Data Collection for Interactive Learning through the Dialog

TL;DR

This paper introduces a new dataset of 1900 natural dialogs designed to facilitate research on interactive learning in dialog systems, addressing the challenge of fact sparsity by enabling systems to learn from user interactions.

Contribution

The paper provides a novel dataset for testing interactive learning in dialog systems, supporting the development of systems that can learn facts dynamically during conversations.

Findings

Dataset enables simulation of interactive fact learning.

Supports testing of dialog systems' ability to acquire new knowledge.

Addresses fact sparsity in open domain dialog systems.

Abstract

This paper presents a dataset collected from natural dialogs which enables to test the ability of dialog systems to learn new facts from user utterances throughout the dialog. This interactive learning will help with one of the most prevailing problems of open domain dialog system, which is the sparsity of facts a dialog system can reason about. The proposed dataset, consisting of 1900 collected dialogs, allows simulation of an interactive gaining of denotations and questions explanations from users which can be used for the interactive learning.