Incremental Language Processing
One of my research goals is to help provide dialogue systems with the ability to incrementally understand, predict, and respond to human speech in real-time. This can make speaking with systems more natural, by enabling systems to quickly signal their understanding of speech using human-like feedback mechanisms such as back-channels, interruptions, and collaborative completions.
Demo Videos
A demonstration of high-speed incremental understanding of image descriptions:
See our SigDial 2015 paper for the details.
A demonstration of collaborative utterance completion in the SASO-EN virtual human dialogue system (4.2M MP4 video)
See our SIGDIAL 2009 paper for the details.
Selected Publications
-
Using Reinforcement Learning to Model Incrementality in a Fast-Paced Dialogue Game, Ramesh Manuvinakurike, David DeVault, and Kallirroi Georgila, in the 18th Annual SIGdial Meeting on Discourse and Dialogue (SIGDIAL), 2017.
-
"So, which one is it?" The effect of alternative incremental architectures in a high-performance game-playing agent Maike Paetzel, Ramesh Manuvinakurike, and David DeVault, in the 16th Annual SIGdial Meeting on Discourse and Dialogue (SIGDIAL), 2015. Best Paper Award Winner
-
Incremental interpretation and prediction of utterance meaning for interactive dialogue David DeVault, Kenji Sagae, and David Traum, in Dialogue & Discourse, 2011.
-
Can I finish? Learning when to respond to incremental interpretation results in interactive dialogue David DeVault, Kenji Sagae, and David Traum, in The 10th Annual SIGDIAL Meeting on Discourse and Dialogue (SIGDIAL 2009), London, UK
-
Towards Natural Language Understanding of Partial Speech Recognition Results in Dialogue Systems Kenji Sagae, Gwen Christian, David DeVault and David Traum, in North American Chapter of the Association for Computational Linguistics - Human Language Technologies (NAACL HLT) 2009, Bounder, Colorado