InSCIt : Information-Seeking Conversations with Mixed-Initiative Interactions

• Published in: Transactions of the Association for Computational Linguistics Vol. 11; pp. 453 - 468
• Main Authors: Wu, Zeqiu, Parish, Ryu, Cheng, Hao, Min, Sewon, Ammanabrolu, Prithviraj, Ostendorf, Mari, Hajishirzi, Hannaneh
• Format: Journal Article
• Language: English
• Published: One Broadway, 12th Floor, Cambridge, Massachusetts 02142, USA MIT Press 18.05.2023; MIT Press Journals, The; The MIT Press
• Subjects: Computational linguistics; Conversation; Datasets; Human-computer interaction; Identification; Knowledge; Queries; Questions
• ISSN: 2307-387X; 2307-387X
• DOI: 10.1162/tacl_a_00559

In an information-seeking conversation, a user may ask questions that are under-specified or unanswerable. An ideal agent would interact by initiating different response types according to the available knowledge sources. However, most current studies either fail to or artificially incorporate such agent-side initiative. This work presents , a dataset for formation- eeking onversations with mixed-initiative n eractions. It contains 4.7K user-agent turns from 805 human-human conversations where the agent searches over Wikipedia and either directly answers, asks for clarification, or provides relevant information to address user queries. The data supports two subtasks, evidence passage identification and response generation, as well as a human evaluation protocol to assess model performance. We report results of two systems based on state-of-the-art models of conversational knowledge identification and open-domain question answering. Both systems significantly underperform humans, suggesting ample room for improvement in future studies.