方法对比

	弱监督 RoBERTa 分类	弱监督BERT分类
领域	深度学习	深度学习
方法族	Machine learning	Machine learning
起源年份	2019–2020	2017–2020
提出者	Liu et al. (RoBERTa, 2019); weak supervision paradigm: Ratner et al. (2016–2020)	Multiple (Ratner et al. for weak supervision framework; Meng et al. for BERT integration)
类型	Pretrained transformer classifier with weak supervision	Weakly supervised fine-tuning of pre-trained language model
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:1907.11692.	Meng, Y., Zhang, Y., Huang, J., Xiong, C., Ji, H., Zhang, C., & Han, J. (2020). Text Classification Using Label Names Only: A Language Model Self-Training Approach. Proceedings of EMNLP 2020, 9006–9017.
别名	WS-RoBERTa, RoBERTa with weak supervision, weakly supervised transformer classification, noisy-label RoBERTa classifier	WS-BERT, BERT with weak supervision, label-efficient BERT classification, noisy-label BERT fine-tuning
相关	5	6
摘要	Weakly supervised RoBERTa-based classification combines the RoBERTa pretrained transformer with weak supervision — programmatic or heuristic labeling sources — to train powerful text classifiers without requiring a fully hand-labeled dataset. Labeling functions, distant supervision, or crowd-sourced signals generate noisy labels that are aggregated and used to fine-tune RoBERTa for downstream classification tasks.	Weakly supervised BERT-based classification adapts BERT to text classification tasks when only noisy, heuristic, or programmatically generated labels are available instead of clean human annotations. It combines weak supervision frameworks — such as labeling functions and data programming — with BERT's pre-trained language representations to achieve robust classification without expensive hand-labeling.