bpmn-information-extraction
This model is a fine-tuned version of bert-base-cased on a dataset containing 90 textual process descriptions.
The dataset contains 5 target labels:
AGENTTASKTASK_INFOPROCESS_INFOCONDITION
It achieves the following results on the evaluation set:
- Loss: 0.2909
- Precision: 0.8557
- Recall: 0.9247
- F1: 0.8889
- Accuracy: 0.9285
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: 15
Training results
| Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy |
|---|---|---|---|---|---|---|---|
| 2.0586 | 1.0 | 10 | 1.5601 | 0.1278 | 0.1559 | 0.1404 | 0.4750 |
| 1.3702 | 2.0 | 20 | 1.0113 | 0.3947 | 0.5645 | 0.4646 | 0.7150 |
| 0.8872 | 3.0 | 30 | 0.6645 | 0.5224 | 0.6882 | 0.5940 | 0.8051 |
| 0.5341 | 4.0 | 40 | 0.4741 | 0.6754 | 0.8280 | 0.7440 | 0.8541 |
| 0.3221 | 5.0 | 50 | 0.3831 | 0.7523 | 0.8817 | 0.8119 | 0.8883 |
| 0.2168 | 6.0 | 60 | 0.3297 | 0.7731 | 0.8978 | 0.8308 | 0.9079 |
| 0.1565 | 7.0 | 70 | 0.2998 | 0.8195 | 0.9032 | 0.8593 | 0.9128 |
| 0.1227 | 8.0 | 80 | 0.3227 | 0.8038 | 0.9032 | 0.8506 | 0.9099 |
| 0.0957 | 9.0 | 90 | 0.2840 | 0.8431 | 0.9247 | 0.8821 | 0.9216 |
| 0.077 | 10.0 | 100 | 0.2914 | 0.8252 | 0.9140 | 0.8673 | 0.9216 |
| 0.0691 | 11.0 | 110 | 0.2850 | 0.8431 | 0.9247 | 0.8821 | 0.9285 |
| 0.059 | 12.0 | 120 | 0.2886 | 0.8564 | 0.9301 | 0.8918 | 0.9285 |
| 0.0528 | 13.0 | 130 | 0.2838 | 0.8564 | 0.9301 | 0.8918 | 0.9305 |
| 0.0488 | 14.0 | 140 | 0.2881 | 0.8515 | 0.9247 | 0.8866 | 0.9305 |
| 0.049 | 15.0 | 150 | 0.2909 | 0.8557 | 0.9247 | 0.8889 | 0.9285 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.0+cu116
- Datasets 2.8.0
- Tokenizers 0.13.2