MAS2S: Asking Clarification Questions for Information Seeking in Task-Oriented Dialogues

Forum topic · 小凯 · 2026-07-05

Summary

This paper, 'Towards Asking Clarification Questions for Information Seeking on Task-Oriented Dialogues' (Feng, Rahmani, Lipani, Yilmaz; arXiv:2305.13690, May 2023), addresses a key limitation of task-oriented dialogue systems: users often cannot fully describe complex information needs, and systems may hold ambiguous or missing information about users. The authors propose MAS2S (Multi-Attention Seq2Seq Network), a model that generates clarification questions to resolve ambiguity about both the user's information needs and user profile before providing task-oriented information seeking results. They also extend an existing dataset to create CLARIT, a publicly released corpus of approximately 100k task-oriented information seeking dialogues (code and data at github.com/sweetalyssum/clarit). Experiments on CLARIT show MAS2S outperforms baselines on both clarification question generation and answer prediction. The work is relevant to conversational search, clarifying question generation, and personalization in dialogue-based information access.

Towards Asking Clarification Questions for Information Seeking on Task-Oriented Dialogues

Authors: Yue Feng, Hossein A. Rahmani, Aldo Lipani, Emine Yilmaz Published: 2023-05-23 Source: arXiv:2305.13690 Category: Search Assistance

Abstract (original)

> Task-oriented dialogue systems aim at providing users with task-specific services. Users of such systems often do not know all the information about the task they are trying to accomplish, requiring them to seek information about the task. To provide accurate and personalized task-oriented information seeking results, task-oriented dialogue systems need to address two potential issues: 1) users' inability to describe their complex information needs in their requests; and 2) ambiguous/missing information the system has about the users. In this paper, we propose a new Multi-Attention Seq2Seq Network, named MAS2S, which can ask questions to clarify the user's information needs and the user's profile in task-oriented information seeking. We also extend an existing dataset for task-oriented information seeking, leading to the CLARIT dataset which contains about 100k task-oriented information seeking dialogues that are made publicly available. Experimental results on CLARIT show that MAS2S outperforms baselines on both clarification question generation and answer prediction.

Key Points

Problem: Task-oriented dialogue users may struggle to fully express complex information needs, and the system may have ambiguous or missing information about the user — both degrade personalized information seeking.
Proposed model: MAS2S, a Multi-Attention Seq2Seq Network that asks clarification questions about the user's information needs and user profile before answering.
Dataset: CLARIT, an extension of an existing task-oriented information seeking dataset, containing ~100k publicly available dialogues.
Results: MAS2S outperforms baselines on both clarification question generation and answer prediction on CLARIT.
Resources: Dataset and code are publicly available at https://github.com/sweetalyssum/clarit.

Context and Significance

Clarification question generation is an important component of conversational information seeking. This work tackles it in the task-oriented setting, where personalization (user profile) and task completion both matter. It complements related research on ambiguity handling in open-domain QA and open-domain information-seeking conversations.

Related Work

Asking Clarification Questions to Handle Ambiguity in Open-Domain QA (arXiv:2305.13808)
Asking Clarifying Questions in Open-Domain Information-Seeking Conversations (DOI: 10.1145/3331184.3331265)

Limitations and Outlook

As with many dialogue-model papers, evaluation is on a benchmark dataset rather than live users, so real-world satisfaction, latency, and safety considerations remain open questions for industrial deployment.