5th International Conference on Spoken Language Processing

Sydney, Australia
November 30 - December 4, 1998

VPQ: A Spoken Language Interface to Large Scale Directory Information

Bruce Buntschuh (1), Candace A. Kamm (1), Giuseppe Di Fabbrizio (1), Alicia Abella (1), Mehryar Mohri (1), Shrikanth Narayanan (1), I. Zeljkovic (1), R.D. Sharp (1), Jeremy H. Wright (1), S. Marcus (2), J. Shaffer (3), R. Duncan (3), J.G. Wilpon (1)

(1) AT&T Labs - Research, USA
(2) International Asset Systems Ltd, USA
(3) Mississippi State University, USA

VPQ (Voice Post Query) is a dialog system for spoken access to information in the AT&T personnel database (>120,000 entries). An explicit design goal is for the initial interaction with the system to be rather unconstrained and to rely on tighter, prompt-constrained, dialog only when absolutely necessary. The purpose of VPQ is to explore and exploit the capabilities of ``state-of-the-art'' speech recognition systems for this high-perplexity task, and to develop the natural language understanding and dialog control components necessary for effective and efficient user interactions. The VPQ task includes simple interactions, where the initial request is unambiguous and the system's response provides the desired information or completes a call to the requested person, as well as more complex interactions where ambiguities or errors require dialog-driven resolution. This paper highlights the inherent challenges in this task, the major components of the system, the rationale for their design, and how they perform.

Full Paper

Bibliographic reference.  Buntschuh, Bruce / Kamm, Candace A. / Fabbrizio, Giuseppe Di / Abella, Alicia / Mohri, Mehryar / Narayanan, Shrikanth / Zeljkovic, I. / Sharp, R.D. / Wright, Jeremy H. / Marcus, S. / Shaffer, J. / Duncan, R. / Wilpon, J.G. (1998): "VPQ: a spoken language interface to large scale directory information", In ICSLP-1998, paper 0877.