Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Incremental Elaboration in Generating and Interpreting Spontaneous Speech

Michio Okada, Satoshi Kurihara, Ryohei Nakatsu

NTT Basic Research Laboratories, Kanagawa, Japan

In this paper, we outline a computational model for incremental elaboration and some other disfluencies in spontaneous speech. In the model, spontaneity and disfluencies are assumed to be behaviors that emerge as the result of many local interactions between routine activities and their environment. Incremental elaboration can be thought to arise from interaction among heterogeneous modules. In the model presented here, we also introduce a framework for accomplishing desired actions with an under-specification control strategy. To specify the behavior, this strategy tries to indirectly arrange conditions (a situation) for the emergence of the desired actions. We then show that incremental elaboration is also a strategy for generating and interpreting an utterance with minimal specification. The model presented here provides a way of thinking about the design of an interpreting system that is robust and efficient architecture for spontaneous speech.

Full Paper

Bibliographic reference.  Okada, Michio / Kurihara, Satoshi / Nakatsu, Ryohei (1994): "Incremental elaboration in generating and interpreting spontaneous speech", In ICSLP-1994, 103-106.