• research_dialogue_lab

research_dialogue_lab

Documentation for GoDiS

Using the information state update approach as implemented in TrindiKit, I have developed
a theory of dialogue management (“Issue-based Dialogue Management” or IBDM)
and implemented it in a dialogue system, GoDiS (Larsson, 2002). GoDiS provides general
solutions to several pervasive problems and issues in dialogue, such as: establishing shared
understanding, user initiative, dealing with multiple conversational threads, information
sharing between threads, revising information, multimodal interaction, multilinguality,
plug-and-play of dialogue system applications, etc.). Providing general solutions to general
problems in dialogue means they are solved once and for all, and this in turn means that
new applications which use these general solutions can be developed quicker and cheaper.

There has never been a proper manual for GoDiS; the closest thing to a manual is Larsson
(2002) but this is unsuitable for at least two reasons: (1) It does not provide a comprehensive
overview of the system (nor is it intended to), and (2) it is not up-to-date. A proper manual
should allow a beginner to understand the basics of the underlying theoretical approach,
explain the implementation, and provide detailed information about how to implement
new GoDiS applications.

It is likely that the GoDiS system itself will be slightly modified in the course of writing
a manual, as rough solutions (“hacks”) must be replaced to enable a comprehensive
descriptions.

Completion of the GoDiS manual is expected to take 1 month.

Project duration: 2008-03-01 - 2008-12-31

 

Dialogue Systems and Type Theory

To facilitate rapid development of dialogue systems by specifying them as type theoretical grammars (CLT PostDoc project)

Duration: Jan–Dec 2008

Participant: Peter Ljunglöf, CLT PostDoc researcher

Abstract:

Ranta and Cooper (2004) describe how a dialogue system can be implemented in a syntax editor based on type theory, originally developed for editing mathematical proofs. However, only information-seeking dialogues are supported, and the backbone is a fairly simple form-based dialogue system.

In this project I want to build upon their ideas to implement features from ISU-based dialogue systems, to extend the range of dialogue type in include, e.g., action-oriented dialogue. This will be done by specifying information states, update rules, dialogue plans, semantics and syntax in the same type-theoretical framework.

References:

Aarne Ranta and Robin Cooper. Dialogue systems as proof editors. Journal of Logic, Language and Information, 13(2):225–240, April 2004.

X
Loading