Uporedite metode

	Полу-надгледано одговарање на питања	Полу-надгледани Трансформер
Oblast	Duboko učenje	Duboko učenje
Porodica	Machine learning	Machine learning
Godina nastanka	2006–2020	2018–2019
Tvorac	Multiple (Chapelle et al.; Zhu; Clark et al. for NLP applications)	Devlin, J. et al. (BERT); broader SSL-Transformer paradigm community
Tip	Semi-supervised learning applied to extractive/generative QA	Semi-supervised deep learning
Temeljni izvor	Clark, K., Luong, M.-T., Le, Q. V., & Manning, C. D. (2020). ELECTRA: Pre-training Text Encoders as Discriminators Rather Than Generators. In Proceedings of ICLR 2020.	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.
Drugi nazivi	Semi-supervised QA, Self-training for QA, Pseudo-labeled Question Answering, SSL-QA	semi-supervised transformer model, SSL transformer, transformer with self-supervised pre-training, semi-supervised attention model
Srodne	6	5
Sažetak	Semi-supervised question answering (QA) trains a model on a small labeled set of question-answer pairs, then generates pseudo-labels on a large unlabeled corpus and retrains iteratively. This self-training loop dramatically increases effective training data without the cost of full manual annotation, achieving strong performance on reading comprehension, open-domain QA, and machine reading tasks.	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.