Porovnať metódy
Prezrite si vybrané metódy vedľa seba; riadky, ktoré sa líšia, sú zvýraznené.
| Adaptácia otázok a odpovedí pre konkrétnu doménu× | Doménovo-adaptívna klasifikácia založená na modeli BERT× | |
|---|---|---|
| Odbor | Hlboké učenie | Hlboké učenie |
| Rodina | Machine learning | Machine learning |
| Rok vzniku | 2019–2020 | 2019–2020 |
| Tvorca≠ | Multiple (e.g., Garg et al.; Yue et al.) | Gururangan et al. (2020); earlier domain-specific instances include Lee et al. (2020) — BioBERT |
| Typ≠ | Domain adaptation for extractive/generative QA | Domain-adaptive pre-training followed by supervised fine-tuning |
| Pôvodný zdroj≠ | Garg, S., Vu, T., & Moschitti, A. (2020). TANDA: Transfer and Adapt Pre-Trained Transformer Models for Answer Sentence Selection. Proceedings of the AAAI Conference on Artificial Intelligence, 34(5), 7780–7788. DOI ↗ | Gururangan, S., Marasovic, A., Swayamdipta, S., Lo, K., Beltagy, I., Downey, D., & Smith, N. A. (2020). Don't Stop Pretraining: Adapt Language Models to Domains and Tasks. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (ACL 2020), 8342–8360. DOI ↗ |
| Ďalšie názvy | DA-QA, domain-adapted QA, domain-specific question answering, cross-domain question answering | DAPT BERT classification, domain-adaptive pre-training, domain-specific BERT fine-tuning, BERT DAPT |
| Príbuzné | 6 | 6 |
| Zhrnutie≠ | Domain-adaptive Question Answering (DA-QA) adapts a pre-trained language model — typically BERT or RoBERTa — first trained on general QA benchmarks such as SQuAD to answer questions accurately in a new target domain (e.g., biomedical, legal, financial) where labelled data is scarce. Combining domain-adaptive pre-training with task fine-tuning yields substantially stronger performance than direct fine-tuning alone. | Domain-adaptive BERT-based classification extends the standard fine-tuning pipeline by first continuing BERT's masked-language-model pre-training on a large corpus of in-domain unlabeled text, then fine-tuning the adapted model on labeled examples for the target classification task. This two-stage approach closes the vocabulary and distributional gap between BERT's general pre-training corpus and specialized domains such as biomedicine, law, finance, or social-media text. |
| ScholarGateDátová sada ↗ |
|
|