Modelling utterance generation in conversational dialogue systems
The GenDial project spans research in several disciplines including computer science, human-human and human-machine spoken
interaction and language technology. Our long term research goal is to develop a new class of conversational spoken dialog systems
which to a large extent follows the principles of human-human interaction. A natural part of human conversation is to choose what
we say, when we say it and how we say it depending on our intentions, our conversational partners and the current state of the
dialogue. For machines to be perceived as natural conversational partners the output needs to be coherent with such an adaptive
behaviour. This proposal addresses a challenge that so far to a large extend has been neglected - the utterance generation process.
We will focus on how this process could be made context dependent and how such a generation model should be integrated in a
spoken dialog system. The project plan includes: data analysis on human-human and human-machine interaction; development of
speech generation models; integration of the models in our existing multimodal dialog system platform; evaluation of the models in
isolation and as part of a conversational dialog system. The proposed research on utterance generation will strengthen the
knowledge in an area that so far has attracted limited research effort despite fact that relevant and well formed system output is of
great importance for the perception and acceptability of spoken dialogue systems.
Group: Speech Communication and Technology
Rolf Carlson (Project leader)
Duration: 2008 - 2010
Human interaction as a model for spoken dialogue system behaviour. Doctoral dissertation. [abstract] [pdf] (2010).
Proceedings of SigDial. Tokyo, Japan. [pdf] (2010). Middleware for Incremental Processing in Conversational Agents. In
Proceedings of SIGdial (pp. 1-8). Tokyo, Japan. (*) [abstract] [pdf] (2010). Towards Incremental Speech Generation in Dialogue Systems. In
(*) Best Paper Award at SIGdial 2010
Proceedings of the 12th Conference of the European Chapter of the Association for Computational Linguistics (EACL-09). Athens, Greece. [abstract] [pdf] (2009). A general, abstract model of incremental dialogue processing. In
Proceedings of SigDial 2009. London, UK. [abstract] [pdf] (2009). Attention and interaction control in a human-human-computer dialogue setting. In
Proceedings of the 12th Conference of the European Chapter of the Association for Computational Linguistics (EACL-09). Athens, Greece. [abstract] [pdf] (2009). Incremental dialogue processing in a micro-domain. In
Perception in Multimodal Dialogue Systems - Proceedings of the 4th IEEE Tutorial and Research Workshop on Perception and Interactive Technologies for Speech-Based Systems, PIT 2008, Kloster Irsee, Germany, June 16-18, 2008. (pp. 252-255). Berlin/Heidelberg: Springer. [abstract] [pdf] (2008). Human-likeness in utterance generation: effects of variability. In