Spoken Dialogue System Development With User SimulationTuesday 8th August 2006 at 11am
AbstractRecently user simulation has become an attractive topic for developers of dialogue interfaces which, due to the numerous software components, require extensive development time to configure and refine. At early stages, there is a chicken-and-egg problem whereby good quality user data cannot be obtained without a reasonably robust system. This talk will present work in creating a conversational interface that provides information about restaurant listings with the aid of a user simulator. We will discuss how synthetic user dialogues are created and how these have accelerated system prototyping. We detail one application of the simulator where it is used to automatically induce domain dependent data used to train the speech recognizer language model. The method can yield a synthetic training corpus from out-of-domain data and a small amount of in-domain development data. Short resumeFrom 2001, Grace Chung worked as a senior research scientist in Spoken Dialogue Systems, at the Corporation for National Research Initiatives, Virginia, USA. Jointly, she held an adjunct teaching position at Georgetown University, Washington DC. Prior to Washington, Grace was a researcher at the Spoken Language Systems Group at the MIT Laboratory for Computer Science where she received her Ph.D. (2001) and S.M. (1998). She also has a B.Eng. in Electrical Engineering and B.Sc. from the University of New South Wales. Her research interests focus on many aspects of spoken dialogue systems development. These include modeling unknown words and dynamic vocabulary for speech recognition to rapid prototyping for dialogue systems development with user simulation. |