Modeling Pronunciation Variation for Automatic Speech Recognition Rolduc, The Netherlands May 4-6, 1998

Pronunciation Modelling in the RWTH Large Vocabulary Speech Recognizer

K. Beulen, S. Ortmanns, A. Eiden, S. Martin, L. Welling, J. Overmann, Hermann Ney

Lehrstuhl fur Informatik VI, RWTH Aachen, University of Technology, Aachen, Germany

In this paper we describe the application of pronunciation variants for our large vocabulary continuous speech recognizer. We will explain how the pronunciation variants were used in training and recognition and give some recognition results on three different corpora. The recognition tests were performed on the Wall Street Journal (WSJ) November 92 development and evaluation corpora (5 000 words), the North American Business (NAB) HI development corpus (20000 words) and on the Verbmobil 1996 evaluation corpus (5 000 words). For the WSJ and NAB corpora, a slight improvement in recognition accuracy can be observed, while for the Verbmobil corpus the error rate remains unchanged. In addition, we will discuss the incorporation of phrases in combination with pronunciation variants in the pronunciation lexicon as well as the language model. The recognition results on the WSJ November 92 development and evaluation corpora show that the main improvement due to phrases is caused by the language model.

Full Paper

Bibliographic reference.  Beulen, K. / Ortmanns, S. / Eiden, A. / Martin, S. / Welling, L. / Overmann, J. / Ney, Hermann (1998): "Pronunciation modelling in the RWTH large vocabulary speech recognizer", In MPV-1998, 13-16.