Dynamic Label Name Refinement for Few-Shot Dialogue Intent Classification
Gyutae Park, Ingeol Baek, ByeongJeong Kim, Joongbo Shin, Hwanhee Lee
TL;DR
This paper tackles the difficulty of few-shot dialogue intent classification when numerous intents exhibit semantic overlap. It introduces a retrieval-augmented in-context learning framework augmented by dynamic label renaming, where a large language model refines intent labels based on retrieved examples before final classification. Empirical results across DialoGLUE and HINT3 datasets show substantial accuracy gains and reduced label overlap, with gains persisting across model scales and particularly benefiting semantically similar intents. The approach yields more interpretable, semantically coherent labels and demonstrates practical impact for improving robustness in low-resource dialogue systems, albeit with added computational cost.
Abstract
Dialogue intent classification aims to identify the underlying purpose or intent of a user's input in a conversation. Current intent classification systems encounter considerable challenges, primarily due to the vast number of possible intents and the significant semantic overlap among similar intent classes. In this paper, we propose a novel approach to few-shot dialogue intent classification through in-context learning, incorporating dynamic label refinement to address these challenges. Our method retrieves relevant examples for a test input from the training set and leverages a large language model to dynamically refine intent labels based on semantic understanding, ensuring that intents are clearly distinguishable from one another. Experimental results demonstrate that our approach effectively resolves confusion between semantically similar intents, resulting in significantly enhanced performance across multiple datasets compared to baselines. We also show that our method generates more interpretable intent labels, and has a better semantic coherence in capturing underlying user intents compared to baselines.