This model is a fine-tuned version of PlanTL-GOB-ES/roberta-base-biomedical-clinical-es on the CRAFT dataset. It achieves the following results on the evaluation set:
This model performs Named Entity Recognition for 6 entity tags: Sequence, Cell, Protein, Gene, Taxon, and Chemical from the CRAFT(Colorado Richly Annotated Full Text) Corpus in English. Entity tags have been normalized and replaced from the original three letter code to a full name e.g. B-Protein, I-Chemical.
This model is trained on augmented data created using Entity Replacement. 20% of the entities were replaced using a list of entities for each entity tag obtained from the official ontologies for each entity class. Three datasets (original, augmented, MT translated CRAFT) were concatenated.
More information needed
More information needed
The following hyperparameters were used during training:
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 0.0624 | 1.0 | 4078 | 0.1844 | 0.8002 | 0.7923 | 0.7963 | 0.9607 |
| 0.0284 | 2.0 | 8156 | 0.1937 | 0.8394 | 0.7988 | 0.8186 | 0.9637 |
| 0.0118 | 3.0 | 12234 | 0.2007 | 0.8285 | 0.8232 | 0.8258 | 0.9649 |
| 0.0043 | 4.0 | 16312 | 0.2224 | 0.8298 | 0.8306 | 0.8302 | 0.9659 |
pytorch_model.bin 502 MBtraining_args.bin 3.1 kB