Comparing dialogue strategies for learning grounded language from human tutors

Yu, Yanchao; Lemon, Oliver; Eshghi, Arash

Abstract

We address the problem of interactively learning perceptually grounded word meanings in a multimodal dialogue system. Human tutors can correct, question, and confirm the statements of a dialogue agent which is trying to interactively learn the meanings of perceptual words, e.g.\ colours and shapes.
We show that different learner and tutor dialogue strategies lead to different learning rates, accuracy of learned meanings, and effort/costs for human tutors. For example, we show that a learner which can handle corrections in dialogue, and its own uncertainty about what it sees, can learn meanings that are as accurate as a fully-supervised learner, but with less cost/effort to the human tutor.

Citation

Yu, Y., Lemon, O., & Eshghi, A. (2016, July). Comparing dialogue strategies for learning grounded language from human tutors. Presented at 20th Workshop Series on the Semantics and Pragmatics of Dialogue 201, New Brunswick, US

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 20th Workshop Series on the Semantics and Pragmatics of Dialogue 201
• Start Date: Jul 16, 2016
• End Date: Jul 18, 2016
• Publication Date: 2016
• Deposit Date: Jun 29, 2023
• Pages: 44-54
• Book Title: Proceedings of the 20th Workshop on the Semantics and Pragmatics of Dialogue