bert-finetuned-ner

This model is a fine-tuned version of bert-base-cased on tner/ontonotes5 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1136
• Precision: 0.8655
• Recall: 0.8901
• F1: 0.8776
• Accuracy: 0.9798

Model Description

The model can be used for Token Classification, it has been fine-tuned for the purpose of Named Entity Recognition(NER).

Dataset Summary

Ontonotes5 NER dataset formatted in a part of TNER project.

Entity Types: CARDINAL, DATE, PERSON, NORP, GPE, LAW, PERCENT, ORDINAL, MONEY, WORK_OF_ART, FAC, TIME, QUANTITY, PRODUCT, LANGUAGE, ORG, LOC, EVENT

An example of train looks as follows.

{
    'tags': [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4, 5, 0, 0, 0, 0, 11, 12, 12, 12, 12, 0, 0, 7, 0, 0, 0, 0, 0],
    'tokens': ['``', 'It', "'s", 'very', 'costly', 'and', 'time', '-', 'consuming', ',', "''", 'says', 'Phil', 'Rosen', ',', 'a', 'partner', 'in', 'Fleet', '&', 'Leasing', 'Management', 'Inc.', ',', 'a', 'Boston', 'car', '-', 'leasing', 'company', '.']
}

Label ID

{
    "O": 0,
    "B-CARDINAL": 1,
    "B-DATE": 2,
    "I-DATE": 3,
    "B-PERSON": 4,
    "I-PERSON": 5,
    "B-NORP": 6,
    "B-GPE": 7,
    "I-GPE": 8,
    "B-LAW": 9,
    "I-LAW": 10,
    "B-ORG": 11,
    "I-ORG": 12, 
    "B-PERCENT": 13,
    "I-PERCENT": 14, 
    "B-ORDINAL": 15, 
    "B-MONEY": 16, 
    "I-MONEY": 17, 
    "B-WORK_OF_ART": 18, 
    "I-WORK_OF_ART": 19, 
    "B-FAC": 20, 
    "B-TIME": 21, 
    "I-CARDINAL": 22, 
    "B-LOC": 23, 
    "B-QUANTITY": 24, 
    "I-QUANTITY": 25, 
    "I-NORP": 26, 
    "I-LOC": 27, 
    "B-PRODUCT": 28, 
    "I-TIME": 29, 
    "B-EVENT": 30,
    "I-EVENT": 31,
    "I-FAC": 32,
    "B-LANGUAGE": 33,
    "I-PRODUCT": 34,
    "I-ORDINAL": 35,
    "I-LANGUAGE": 36
}

Data Splits

Name	Train	Validation	Test
ontonotes5	59924	8528	8262

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
0.0746	1.0	7491	0.0808	0.8399	0.8813	0.8601	0.9777
0.0527	2.0	14982	0.0801	0.8558	0.8884	0.8718	0.9795
0.0264	3.0	22473	0.0901	0.8624	0.8905	0.8762	0.9798
0.0147	4.0	29964	0.1057	0.8638	0.8880	0.8757	0.9796
0.0091	5.0	37455	0.1136	0.8655	0.8901	0.8776	0.9798

Framework versions

• Transformers 4.56.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.0