|
||||||||||||||||||||||||||
Invited Talks
| Parameters of Dialog analysis | |
| Jens Allwood | Göteborg U |
| The notion of "dialog" as well as the related notions of "dialog analysis" and "dialog theory" have become increasingly popular in recent years. The popularity of the notion of "dialog" includes several disciplines such as philosophy, linguistics, computer science, psychology, sociology, anthropology, literary studies etc. The goal of this paper is to give an overview of some possible parameters of dialog analysis. These parameters can then be used to locate different approaches to dialog in relation to each other. The parameters will first be brielfly characterized and then some of them will be commented on more in depth. | |
| Uptake and its role in conversation | |
| Herbert H. Clark | Stanford U |
| Illocutionary acts have been viewed in two main ways. In the first, which originated with Searle, they are acts that speakers perform autonomously. In the second, proposed by Austin, they are acts performed by speakers in coordination with their addressees. In my terminology, they are participatory acts, which are one individual's parts of joint acts. By now Searle's view is standard, and Austin's is almost forgotten. The essential difference between them is this: In Austin's view the content of an illocutionary act is determined in part by the addressee's uptake, whereas in Searle's view it is not. I review a range of phenomena from spontaneous conversation as evidence for Austin's view and against Searle's. What finally counts, I argue, is not what speakers mean, but what they are jointly taken to mean. | |
| On Safe Updates by Means of Supported Assertions | |
| Paul Dekker | U Amsterdam |
| With this talk I look upon classical systems of update semantics from the wider perspective of information exchange, and present independent compositional statements of the content of, update with, and support for the expressions of a first order system. It is shown that a proper update with the contents of supported utterances is safe, in the sense that it does not corrupt the information distributed over the interlocutors. Our pragmatic outlook on update and support is furthermore shown to be productive, in that it suggests a plausible analysis of functional dependencies and of certain cases of what has been called quantificational and modal subordination. | |
| Trying to Understand Misunderstanding: How Robust Can Spoken Natural Language Dialogue Systems Be? | |
| Ronnie Smith | East Carolina University |
| A ubiquitous problem with AI systems is the difficulty in "knowing that you don't know." Spoken natural language dialog systems are not exempt from this dilemma. This talk will overview several different studies that have been undertaken in order to make dialog systems behave more robustly in the presence of miscommunication. The studies use data collected via experimental interaction with the Circuit Fix-It Shop, a dialog system originally constructed to test the validity of a model for integrated dialog processing that is continuing to be used as a testbed for evaluating the effectiveness of techniques for the prevention, detection, and repair of miscommunication in human-computer dialog. | |
| Overview of TRINDI Project Results | |
| The TRINDI Consortium | |