Mid-Air Hand Gestures for Post-Editing of Machine Translation

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 6763–6773 August 1–6, 2021. ©2021 Association for Computational Linguistics 6763 Mid-Air Hand Gestures for Post-Editing of Machine Translation Rashad Albo Jamara1, Nico Herbig1, Antonio Kr¨uger1, Josef van Genabith1,2 1German Research Center for Artificial Intelligence (DFKI), Saarland Informatics Campus, Germany 2Department of Language Science and Technology, Saarland University, Germany rashad.jamara@gmail.com {nico.herbig, krueger, josef.van genabith}@dfki.de Abstract To translate large volumes of text in a globally connected world, more and more translators are integrating machine translation (MT) and post-editing (PE) into their translation work- flows to generate publishable quality transla- tions. While this process has been shown to save time and reduce errors, the task of trans- lation is changing from mostly text production from scratch to fixing errors within useful but partly incorrect MT output. This is affecting the interface design of translation tools, where better support for text editing tasks is required. Here, we present the first study that investi- gates the usefulness of mid-air hand gestures in combination with the keyboard (GK) for text editing in PE of MT. Guided by a ges- ture elicitation study with 14 freelance trans- lators, we develop a prototype supporting mid- air hand gestures for cursor placement, text se- lection, deletion, and reordering. These ges- tures combined with the keyboard facilitate all editing types required for PE. An evaluation of the prototype shows that the average edit- ing duration of GK is only slightly slower than the standard mouse and keyboard (MK), even though participants are very familiar with the latter, and relative novices to the former. Fur- thermore, the qualitative analysis shows posi- tive attitudes towards hand gestures for PE, es- pecially when manipulating single words. 1 Introduction In a well-connected world, translation is of ever- increasing importance (Bassnett, 2013). To meet translation demands, machine translation (MT) is often employed as a cheaper and faster alterna- tive to human translation (HT) (O’Brien, 2012). Even though MT has improved drastically over the last 5 years, discussions about reaching hu- man parity are still ongoing (L¨aubli et al., 2020) and limited to a small set of language pairs and domains for which ample training data is avail- able. For most application scenarios, however, MT quality is far from reaching the quality of highly trained professionals. In an attempt to combine the best of both worlds, post-editing (PE) is becom- ing common practice, where human translators use raw MT output and make the necessary changes to produce an acceptable level of quality (Kopo- nen, 2016). Although translators have approached PE with fear and skepticism (Lagoudaki, 2009), more recent studies found that nowadays transla- tors are more open to it and that much of the origi- nal dislike was attributed to outdated perceptions of MT quality (Plitt and Masselot, 2010; Green et al., 2013). Independent of translators’ percep- tions, studies found that PE increases productivity and decreases errors compared to translation from scratch (Green …

Content truncated to 500 words