Generalized Intent Discovery: Learning from Open World Dialogue System
This addresses the limitation of traditional intent classifiers in practical dialogue systems by enabling discovery of new intents, though it is incremental as it builds on existing classification tasks.
The paper tackles the problem of intent classification in dialogue systems by introducing Generalized Intent Discovery (GID), which aims to extend classifiers to handle both in-domain and out-of-domain intents, and constructs three datasets and proposes frameworks to guide future research.
Traditional intent classification models are based on a pre-defined intent set and only recognize limited in-domain (IND) intent classes. But users may input out-of-domain (OOD) queries in a practical dialogue system. Such OOD queries can provide directions for future improvement. In this paper, we define a new task, Generalized Intent Discovery (GID), which aims to extend an IND intent classifier to an open-world intent set including IND and OOD intents. We hope to simultaneously classify a set of labeled IND intent classes while discovering and recognizing new unlabeled OOD types incrementally. We construct three public datasets for different application scenarios and propose two kinds of frameworks, pipeline-based and end-to-end for future work. Further, we conduct exhaustive experiments and qualitative analysis to comprehend key challenges and provide new guidance for future GID research.