Human-like informative conversations: Better acknowledgements using conditional mutual information

TL;DR

This paper introduces a new method called Fused-PCMI that enhances dialogue agents' ability to acknowledge prior turns by leveraging pointwise conditional mutual information, resulting in more human-like and contextually specific responses.

Contribution

The paper proposes Fused-PCMI, a novel approach that improves response specificity and acknowledgment in dialogue systems by applying linguistic principles and mutual information measures.

Findings

Fused-PCMI outperforms Max-PMI in human preference tests 60% of the time.

Higher $ ext{pcmi}_h$ correlates with better acknowledgment in responses 74% of the time.

Linguistic principles can effectively guide improvements in dialogue agent behavior.

Abstract

This work aims to build a dialogue agent that can weave new factual content into conversations as naturally as humans. We draw insights from linguistic principles of conversational analysis and annotate human-human conversations from the Switchboard Dialog Act Corpus to examine humans strategies for acknowledgement, transition, detail selection and presentation. When current chatbots (explicitly provided with new factual content) introduce facts into a conversation, their generated responses do not acknowledge the prior turns. This is because models trained with two contexts - new factual content and conversational history - generate responses that are non-specific w.r.t. one of the contexts, typically the conversational history. We show that specificity w.r.t. conversational history is better captured by Pointwise Conditional Mutual Information ($\text{pcmi}_h$) than by the established use of Pointwise Mutual Information ($\text{pmi}$). Our proposed method, Fused-PCMI, trades off $\text{pmi}$ for $\text{pcmi}_h$ and is preferred by humans for overall quality over the Max-PMI baseline 60% of the time. Human evaluators also judge responses with higher $\text{pcmi}_h$ better at acknowledgement 74% of the time. The results demonstrate that systems mimicking human conversational traits (in this case acknowledgement) improve overall quality and more broadly illustrate the utility of linguistic principles in improving dialogue agents.