My research interests lie in the areas of conversational agents and spoken language interfaces. Broadly speaking, I am interested in developing techniques that allow interactive systems to act robustly in the real-world, and to acquire knowledge automatically and learn from their own experiences (i.e. without explicit supervision from their developers.)
My dissertation is focused on developing techniques for increasing the robustness and reliability of spoken language interfaces. Some of the questions under scrutiny are: how does a system “know that it doesn’t know”? How do we develop systems that can monitor and accurately update their beliefs? What set of strategies can be used to set a conversation back on track, and what are the typical user behaviors in response to these strategies? What techniques can be used to learn the optimal system behavior on-line, from detected error segments, and how do we make these systems adapt and improve their performance over time?
As infrastructure for my research in error handling, I have developed RavenClaw, a freely-available, open-source dialog management framework for task-oriented domains. RavenClaw has already been used to build a number of spoken dialog systems spanning different domains and interaction types. Some of them, such as RoomLine, Let's Go! Public and Conquest have been successfully deployed into the real-world.
I am currently organizing the Dialogs on Dialogs reading group at CMU, focused on research and current developments in dialog management and generally the upper levels of spoken dialog systems. If you're interested in joining this group, let me know. Together with other members of the Dialogs on Dialogs group, I was an organizer for the 1st Young Researchers' Roundtable on Spoken Dialog Systems workshop. The workshop was held on September 1st 2005, in conjunction with Interspeech-2005. A second edition was succesfully held this year.