Fast word-graph generation for spontaneous conversational speech translation

This paper introduces the latest advances in research at ATR on speech translation for spontaneous conversations, especially focusing on speech recognition e orts. For recognition, we employ a word search technique that generates moderate sized word graphs in real-time. To cope with a variety in length of utterances, e.g., word, phrase, sentence fragment, sentence, and concatenated sentences in spontaneous speech, we have adopted a two pass search strategy that uses variable-order word n-gram statistics in the rst stage and task dependent language constraints in the second stage. This strategy is evaluated using the \ATRTravel Arrangement" corpus.