Zero-Shot Text Classification with Self-Training
This addresses the problem of unreliable zero-shot classification for users needing efficient adaptation without domain expertise, though it is incremental as it builds on existing self-training techniques.
The paper tackles the instability and performance issues of zero-shot text classification by proposing a plug-and-play self-training method that fine-tunes the classifier on its most confident predictions, leading to significant performance gains across various tasks.
Recent advances in large pretrained language models have increased attention to zero-shot text classification. In particular, models finetuned on natural language inference datasets have been widely adopted as zero-shot classifiers due to their promising results and off-the-shelf availability. However, the fact that such models are unfamiliar with the target task can lead to instability and performance issues. We propose a plug-and-play method to bridge this gap using a simple self-training approach, requiring only the class names along with an unlabeled dataset, and without the need for domain expertise or trial and error. We show that fine-tuning the zero-shot classifier on its most confident predictions leads to significant performance gains across a wide range of text classification tasks, presumably since self-training adapts the zero-shot model to the task at hand.