Important Dates | Submission Procedure | Organizing Committee |
Workshop program | Further information |

Description of the Workshop

The second workshop will continue along the same lines. The topic of dialogue systems and the type and management of their knowledge is becoming increasingly important not only for applications utilizing spoken natural language interaction but also for applications that allow interaction with several modalities and especially in real environments where several other information sources must be coordinated.

The workshop will especially concentrate on multimodal interfaces, which in the past five or ten years have seen a growing interest. Providing systems with multimodal behaviour is seen as a way of allowing for more natural human-machine communication, but it also has applications in new areas like robotics, text mining, and interaction for disabled and mobile communication.

Multimodal dialogue systems can facilitate communication since the interpretation of the communicative acts can be based on input from different modality devices and errors in one channel can be compensated by the information coming from another channel. On the other hand, the use of additional modalities is also likely to introduce ambiguities and uncertainty, and thus multimodality puts heavy demand on flexible interaction modelling. The need to merge input from different sources requires extensive reasoning capabilities both in understanding and responding to partial, erroneous, multi-channel input. The workshop focuses especially on these aspects: the coordination and integration of multimodal inputs and the ways in which multimodal inputs reinforce and complement each other, and the role of dialogue in multimodal interaction. An important issue is also the representation of multimodal inputs, and it can be mentioned that last year the W3C initiative has issued a document on Voice Markup Languages including requirements for standards concerning multimodal input and output (http://www.w3.org/TR/voice-dialog-reqs).

The following topics are of primary interest to the workshop:

• standards and generic methodologies for modality integration. Such methodologies should enable systems to make the most of the redundancy introduced by multimodality and make multimodal systems more easily scalable and portable across domains.
• robust recognition and coordination of input from various modalities, including text, speech, and graphics, and also from such "new" devices as data gloves, touch screens and eye-gaze trackers.
• interpretation of multimodal input in relation to a dialogue model and the user model and the task at hand. Especially this includes parsing and coordination of various types of feedback.
• studies on modelling the attentional and focus structure
• natural and coherent generation of multimodal output: the choice between single and multiple modalities. How to best present different information pieces and how to coordinate and time various modalities so as to form a coherent response.
• new techniques developed for the coordination and control of multimodal inputs. Especially, various supervised and unsupervised learning algorithms have already successfully been applied in AI, e.g. in robotic motor-control problems, and the application of these techniques in multimodal environment has just started.
• how to combine modalities to facilitate efficient interaction, especially empirical investigations on such interaction.

Who should attend

It should be of interest also for anyone studying dialogue and multimodal interfaces and how to coordinate different information sources. This involves theoretical as well as practical research, e.g. empirical evaluations of usability, formalization of dialogue phenomena and development of intelligent interfaces for various applictions, not excluding such areas as robotics.

The workshop will encourage the participation of both system builders and theoretically oriented researchers, thus creating a forum for discussion across vocational and disciplinary borders. While taking practical applications and implemented dialogue systems as our point of departure, we emphasize the potential contributions of theoretical and empirical research: applications are the best testbeds for evaluating the usefulness and originality of theories and ideas. There is a certain lack of communication between theorists and system builders, and the workshop aims at creating an atmosphere that allows for a productive interaction between these two groups.

Workshop format

Each paper will be given ample time for discussion, more than what is customary at a conference. As said above, we encourage contributions of a critical or comparative nature that provide fuel for discussion. We also invite people to share their experiences of implementing and coordinating knowledge modules in their dialogue systems, and integrating dialogue components to other applications.

Workshop program

Submissions

Electronic submissions (in postscript format) should be sent to Jan Alexandersson at: jan.alexandersson@dfki.de
Alternatively, three hardcopies can be mailed to the corresponding address below.

Jan Alexandersson
German Research Center for Artificial Intelligence, DFKI GmbH
Stuhlsatzenhausweg 3
66 123 Saarbrücken
Germany

The accepted papers will be published in the workshop proceedings which will be distributed to each participant.

Important dates

First call for papers: 1 December, 2000

Second call for papers: 15 January, 2001

DEADLINES FOR SUBMISSIONS HAVE BEEN EXTENDED. NEW DEADLINES:

Deadline for submission of papers: 9 March, 2001

Notification of acceptance: 3 April, 2001

Camera-ready paper for the workshop notes: 24 April, 2001

Workshop: 5 August, 2001

Program Committee

Organizing Committee