| license: eupl-1.1 | |
| SeTABERTa is a new multilingual language model pretained from scratch using various Open Access text repositories: EU legislation, research articles, EU public documents and US patents. | |
| 2/3 of training data is English. The other part of data covers EU24 languages. | |
| The model was trained on JRC Big Data Platform. The model can be fine-tuned for other tasks. |