A crowd-sourcing tool for collecting task-oriented spoken dialogues

Access full-text files

Date

2023-04-21

Authors

Salem, Rishi

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Dialogue skills are crucial in everyday life, and machine learning tools that model and analyze data can be essential in helping humans with dialogue-based tasks. However, most dialogue models rely solely on text data, given the lack of spoken dialogue datasets. In this thesis, we build the DARE interface, which is designed to collect spoken dialogues. While it can be used with any task-oriented spoken dialogue, we focus on the example case of collecting dialogue on simple negotiation. This paper details the tools used in the development of the DARE interface, the design choices made, and the initial pilot study of using the system to collect spoken negotiation dialogues.

Description

LCSH Subject Headings

Citation