方法对比

	基于RoBERTa的半监督分类	半监督式 Transformer
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2019–2020	2018–2019
提出者	Liu et al. (RoBERTa, 2019); semi-supervised adaptation by the NLP community	Devlin, J. et al. (BERT); broader SSL-Transformer paradigm community
类型	Semi-supervised fine-tuning of a pretrained language model	Semi-supervised deep learning
开创性文献	Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., & Stoyanov, V. (2019). RoBERTa: A Robustly Optimized BERT Pretraining Approach. arXiv preprint arXiv:1907.11692.	Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. Proceedings of NAACL-HLT 2019, 4171–4186.
别名	Semi-supervised RoBERTa, RoBERTa with semi-supervised learning, SSL-RoBERTa classification, RoBERTa pseudo-label classification	semi-supervised transformer model, SSL transformer, transformer with self-supervised pre-training, semi-supervised attention model
相关	6	5
摘要	Semi-supervised RoBERTa-based classification combines a large pretrained RoBERTa language model with both a small labeled dataset and a larger pool of unlabeled text. By generating pseudo-labels or enforcing consistency on unlabeled examples, the method extracts supervisory signal from unannotated data, yielding stronger classifiers when ground-truth annotations are scarce.	Semi-supervised learning with Transformer architectures leverages large quantities of unlabeled data alongside a small labeled set to train powerful sequence models. The dominant pattern — exemplified by BERT — first pre-trains the Transformer on unlabeled data using self-supervised objectives such as masked token prediction, then fine-tunes it on the labeled task. This two-stage approach dramatically reduces the labeled data needed to achieve strong performance.