How People Talk To Computers Robots And Other Artificial Communication Partners

-- KerstinFischer -- 13 Jan 2006

Workshop April 21-23, 2006

Organisers:

Kerstin Fischer, University of Bremen (SFB/TR8), kerstinf@uni-bremen.de

Anton Batliner, University of Erlangen, batliner@informatik.uni-erlangen.de

Conference Site:

Hanse Wissenschaftskolleg (HWK), Delmenhorst, Germany

Aims and Motivation

The workshop aims at bringing together researchers from linguistics, psychology, and computer science for the joint development of a catalogue of research desiderata, the formulation of concrete research questions and the definition of different research positions focussing on how people talk to artificial communication partners.

There is a growing body of research on the design of artificial communication partners, such as dialogue systems, robots, ECAs and so on, and thus conversational interfaces are becoming more and more sophisticated. However, so far such systems do not meet the expectations of ordinary users. One reason that prevents systems being perceived as useful and fully functional may be that there is still very little known about the ways human users actually address such conversational interfaces. How naïve speakers really interact with such systems and the language that they use to do so cannot be deduced by intuition; effective language of this kind is simply not available to introspection. Moreover, empirical linguistic and psychological studies of the ways people talk to artificial communication partners so far have yielded only very particular, corpus-, domain- or situation-specific results. What is needed, therefore, is to bring together results from various different scenarios in order to achieve a more general picture of the determining factors of different ways of talking to artificial agents, such as dialogue systems, ECAs, robots and the like, aiming at a model that promises both reusability of results achieved in different human-computer situations and predictability with respect to behaviours that may be expected of new human-computer interfaces.

In this area, researchers have only just begun to explore the role of central pragmatic mechanisms, such as recipient design, alignment, and interactional strategies, such as feedback, in communication with artificial communication partners. Here, psychological and linguistic studies will certainly reveal dialogue strategies that support dialogue system design.

Furthermore, system design may profit from the identification of different user groups. For instance, a compromise between fully speaker-independent systems (word-error rate too high) and fully speaker-dependent systems (low word-error rate but confined to one speaker) might be to establish different types of speakers according to their linguistic behaviour and to establish different recognizers especially tailored for these different groups.

Finally, the fact that speakers align to their communication partners should be exploited by shaping the linguistic behaviour of speakers in a way which is most useful for the system to understand. This involves issues of initiative, feedback, and dialogue act modelling.

The topic of the workshop is thus highly relevant from theoretical and practical perspectives. It addresses one of the most urgent deadlocks in current dialogue system design and evokes an interdisciplinary perspective on the problem, providing theoretically interesting and practical ways out of current dilemmas. The workshop will furthermore serve to connect scientists from different disciplines.

We want to bring together researchers particularly interested in the following questions:

• Which different types of linguistic behaviours (phonetic, prosodic, syntactic, lexical, conversational) can be found in communication with artificial communication partners?

• Do these types of behaviours cluster in particular ways such that some behaviours tend to co-occur with others so that different types of users become apparent?

• Are there particular linguistic means to identify different types of users (unobtrusively and online)?

• Which aspects of the design condition which kinds of behaviours?

• Which roles do recipient design, alignment, and feedback play in the communication with artificial communication partners?

• Which kinds of problems in dialogue modelling and automatic speech processing can be prevented by modelling different kinds of linguistic behaviours and different types of users?

Projected outcomes of the workshop include, besides the interdisciplinary networking of the participants, the formulation of concrete research questions, an increased awareness of the problems of human-computer communication, the augmentation of available research and evaluation methodologies which are inspired by the interdisciplinary discussion, and concrete methods for the realisation of findings in adaptive system design. Suitable means of publication are under discussion.

Invited speakers:

• Holly Branigan (Edinburgh)
• Jamie Pearson (Edinburgh)
• Ludwig Hitzenberger (Regensburg)
• Fiorella de Rosis (Bari)

To apply for participation in the workshop, please send a one-page summary of how your work relates to the topics outlined above before February 15th, 2006. Since workshop contributions will consist of one or more short impulse presentations on the workshop topics, please indicate what you would like to present and how much time you would like to have for each presentation. This abstract will also be made available to the other participants at the beginning of the workshop. You’ll be informed about acceptance or rejection by February 25th.

Costs:

You will only have to take care of your travelling costs. All other costs, including two nights in a hotel near to the conference site, will be taken care of by the SFB/TR8 ‘Spatial Cognition’ and the Hanse Wissenschaftskolleg. The number of participants may therefore not exceed 25.

Contact:

Kerstin Fischer kerstinf@uni-bremen.de

Anton Batliner batliner@informatik.uni-erlangen.de