Watch the Neighbors: A Unified K-Nearest Neighbor Contrastive Learning Framework for OOD Intent Discovery

TL;DR

This paper introduces a unified K-nearest neighbor contrastive learning framework that improves out-of-domain intent discovery in dialogue systems by addressing in-domain overfitting and bridging representation learning with clustering.

Contribution

It proposes a novel KCL and KCC method that jointly enhances in-domain discriminative features and out-of-domain clustering, outperforming existing approaches.

Findings

Significant performance improvements on three benchmark datasets.

Effective mitigation of in-domain overfitting.

Enhanced clustering quality through hard negative mining.

Abstract

Discovering out-of-domain (OOD) intent is important for developing new skills in task-oriented dialogue systems. The key challenges lie in how to transfer prior in-domain (IND) knowledge to OOD clustering, as well as jointly learn OOD representations and cluster assignments. Previous methods suffer from in-domain overfitting problem, and there is a natural gap between representation learning and clustering objectives. In this paper, we propose a unified K-nearest neighbor contrastive learning framework to discover OOD intents. Specifically, for IND pre-training stage, we propose a KCL objective to learn inter-class discriminative features, while maintaining intra-class diversity, which alleviates the in-domain overfitting problem. For OOD clustering stage, we propose a KCC method to form compact clusters by mining true hard negative samples, which bridges the gap between clustering and representation learning. Extensive experiments on three benchmark datasets show that our method achieves substantial improvements over the state-of-the-art methods.