)." This project has received funding from the European Union's Horizon 2020 program for ICT under grant agreement no. 645452.

Josef van Genabith

Josef van Genabith is a Scientific Director at DFKI, the German Research Centre for Artificial Intelligence, where he heads the Multilingual Technologies Group (MLT), and jointly with Prof. Hans Uzskoreit, the Language Technology Lab (LT). He holds the Chair of Translation-Oriented Language Technologies at the University of the Saarland, Germany. He was the Founding Director of CNGL, the Centre for Next Generation Localisation (now ADAPT), in Dublin, Ireland, and was a Professor in the School of Computing at Dublin City University (DCU), Ireland.

randomness