Hironobu Takagi

Development of an eye-movement enhanced translation support system

Translation support systems have the ability to retrieve information quickly from large databases and dictionaries. However, current systems lack the ability to present relevant information automatically, at just the right moment, without interruptions. To allow a user seamless retrieval, an ideal system would be able to observe a translators behavior, predict their translation processes and automatically retrieve appropriate information at just the right moment. We focused on eye movements as an aspect of user behavior that can reveal users translation processes. They move their eyes during translation, both to read text and to obtain various information. We thought that eye movements could therefore provide a valuable insight into the translation process. Based on this fundamental idea, we investigated eye movements during translation and found two characteristic patterns. One indicates that a user is at a standstill in the translation process and the other reveals when superfluous information is being displayed. Then we developed a translation support system that is able to detect these patterns. The system automatically retrieves information from the corpora databases at the appropriate time, and removes displayed information when the system predicts that it is superfluous. We refer to the results of preliminary experiments indicating the relation between eye movements and translation processes. Then we describe the implementation of the translation system, review related works, discuss the systems effects and problems, and end by considering the future direction of our research.