Task2Dial: A Novel Task and Dataset for Commonsense-enhanced Task-based Dialogue Grounded in Documents

• DOI: 10.18653/v1/2022.dialdoc-1.21
• Publication Date: 2022-06-03T01:34:53Z
• Authors: Carl Strathearn; Dimitra Gkatzia
• Keywords: Commonsense knowledge; SemEval; Commonsense reasoning

This paper proposes a novel task on commonsense-enhanced task-based dialogue grounded in documents and describes the Task2Dial dataset, dataset of document-grounded dialogues, where an Information Giver (IG) provides instructions (by consulting document) to Follower (IF), so that latter can successfully complete task. In this unique setting, IF ask clarification questions which may not be underlying document require commonsense knowledge answered. The poses new challenges: (1) its human reference texts show more lexical richness variation than other datasets; (2) generating from set requires paraphrasing as instructional responses might have been modified document; (3) knowledge, since necessarily (4) planning based context, steps need provided order. contains dialogues with average 18.15 number turns 19.79 tokens per turn, compared 12.94 12 respectively existing datasets. As such, learning promises natural, varied less template-like system utterances.