JANUS 93: towards spontaneous speech translation

We present rst results from our eorts toward translation of spontaneously spoken speech. Improvements include increasing coverage, robustness, generality and speed of JANUS, the speech-to-speech translation system of Carnegie Mellon and Karlsruhe University. Recognition and Machine Translation Engine have been upgraded to deal with requirements introduced by spontaneous human to human dialogs. To allow for development and evaluation of our system on adequate data, a large database with spontaneous scheduling dialogs is being gathered for En-glish, German and Spanish. 1. OVERVIEW JANUS [1, 2] has been among early systems to attempt the translation of spoken dialogs. It had initially been built based on a speech database of 12 read dialogs of the conference registration task, encompassing a vocabulary of around 500 words. It was designed as a speaker-independent system which translates spoken utterances from English and also from German into one of German, English or Japanese.