SEECAT: ASR & Eye-tracking enabled computer-assisted translation

SEECAT: ASR & Eye-Tracking Enabled Computer-Assisted Translation Mercedes Garc´ıa-Mart´ınez1, Karan Singla2, Aniruddha Tammewar2, Bartolom´e Mesa-Lao1, Ankita Thakur3, Anusuya M.A.4, Srinivas Bangalore5, Michael Carl1 1CRITT - Copenhagen Business School, Denmark, 2IIIT Hyderabad, India, 3JSSATE, Noida, India 4SJCE, Mysore,India, 5AT&T Research Labs, United States Abstract Typing has traditionally been the only in- put method used by human translators working with computer-assisted transla- tion (CAT) tools. However, speech is a nat- ural communication channel for humans and, in principle, it should be faster and easier than typing from a keyboard. This contribution investigates the integration of automatic speech recognition (ASR) in a CAT workbench testing its real use by human translators while post-editing ma- chine translation (MT) outputs. This pa- per also explores the use of MT com- bined with ASR in order to improve recog- nition accuracy in a workbench integrat- ing eye-tracking functionalities to collect process-oriented information about trans- lators’ performance. 1 Introduction Human-aided machine translation is gradually be- coming a common practice for language service providers (LSPs) as opposed to machine-aided hu- man translation. Depending on the nature of the text, more and more LSPs pre-translate the source text using existing translation memories (TMs) and then automatically translate the remaining text us- ing an MT engine. Then human translators cor- rect and adapt, i.e. post-edit, the output from both TMs and MT to produce different levels of transla- tion quality. Improving and maximizing the poten- tials of a post-editing workbench is thus one of the c⃝ 2014 The authors. This article is licensed under a Creative Commons 3.0 licence, no derivative works, attribution, CC- BY-ND. priorities set by both the industry and the research community (Mesa-Lao, 2012). The motivation be- hind this paper comes from a desire to know how different input modalities in a computer-assisted translation (CAT) workbench can be of greater support to translation professionals. Keyboards are the most widely used input de- vice for text production and they seem to be the easiest input method when only minor changes are needed. However, in the context of post-editing, when the text requires major changes (e.g. editing larger segments of text), typing could be optimized using other input modalities. Moreover, if the post- editor is not a touch typist, then she has to switch visual attention back and forth between the screen and the keyboard making the task more complex. A possible solution for this profile of users could be the use of other input methods, such as ASR or hand-writing, in addition to traditional typing (Hauptmann and Rudnicky, 1990). The comparison between ASR and typing as in- put methods can be done based on task duration, i.e. measuring the time needed to type against the ASR rate including possible corrections to fix recognition inaccuracies. Studies on input dura- tions have shown that ASR input can be faster (Chen, 2006; Vidal et al., 2006). This paper is structured along the following lines: The first section presents the CASMACAT1 1CasMaCat: Cognitive Analysis and Statistical Methods for Advanced Computer Aided Translation. …

Content truncated to 500 words