Dialogue

Abstract:
Dialogue is at the same time the most fundamental and broadly used form of language, as well as the most complex one. And most of all, dialogue is the most natural medium of communication for human beings. It is these aspects of dialogue that make its modelling a research area of its own, and one of greatest interest at that. In many cases of interaction between humans and machines, using dialogue may enables better results with regard to the target each time.In these series of lectures we are going to address the issue of dialogue and dialogue management.In the first lecture we are going to look at some existing dialogue systems and evaluate their performance with regard to the general characteristics that a dialogue system should exhibit. We will consider different approaches to dialogue management and assess their methods of dealing with different dialogue characteristics. In the second lecture we will have a closer look at systems based on finite state automata (FSA). We will consider their possibilities and limitations by exploring the dialogue specifications necessary and their manipulation in FSA. In the case where the system does not currently cater for all desirable dialogue properties, we will weigh the possibility of extending it.In this lecture, we will come to more complex dialogue modeling and will discuss a more powerful systems We will talk about speech and dialogue acts, their relation and the way they are used to facilitate dialogue modeling. We will also address the issue of general system architectures to the extent that they partake of or influence dialogue management.

Table of Contents

Examples of Dialogue Systems
In this Chapter, we give examples in the form of dialogue fragments in the context of two existing dialogue systems: a speaking elevator and a route planning system. This will serve as background when we look at the details of the elevator system later (» FSA-based Dialogue Systems).

Finite States Techniques for Dialogue Processing
A common technique in dialogue processing is the use of a kind of FSA to specify the dialogues that a sysem can handle. The Saarbrücken Speaking Elevator mentioned earlier (» Speaking Elevator: Examples of Required Behaviour) is based on such a FSA. We will now introduce FSA techniques for dialogue processing and then discuss the elevator's underlying FSA.

Speech Acts and Dialogue Management
In this lecture we are going to talk about the use of speech acts in dialogue management. In » Introduction to Speech Acts we will give an introduction to what speech acts are.» Speech Acts in Dialogue Management we will concentrate on how speech acts can facilitate dialogue management and » Example Dialogue Annotation Schemes will briefly consider some examples of existing dialogue act taxonimies.In » TRIPS the TRIPS system architecture will be considered. This is a system based on speech acts. Looking into it will enable an insight into how the use of speech acts can be integrated in dialogue management and taken advantage of.The practical exercise in » Practical exercise. will ask you to annotate a small dialogue excerpt in order to get better acquainted with the use of speech acts and the problems involved.

Practical Session (MiLCA-Summer-School Tübingen, 2003)
We now come to the practical exercises. Please work in small groups.The exercises are about
  1. Testing and Evaluating the Philips train information system.
  2. Getting used to the CLT-tool by extending a small travel-information dialogue by some useful features.
  3. Using the CLT-tool to design a speaking elevator dialogue model. You will then test and evaluate your dialogue model using a Lego-robot and hopefuly enhance it based on your eveluation.


Bibliography
Glossary