legal-ner-roberta

This model is a fine-tuned version of roberta-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1077
• Overall Precision: 0.7753
• Overall Recall: 0.8653
• Overall F1: 0.8178
• Overall Accuracy: 0.9768
• F1 Case Number: 0.7298
• F1 Court: 0.8661
• F1 Date: 0.9777
• F1 Gpe: 0.7186
• F1 Judge: 0.9127
• F1 Lawyer: 0.8840
• F1 Org: 0.5699
• F1 Other Person: 0.8598
• F1 Petitioner: 0.7920
• F1 Precedent: 0.6789
• F1 Provision: 0.8964
• F1 Respondent: 0.7281
• F1 Statute: 0.8662
• F1 Witness: 0.8837

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: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Overall Precision	Overall Recall	Overall F1	Overall Accuracy	F1 Case Number	F1 Court	F1 Date	F1 Gpe	F1 Judge	F1 Lawyer	F1 Org	F1 Other Person	F1 Petitioner	F1 Precedent	F1 Provision	F1 Respondent	F1 Statute	F1 Witness
0.4031	1.0	344	0.1562	0.6489	0.7623	0.7010	0.9625	0.6507	0.8539	0.9644	0.5871	0.8140	0.8629	0.432	0.7338	0.5209	0.5802	0.7653	0.4011	0.8201	0.0
0.1274	2.0	688	0.1167	0.7342	0.8412	0.7841	0.9732	0.7047	0.8511	0.9733	0.6425	0.9243	0.8662	0.5013	0.8647	0.7278	0.6228	0.8757	0.6277	0.8565	0.7544
0.0869	3.0	1032	0.1077	0.7753	0.8653	0.8178	0.9768	0.7298	0.8661	0.9777	0.7186	0.9127	0.8840	0.5699	0.8598	0.7920	0.6789	0.8964	0.7281	0.8662	0.8837

Framework versions

• Transformers 4.57.1
• Pytorch 2.6.0+cu124
• Datasets 4.4.0
• Tokenizers 0.22.1