Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

Integrating Multiple Knowledge Sources for Word Hypotheses Graph Interpretation

Volker Warnke, Florian Gallwitz, Anton Batliner, Jan Buckow, R. Huber, Elmar Nöth, A. Höthker

Universität Erlangen-Nürnberg, Lehrstuhl für Mustererkennung (LME), Erlangen, Germany

We present an integrated approach for the interpretation of word hypotheses graphs (WHGs) using multiple knowledge sources. Commonly, different knowledge sources in speech understanding are applied sequentially. Typically, speech understanding systems, such as the {\vm} speech-to-speech translation system, first use a word recognizer to determine word hypotheses, only based on acoustic and language model (LM) information. The resulting word sequences or WHGs are then segmented according to syntactic and/or prosodic information. Finally, these segments are interpreted by a parser or a stochastic process. Thus, it is impossible to use the knowledge of the syntactic-prosodic process, the parser or any other subsequent process to find the best word sequence. In our new approach we use acoustic, prosodic and LM information to determine the best word chain, to detect syntactic/prosodic/pragmatic phrase boundaries and to classify dialog acts in one integrated search procedure, based on a WHG or a word lattice.

Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Warnke, Volker / Gallwitz, Florian / Batliner, Anton / Buckow, Jan / Huber, R. / Nöth, Elmar / Höthker, A. (1999): "Integrating multiple knowledge sources for word hypotheses graph interpretation", In EUROSPEECH'99, 235-238.