bpmn-information-extraction-v2

This model is a fine-tuned version of bert-base-cased on a dataset containing 104 textual process descriptions.

The dataset and the training scripts can be found here: https://github.com/jtlicardo/process-visualizer/tree/main/src/token_classification

The dataset contains 5 target labels:

• AGENT
• TASK
• TASK_INFO
• PROCESS_INFO
• CONDITION

It achieves the following results on the evaluation set:

• Loss: 0.2179
• Precision: 0.8826
• Recall: 0.9246
• F1: 0.9031
• Accuracy: 0.9516

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
1.9945	1.0	12	1.5128	0.2534	0.3730	0.3018	0.5147
1.2161	2.0	24	0.8859	0.2977	0.4524	0.3591	0.7256
0.6755	3.0	36	0.4876	0.5562	0.7262	0.6299	0.8604
0.372	4.0	48	0.3091	0.7260	0.8413	0.7794	0.9128
0.2412	5.0	60	0.2247	0.7526	0.8571	0.8015	0.9342
0.1636	6.0	72	0.2102	0.8043	0.8968	0.8480	0.9413
0.1325	7.0	84	0.1910	0.8667	0.9286	0.8966	0.9500
0.11	8.0	96	0.2352	0.8456	0.9127	0.8779	0.9389
0.0945	9.0	108	0.2179	0.8550	0.9127	0.8829	0.9429
0.0788	10.0	120	0.2203	0.8830	0.9286	0.9052	0.9445
0.0721	11.0	132	0.2079	0.8902	0.9325	0.9109	0.9516
0.0617	12.0	144	0.2367	0.8797	0.9286	0.9035	0.9445
0.0615	13.0	156	0.2183	0.8859	0.9246	0.9049	0.9492
0.0526	14.0	168	0.2179	0.8826	0.9246	0.9031	0.9516

Framework versions

• Transformers 4.26.1
• Pytorch 1.13.1+cu116
• Datasets 2.10.0
• Tokenizers 0.13.2