Effectiveness of Pre-training for Few-shot Intent Classification
Haode Zhang, Yuwei Zhang, Li-Ming Zhan, Jiaxin Chen, Guangyuan Shi, Albert Y. S. Lam, Xiao-Ming Wu
TL;DR
The paper tackles few-shot intent classification for task-oriented dialogues by proposing IntentBERT, a model created by fine-tuning BERT on roughly 1,000 publicly labeled utterances and capable of cross-domain generalization without target-domain fine-tuning. It further introduces a joint pre-training scheme that leverages unlabeled data from the target domain to boost representations, yielding substantial gains over existing pre-training methods. The approach emphasizes data efficiency and practical applicability, showing that most intent detection tasks share a common underlying structure that can be learned from small, task-relevant data. Overall, IntentBERT offers a simple, effective, and adaptable pathway for rapid deployment of robust few-shot intent classifiers across diverse domains.
Abstract
This paper investigates the effectiveness of pre-training for few-shot intent classification. While existing paradigms commonly further pre-train language models such as BERT on a vast amount of unlabeled corpus, we find it highly effective and efficient to simply fine-tune BERT with a small set of labeled utterances from public datasets. Specifically, fine-tuning BERT with roughly 1,000 labeled data yields a pre-trained model -- IntentBERT, which can easily surpass the performance of existing pre-trained models for few-shot intent classification on novel domains with very different semantics. The high effectiveness of IntentBERT confirms the feasibility and practicality of few-shot intent detection, and its high generalization ability across different domains suggests that intent classification tasks may share a similar underlying structure, which can be efficiently learned from a small set of labeled data. The source code can be found at https://github.com/hdzhang-code/IntentBERT.