Sixth European Conference on Speech Communication and Technology
Spoken dialogue systems have to cope with well known problems of spontaneous speech such as ungrammaticalities, hesitations or corrections. Besides, unrestricted speech poses the problem of paraphrasing, rising the number of utterances exceeding the linguistic coverage of the system. Concerning the linguistic processing of spontaneous speech, partial parsing represents a good method to achieve suficient robustness against these phenomena. In simple application domains however, mostly systems are implemented that do completely without speech understanding. In our paper we present an approach for robust linguistic speech processing and understanding for simple application domains. Our approach provides a complete analysis of user utterances based on partial analysis of semantically relevant sub-parts of the utterance. The semantic representations may be directly mapped onto data base indices thus bypassing the interpretation by the dialogue manager. On the other hand, the integration into complex dialogue systems allows for free dialogues and recovery strategies in case of system faults and misunderstandings.
Full Paper (PDF) Gnu-Zipped Postscript
Bibliographic reference. Boros, Manuela / Heisterkamp, Paul (1999): "Linguistic phrase spotting in a simple application spoken dialogue system", In EUROSPEECH'99, 1983-1986.