The NAIST Simultaneous Translation Corpus

Abstract

This chapter describes an English-Japanese/Japanese-English simultaneous interpretation corpus collected at the Nara Institute of Science and Technology (NAIST). There are two main features of the corpus that set it apart from others. The first is that it contains recorded interpretation results from professional simultaneous interpreters with different amounts of experience. This makes it possible to compare the differences between interpreters of different levels, elucidating the effect of interpreter experience on the objective and subjective qualities of results. The second feature is that part of the corpus also has been translated. This data makes it possible to compare and contrast the results when a particular talk is translated from text without time constraints (using the translation data) or from speech with time constraints (using the simultaneous interpretation data). The corpus contains a total of 387k words worth of data, with the material covering lectures and news. All transcriptions are time aligned. The corpus will be helpful to analyze differences in interpretation styles, and may also be used as a reference in the construction of simultaneous interpretation systems.