Bringing humans into the loop: Localization with MT at Traslán

Bringing Humans into the Loop: Localization with Machine Translation at Trasl´an Declan Groves Trasl´an Co. Wicklow, Ireland dgroves@traslan.ie http://www.traslan.ie Abstract Trasl´an makes full use of MT during our translation workflow, where the raw output from our Machine Translation (MT) system is passed onto human translators who perform post-editing (if necessary) to arrive at the final translation. Within Trasl´an we have found that using MT has enabled us to increase the speed, accu- racy and consistency of translation - elements which allow us to process larger amounts of translation; with quicker turnaround times, which in turn has resulted in overall savings of approx. 20% so far. One of the main challenges in using MT within a commercial setting is getting human translators to adopt and make full use of the technology. Within Trasl´an we overcome this obstacle by working closely and intensively with our translators, getting them involved di- rectly in the development process. Doing so enables translators in turn to train new users of the system and to communicate effectively to other translators the benefits of integrating MT into the translation pipeline. 1 Introduction As the demand for commercial translation increases the demand for automation increases with MT, to- gether with post-editing, presenting itself as a prac- tical solution. User acceptance of MT is vital in implementing this type of technology in a commer- cial setting. Although the quality of the MT out- put plays an important role in this acceptance, it is not the sole decisive factor; user perception is key (Hutchins, 2001). However, historically, translators are not known for their open acceptance and use of MT technology. Common arguments from transla- tors against the use of MT include a dislike for cor- recting repetitive errors, a fear of losing language proficiency by working with poor quality output, and a dislike of having one’s freedom of expression lim- ited (O’Brien, 2006). Often, MT is wrongly per- ceived as a replacement for translators, when in fact it needs to be viewed as one of several tools in the users’ work environment to help their productivity and efficiency. It is important that translators have a positive view of MT and that they are not in any way made to feel redundant (Valderr´abanos et al., 2003). After all, there will always be a need for hu- man translators; MT will only ever serve to comple- ment human translation, not to replace it (DePalma, 2006). As there is relatively little research available dis- cussing the successful implementation of MT within translation companies, in this paper we aim to shed some light on our own experiences with MT and how we manage to keep our human translators in the loop when translating as part of a localization task. In MT research, systems are often developed without any consideration for who the final user may be, however, in a commercial setting the user is paramount. We are not seeking translation per- fection, but instead we need to base our evaluation of system performance on the …

Content truncated to 500 words