Eye-tracking Post-editing Behaviour in an Interactive Translation Prediction Environment

Post-editing machine translation is becoming a common practise within the localization industry as opposed to full human translation. Improving the potentials of a post-editing workbench is thus one of the priorities set by both the industry and researchers when addressing the technological challenges faced by translators. This paper reports on a preliminary pilot test within the CasMaCat project. Based in user activity data (key-logging and eye-tracking), this project aims at defining the functionalities of a new translator's workbench focusing on post-editing and advanced computer-aided translation methods. The main aim of this preliminary pilot was to assess one of the new features implemented in the second prototype of the workbench: the interactive translation prediction (ITP) feature. This ITP feature is set to provide translators with different suggestion as they post-edit. For this purpose 6 translators were asked to post-edit 1,000 words from English to Spanish in five different tasks while their eye movements were being tracked. Each task was designed to test different modalities of ITP. Translators were also asked to fill out a questionnaire expressing their attitudes towards each ITP modality. Eye-tracking data was used to correlate participant's satisfaction with ITP and the post-editing behaviour revealed by the eye-tracker.