Interactive Machine Translation using Hierarchical Translation Models

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 244–254, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Interactive Machine Translation using Hierarchical Translation Models Jes´us Gonz´alez-Rubio, Daniel Ortiz-Mart´ınez, Jos´e-Miguel Bened´ı, Francisco Casacuberta D. de Sistemas Inform´aticos y Computaci´on Universitat Polit`ecnica de Val`encia Camino de Vera s/n, 46021 Valencia (Spain) {jegonzalez,dortiz,jbenedi,fcn}@dsic.upv.es Abstract Current automatic machine translation sys- tems are not able to generate error-free trans- lations and human intervention is often re- quired to correct their output. Alternatively, an interactive framework that integrates the human knowledge into the translation pro- cess has been presented in previous works. Here, we describe a new interactive ma- chine translation approach that is able to work with phrase-based and hierarchical translation models, and integrates error-correction all in a unified statistical framework. In our experi- ments, our approach outperforms previous in- teractive translation systems, and achieves es- timated effort reductions of as much as 48% relative over a traditional post-edition system. 1 Introduction Research in the field of machine translation (MT) aims to develop computer systems which are able to translate between languages automatically, with- out human intervention. However, the quality of the translations produced by any automatic MT sys- tem still remain below than that of human transla- tion. Typical solutions to reach human-level quality require a subsequent manual post-editing process. Such decoupled post-edition solution is rather inef- ficient and tedious for the human translator. More- over, it prevents the MT system from taking advan- tage of the knowledge of the human translator and, reciprocal, the human translator cannot take advan- tage of the adapting ability of MT technology. An alternative way to take advantage of the exist- ing MT technology is to use them in collaboration with human translators within a computer-assisted translation (CAT) or interactive framework (Isabelle and Church, 1998). The TransType and TransType2 projects (Foster et al., 1998; Langlais and Lapalme, 2002; Barrachina et al., 2009) entailed an interesting focus shift in CAT technology by aiming interaction directly at the production of the target text. These research projects proposed to embed an MT system within an interactive translation environment. This way, the human translator can ensure a high-quality output while the MT system ensures a significant gain of productivity. Particularly interesting is the interactive machine translation (IMT) approach pro- posed in (Barrachina et al., 2009). In this scenario, a statistical MT (SMT) system uses the source sen- tence and a previously validated part (prefix1) of its translation to propose a suitable continuation. Then the user finds and corrects the next system error, thereby providing a longer prefix which the system uses to suggests a new, hopefully better continua- tion. The reported results showed that IMT can save a significant amount of human effort. Barrachina et al,. (2009) provide a thorough de- scription of the IMT approach and describe algo- rithms for its practical implementation. Neverthe- less, we identify two basic problems for which we think there is room for improvement. The first prob- lem …

Content truncated to 500 words