Predicting dialogue acts for a speech-to-speech translation system
Norbert Reithinger, Engel, Ralf, Michael F. Kipp, Martin Klesen · 1996
Bundesministerium fur Bildung Wissenschaft Forschung und Technologie BMBF unter dem Forderkennzeichen IV K gefordert Die Verantwortung fur den Inhalt dieser Arbeit liegt bei dem Autor We present the application of statistical language modeling methods for the prediction of the next dialogue act This prediction is used by dierent modules of the speechtospeech translation system VERBMOBIL The sta tistical approach uses deleted interpolation of ngram frequencies as basis and determines the interpolation weights by a modied version of the standard optimization algorithm Additionally we present and evaluate dierent ap proaches to improve the prediction process eg including knowledge from a dialogue grammar Evaluation shows that including the speaker information and mirroring the data delivers the best results