Multi-Engine and Multi-Alignment Based Automatic Post-Editing and its Impact on Translation Productivity

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2559–2570, Osaka, Japan, December 11-17 2016. Multi-Engine and Multi-Alignment Based Automatic Post-Editing and its Impact on Translation Productivity Santanu Pal1, Sudip Kumar Naskar2, Josef van Genabith1,3 1Saarland University, Saarbr¨ucken, Germany 2Jadavpur University, Kolkata, India 3German Research Center for Artificial Intelligence (DFKI), Germany {santanu.pal, josef.vangenabith}@uni-saarland.de sudip.naskar@cse.jdvu.ac.in Abstract In this paper we combine two strands of machine translation (MT) research: automatic post- editing (APE) and multi-engine (system combination) MT. APE systems learn a target-language- side second stage MT system from the data produced by human corrected output of a first stage MT system, to improve the output of the first stage MT in what is essentially a sequential MT system combination architecture. At the same time, there is a rich research literature on parallel MT system combination where the same input is fed to multiple engines and the best output is selected or smaller sections of the outputs are combined to obtain improved translation output. In the paper we show that parallel system combination in the APE stage of a sequential MT-APE combination yields substantial translation improvements both measured in terms of automatic evaluation metrics as well as in terms of productivity improvements measured in a post-editing experiment. We also show that system combination on the level of APE alignments yields further improvements. Overall our APE system yields a statistically significant improvement of 5.9% relative BLEU over a strong baseline (English–Italian Google MT) and 21.76% productivity increase in a human post-editing experiment with professional translators. 1 Introduction The term Post-Editing (PE) is defined as the correction performed by humans over the translation pro- duced by an MT system (Veale and Way, 1997). It is often understood as the process of improving a translation provided by an MT system with the minimum amount of manual effort (TAUS Report, 2010). While MT is often not perfect, post-editing MT can yield productivity gains as post-editing MT output may require less effort compared to translating the same input manually from scratch. MT outputs are often post-edited by professional translators and the use of MT has become an important part of the translation workflow. A number of studies confirm that post-editing MT output can improve translators’ performance in terms of productivity and it may positively impact on translation quality and consistency (Guerberof, 2009; Plitt and Masselot, 2010; Zampieri and Vela, 2014). The wide use of MT in modern translation workflows in the localization industry, in turn, has resulted in substantial quantities of PE data which can be used to develop APE systems. APE (Knight and Chander, 1994) has been proposed as an automatic method for improving raw MT output, before performing actual human post-editing on it. The approach is based on collecting human corrected output of a first stage MT system and using this to train a system to correct errors produced by the MT system, possibly resulting in a productivity increase in the translation process. The advantage of APE relies on its …

Content truncated to 500 words