Negation_and_Uncertainty_Scope_Detection_mBERT_fine_tuned
This model is a fine-tuned version of bert-base-multilingual-cased on the NUBES dataset. This is a result of the PhD dissertation of Antonio Tamayo. It achieves the following results on the evaluation set:
- Loss: 0.1954
- Neg Precision: 0.9056
- Neg Recall: 0.9332
- Neg F1: 0.9192
- Nsco Precision: 0.8973
- Nsco Recall: 0.9079
- Nsco F1: 0.9026
- Unc Precision: 0.864
- Unc Recall: 0.9019
- Unc F1: 0.8825
- Usco Precision: 0.8210
- Usco Recall: 0.8650
- Usco F1: 0.8424
- Precision: 0.8886
- Recall: 0.9127
- F1: 0.9005
- Accuracy: 0.9698
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 2e-05
- train_batch_size: 8
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 7
Training results
Training Loss | Epoch | Step | Validation Loss | Neg Precision | Neg Recall | Neg F1 | Nsco Precision | Nsco Recall | Nsco F1 | Unc Precision | Unc Recall | Unc F1 | Usco Precision | Usco Recall | Usco F1 | Precision | Recall | F1 | Accuracy |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0.1688 | 1.0 | 1726 | 0.1386 | 0.8885 | 0.9064 | 0.8974 | 0.8706 | 0.8697 | 0.8702 | 0.8326 | 0.8309 | 0.8318 | 0.7694 | 0.8304 | 0.7987 | 0.8621 | 0.8775 | 0.8697 | 0.9633 |
0.111 | 2.0 | 3452 | 0.1455 | 0.8803 | 0.9128 | 0.8962 | 0.8726 | 0.8924 | 0.8824 | 0.8244 | 0.8622 | 0.8429 | 0.7391 | 0.8125 | 0.7741 | 0.8547 | 0.8888 | 0.8714 | 0.9624 |
0.0678 | 3.0 | 5178 | 0.1621 | 0.9204 | 0.9004 | 0.9103 | 0.9051 | 0.8961 | 0.9006 | 0.8571 | 0.8643 | 0.8607 | 0.8454 | 0.7935 | 0.8187 | 0.9010 | 0.8828 | 0.8918 | 0.9679 |
0.043 | 4.0 | 6904 | 0.1590 | 0.9099 | 0.9212 | 0.9155 | 0.8885 | 0.9042 | 0.8963 | 0.8525 | 0.8810 | 0.8665 | 0.8288 | 0.8426 | 0.8356 | 0.8874 | 0.9020 | 0.8946 | 0.9696 |
0.0278 | 5.0 | 8630 | 0.1679 | 0.9140 | 0.9272 | 0.9206 | 0.8977 | 0.9057 | 0.9017 | 0.8656 | 0.8873 | 0.8763 | 0.8315 | 0.8594 | 0.8452 | 0.8939 | 0.9076 | 0.9007 | 0.9696 |
0.0153 | 6.0 | 10356 | 0.1908 | 0.8994 | 0.9359 | 0.9173 | 0.8888 | 0.9101 | 0.8993 | 0.8676 | 0.8894 | 0.8784 | 0.8122 | 0.8638 | 0.8372 | 0.8819 | 0.9137 | 0.8975 | 0.9688 |
0.0082 | 7.0 | 12082 | 0.1954 | 0.9056 | 0.9332 | 0.9192 | 0.8973 | 0.9079 | 0.9026 | 0.864 | 0.9019 | 0.8825 | 0.8210 | 0.8650 | 0.8424 | 0.8886 | 0.9127 | 0.9005 | 0.9698 |
Framework versions
- Transformers 4.29.2
- Pytorch 2.0.1+cu118
- Datasets 2.12.0
- Tokenizers 0.13.3
- Downloads last month
- 118
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.