Abstract
This paper discusses comparative analysis of word-level time delay between Japanese-English and English-Japanese simultaneous interpretations. For this investigation we used the Simultaneous Interpretation Database of Nagoya University in order to conduct quantitative analysis that requires a large scale corpus. Since temporal information was provided only to each utterance in this database, the effective use of it by providing word-level temporal information on the introduction of speech recognition techniques enabled us to observe a large amount of interpreted words with time delay. We analyzed 4,468 pairs of interpreted words retrieved from Japanese- English (J-E) interpretation data and 2,629 pairs of those from English- Japanese (E-J) interpretation data. As a result, it became clear that time delay in E-J interpretation was shorter than that in J-E interpretation and that part-of-speech and grammatical function of words of the source language made distinguished distribution of time-delay in the interpretation.
