roberta-base_legal_ner_finetuned

This model is a fine-tuned version of FacebookAI/roberta-base on the Darrow LegalLens Shared Task NER dataset. It achieves the following results on the evaluation set:

• Loss: 0.2416
• Law Precision: 0.8319
• Law Recall: 0.8785
• Law F1: 0.8545
• Law Number: 107
• Violated by Precision: 0.8361
• Violated by Recall: 0.7183
• Violated by F1: 0.7727
• Violated by Number: 71
• Violated on Precision: 0.5
• Violated on Recall: 0.5
• Violated on F1: 0.5
• Violated on Number: 64
• Violation Precision: 0.6494
• Violation Recall: 0.7032
• Violation F1: 0.6752
• Violation Number: 374
• Overall Precision: 0.6843
• Overall Recall: 0.7143
• Overall F1: 0.6990
• Overall Accuracy: 0.9553

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
• lr_scheduler_warmup_steps: 500
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Law Precision	Law Recall	Law F1	Law Number	Violated by Precision	Violated by Recall	Violated by F1	Violated by Number	Violated on Precision	Violated on Recall	Violated on F1	Violated on Number	Violation Precision	Violation Recall	Violation F1	Violation Number	Overall Precision	Overall Recall	Overall F1	Overall Accuracy
No log	1.0	85	0.7386	0.0	0.0	0.0	107	0.0	0.0	0.0	71	0.0	0.0	0.0	64	0.0	0.0	0.0	374	0.0	0.0	0.0	0.7707
No log	2.0	170	0.3510	0.0	0.0	0.0	107	0.0	0.0	0.0	71	0.0	0.0	0.0	64	0.2072	0.2781	0.2374	374	0.2072	0.1688	0.1860	0.8901
No log	3.0	255	0.2471	0.4265	0.2710	0.3314	107	0.0	0.0	0.0	71	0.3810	0.125	0.1882	64	0.3965	0.4813	0.4348	374	0.3996	0.3523	0.3745	0.9199
No log	4.0	340	0.1996	0.7596	0.7383	0.7488	107	0.5128	0.5634	0.5369	71	0.3827	0.4844	0.4276	64	0.5101	0.6096	0.5554	374	0.5324	0.6136	0.5701	0.9385
No log	5.0	425	0.1984	0.7946	0.8318	0.8128	107	0.64	0.6761	0.6575	71	0.5091	0.4375	0.4706	64	0.5102	0.6684	0.5787	374	0.5669	0.6737	0.6157	0.9449
0.5018	6.0	510	0.2447	0.7456	0.7944	0.7692	107	0.75	0.6761	0.7111	71	0.4068	0.375	0.3902	64	0.6110	0.6845	0.6456	374	0.6296	0.6705	0.6494	0.9465
0.5018	7.0	595	0.2264	0.8125	0.8505	0.8311	107	0.7736	0.5775	0.6613	71	0.4754	0.4531	0.4640	64	0.6276	0.7166	0.6692	374	0.6570	0.6964	0.6761	0.9511
0.5018	8.0	680	0.2243	0.8598	0.8598	0.8598	107	0.7812	0.7042	0.7407	71	0.4912	0.4375	0.4628	64	0.6209	0.7139	0.6642	374	0.6641	0.7094	0.6860	0.9541
0.5018	9.0	765	0.2327	0.7934	0.8972	0.8421	107	0.7808	0.8028	0.7917	71	0.4231	0.5156	0.4648	64	0.6037	0.7005	0.6485	374	0.6346	0.7273	0.6778	0.9547
0.5018	10.0	850	0.2416	0.8319	0.8785	0.8545	107	0.8361	0.7183	0.7727	71	0.5	0.5	0.5	64	0.6494	0.7032	0.6752	374	0.6843	0.7143	0.6990	0.9553

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0
• Datasets 2.21.0
• Tokenizers 0.19.1