Proactive Dialogue Model with Intent Prediction

TL;DR

This paper introduces a lightweight intent-transition prior using a Temporal Bayesian Network to make dialogue models more proactive and efficient by anticipating user intents, demonstrated on MultiWOZ 2.2 data.

Contribution

It presents a novel intent-transition prior integrated into dialogue systems, improving proactivity and efficiency without altering the core language model.

Findings

T-BN achieves Recall@5 = 0.787 and MRR = 0.576 on intent prediction.

BN-guided generation increases Coverage AUC from 0.742 to 0.856.

Reduces turns to reach 75% intent coverage from 3.95 to 2.73.

Abstract

Dialogue models are inherently reactive, responding to the current user turn without anticipating upcoming intents, which leads to redundant interactions in multi-intent settings. We address this limitation by introducing a lightweight intent-transition prior derived from dialogue data and injected into the system prompt at inference time. We instantiate this prior using a Temporal Bayesian Network (T-BN) trained on per-turn intent annotations in MultiWOZ 2.2. The T-BN achieves Recall@5 = 0.787 and MRR = 0.576 on 1,071 held-out USER-turn pairs. In a ground-truth replay over 200 dialogues, BN-guided generation improves Coverage AUC from 0.742 to 0.856 and reduces the number of turns required to reach 75% intent coverage from 3.95 to 2.73. These results show that lightweight intent-transition guidance enables more proactive and efficient dialogue behavior without modifying the underlying language model.