Salience of questions predicts answers less well in conversations
Do we need to answer that question? Salience and Answerability of Potential Questions in Naturalistic Dialogue
Computation and Language
Summary
Sometimes when people talk, they ask questions that guide the conversation, but not all questions get answered. The authors studied thousands of questions from real conversations to see if more noticeable questions are more likely to be answered. They found that while more noticeable questions do tend to get answers, this connection is weak in everyday back-and-forth talks compared to written text. They also noticed that in more organized conversations, people tend to agree more on which questions matter.
What this means in practice
- •For dialogue system developers: Improve AI dialogue models by refining question prediction based on salience in conversational settings.
- •For call center managers: Analyze customer-agent interactions to better identify which questions are most likely to be resolved during calls.
Tested on one dataset.
Authors
Amandine Decker, Maxime Amblard, Ellen Breitholtz
Abstract
We empirically investigate Question Under Discussion based modelling in naturalistic dialogue by studying whether the salience of generated potential questions predicts their subsequent resolution. Building on Wu et al. (2024), we construct a dataset of 7,124 questions automatically generated from utterances and preceding context from the British National Corpus, and annotated for salience and answerability. We find a robust but low positive correlation between salience and answerability in dialogue, indicating that more salient questions are more likely to be addressed. However, this effect is markedly weaker than in monologic text, suggesting that conversational structure is less predictable. We further observe that structured interactions exhibit stronger alignment between annotators than less organised dialogues.