Pursuing and demonstrating understanding in dialogue

David Devault, Matthew Stone

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

The appeal of natural language dialogue as an interface modality is its ability to support open-ended mixed-initiative interaction. Many systems offer rich and extensive capabilities, but must support novice or infrequent users. It is unreasonable to expect untrained users to know the actions they need in advance, or to be able to specify their goals using a regimented scheme of commands or menu options. Dialogue allows the user to talk through their needs with the system and arrive collaboratively at a feasible solution. Dialogue, in short, becomes more useful to users as the interaction becomes more potentially problematic. However, the flexibility of dialogue comes at a cost in system engineering. We cannot expect the user's model of the task and domain to align with the system's. Consequently, the system cannot count on a fixed schema to enable it to understand the user. It must be prepared for incorrect or incomplete analyses of users' utterances, and must be able to put together users' needs across extended interactions. Conversely, the system must be prepared for users that misunderstand it, or fail to understand it. This chapter provides an overview of the concepts, models, and research challenges involved in this process of pursuing and demonstrating understanding in dialogue. We start in Section 3.2 from analyses of human-human conversation. People are no different from systems: they, too, face potentially problematic interactions involving misunderstandings. In response, they avail themselves of a wide range of discourse moves and interactive strategies, suggesting that they approach communication itself as a collaborative process wherein all parties establish agreement, to their mutual satisfaction, on the distinctions that matter for their discussion and on the expressions through which to identify those distinctions. In the literature, this process is often described as grounding communication, or identifying contributions well enough so that they become part of the common ground of the conversation (Clark and Marshall, 1981; Clark and Schaefer, 1989; Clark and Wilkes-Gibbs, 1990; Clark, 1996).

Original languageEnglish (US)
Title of host publicationNatural Language Generation in Interactive Systems
PublisherCambridge University Press
Pages34-62
Number of pages29
Volume9781107010024
ISBN (Electronic)9780511844492
ISBN (Print)9781107010024
DOIs
StatePublished - Jan 1 2013

All Science Journal Classification (ASJC) codes

  • Engineering(all)

Fingerprint

Dive into the research topics of 'Pursuing and demonstrating understanding in dialogue'. Together they form a unique fingerprint.

Cite this