UniNL: Aligning Representation Learning with Scoring Function for OOD Detection via Unified Neighborhood Learning

TL;DR

This paper introduces UniNL, a unified neighborhood learning framework that aligns representation learning with a KNN-based scoring function to improve out-of-domain intent detection in dialogue systems.

Contribution

The paper proposes a novel unified neighborhood learning framework that combines KNN contrastive learning with a KNN-based scoring function for better OOD detection.

Findings

Effective OOD detection on benchmark datasets

Improved alignment between representations and scoring

Superior performance over previous methods

Abstract

Detecting out-of-domain (OOD) intents from user queries is essential for avoiding wrong operations in task-oriented dialogue systems. The key challenge is how to distinguish in-domain (IND) and OOD intents. Previous methods ignore the alignment between representation learning and scoring function, limiting the OOD detection performance. In this paper, we propose a unified neighborhood learning framework (UniNL) to detect OOD intents. Specifically, we design a K-nearest neighbor contrastive learning (KNCL) objective for representation learning and introduce a KNN-based scoring function for OOD detection. We aim to align representation learning with scoring function. Experiments and analysis on two benchmark datasets show the effectiveness of our method.