In our previous research, we have shown that prosody can be used to dramatically improve the performance of the automatic speech translation system VERBMOBIL [9]. The methods to classify prosodic events have been developed on the German sub-corpus of the VERBMOBIL speech database. In this paper we describe how the methods that we developed on the German sub-corpus can be applied to other languages. Preliminary experiments show that these methods are suited for English and Japanese, as well. Efficiency problems are addressed and a new set of features that eliminates most of these problems is presented. The new set of features facilitates a multi-lingual module for prosodic processing. We present an architecture for such a multi-lingual module and discuss the advantages of this approach compared to an approach that uses separate modules for different languages. This multi-lingual module and the new feature set are evaluated w.r.t. computation time, memory requirement, and classification performance. Preliminary results show that the memory requirement can be reduced by at least 70%, whereas the recognition accuracy does not decrease.
Cite as: Buckow, J., Huber, R., Warnke, V., Batliner, A., Nöth, E., Niemann, H. (1999) Multi-lingual prosodic processing. Proc. ETRW on Dialogue and Prosody, 157-162
@inproceedings{buckow99_diapro,
author={Jan Buckow and Richard Huber and Volker Warnke and Anton Batliner and Elmar Nöth and Heinrich Niemann},
title={{Multi-lingual prosodic processing}},
year=1999,
booktitle={Proc. ETRW on Dialogue and Prosody},
pages={157--162}
}