| CRC Colloquium |
| Monday, 11/17/08 16:00 - 18:00 | |
|---|---|
|
Dr. David Schlangen Universität Potsdam Vertretung Professur "Angewandte Computerlinguistik und Texttechnologie", Universität Bielefeld Alignment, Timing, and Incremental Processing In dialogue, it matters not only what you say, but also *when* you say it. Human dialogue participants are normally very good at achieving smooth transitions from one speaker to the other, with very little overlap and few gaps. Dialogue systems, on the other hand, are not so good at this task, and this adds to the perception of them being unnatural and somewhat deficient conversation partners. In this talk I will first relate the achievement of good timing in dialogue to alignment in a more general sense. I will then present some results of our ongoing project on incremental processing in dialogue systems, that is, processing of and reacting to parts of utterances already and not just complete utterances -- a precondition for achieving natural timing of contributions. I will introduce our conceptual framework for modelling incremental processing, and will describe the first prototype we built, a fully incremental dialogue system that can engage in dialogues in a simple domain, number dictation. Because it uses incremental speech recognition and prosodic analysis, the system can give rapid feedback as the user is speaking, with a very short latency of around 200ms. Because it uses incremental speech synthesis and self-monitoring, the system can react to feedback from the user as it is speaking. A comparative evaluation shows that naïve users preferred this system over a non-incremental version, and that it was perceived as more human-like. |
|
| Contact | |
| contact-person: | Gerhard Jäger |
| E-mail: | This e-mail address is being protected from spam bots, you need JavaScript enabled to view it |