The WHoW Framework: An Automatic Approach to Cross-Scenario Conversational Moderation Analysis through Why, How, and Who

ACL ARR 2024 August Submission466 Authors

Abstract: This study proposes the WHoW framework, an automatic approach for analyzing the facilitation strategies of conversational moderation across different scenarios. The framework breaks down moderation decision-making into three key components: motives, dialogue acts, and target speaker. Using this framework, we annotated 5,657 moderation sentences with human input and 15,494 sentences using GPT-4o across 196 episodes from two settings: the Intelligent Squared TV Debate and the RoundTable Radio Panel Discussion. Comparative analysis of these settings demonstrates the framework's cross-domain generalisability, revealing distinct moderation strategies: debate moderators emphasize coordination and facilitate interaction through questions and instructions, while panel discussion moderators prioritize information provision and actively participate in discussions, but are less involved in fostering inter-speaker interactions. This framework shows potential as a tool for exploring and comparing a broader range of moderation scenarios and could be expanded for the development of moderator agents.

Paper Type: Long

Research Area: Discourse and Pragmatics

Research Area Keywords: conversation, communication, human behavior analysis, NLP tools for social analysis,

Contribution Types: Data resources, Data analysis

Languages Studied: English

Submission Number: 466