Võrdle meetodeid
Vaata valitud meetodeid kõrvuti; erinevad read on esile tõstetud.
| Kontrastõpe NLP-s× | Enesest juhendav õppimine× | |
|---|---|---|
| Valdkond≠ | Tekstikaeve | Masinõpe |
| Perekond≠ | Process / pipeline | Machine learning |
| Tekkeaasta≠ | 2020–2021 | 2018–2020 |
| Looja≠ | Gao, Yao & Chen (SimCSE, 2021); Khosla et al. (Supervised Contrastive, 2020) | LeCun, Y. and community (formalized ~2018–2020) |
| Tüüp≠ | Self-supervised / supervised representation learning | Representation learning paradigm |
| Algallikas≠ | Gao, T., Yao, X., & Chen, D. (2021). SimCSE: Simple Contrastive Learning of Sentence Embeddings. Proceedings of EMNLP 2021. link ↗ | LeCun, Y. & Misra, I. (2022). Self-supervised learning: The dark matter of intelligence. Meta AI Blog. https://ai.facebook.com/blog/self-supervised-learning-the-dark-matter-of-intelligence/ link ↗ |
| Rööpnimetused | SimCSE, contrastive sentence embeddings, ContrastiveBERT, Karşıtlık Öğrenmesi — NLP (Contrastive Learning) | SSL, self-supervised pre-training, pretext-task learning, unsupervised representation learning |
| Seotud≠ | 4 | 3 |
| Kokkuvõte≠ | Contrastive learning for NLP is a representation-learning technique — popularised by SimCSE (Gao et al., 2021) and Supervised Contrastive Learning (Khosla et al., 2020) — that trains a text encoder by pulling embeddings of similar text pairs together while pushing embeddings of dissimilar pairs apart. The result is a dense, high-quality embedding space that can be learned with no labels at all, or with minimal supervision, making it especially valuable when annotated data are scarce. | Self-supervised learning (SSL) is a machine-learning paradigm that generates its own supervisory signal directly from unlabeled data by defining an auxiliary pretext task — such as predicting masked words, rotating images, or contrasting augmented views — and uses the learned representations as a powerful starting point for downstream tasks with minimal labeled examples. |
| ScholarGateAndmestik ↗ |
|
|