ner-portuguese-br-bert-cased

This model aims to help reduce the need for models in Portuguese.

How to use:

from transformers import BertForTokenClassification, DistilBertTokenizerFast, pipeline

model = BertForTokenClassification.from_pretrained('rhaymison/ner-portuguese-br-bert-cased')
tokenizer = DistilBertTokenizerFast.from_pretrained('rhaymison/ner-portuguese-br-bert-cased'
                                                    , model_max_length=512
                                                    , do_lower_case=False
                                                    )

nlp = pipeline('ner', model=model, tokenizer=tokenizer, grouped_entities=True)

result = nlp(f"""
A passagem de uma frente fria pelo Rio Grande do Sul e Santa Catarina mantém o tempo instável,
e chove a qualquer hora nos dois estados. Há risco de temporais no sul e leste gaúcho.
No Paraná segue quente, e pancadas de chuva ocorrem a partir da tarde, também com risco de temporais.
""")

###output

[{'entity_group': 'LOC',
  'score': 0.99812114,
  'word': 'Rio Grande do Sul',
  'start': 36,
  'end': 53},
 {'entity_group': 'LOC',
  'score': 0.99795854,
  'word': 'Santa Catarina',
  'start': 56,
  'end': 70},
 {'entity_group': 'LOC',
  'score': 0.997009,
  'word': 'Paraná',
  'start': 186,
  'end': 192}]

He has various named classes. Follow the list below:

• O: 0
• B-ANIM: 1
• B-BIO: 2
• B-CEL: 3
• B-DIS: 4
• B-EVE: 5
• B-FOOD: 6
• B-INST: 7
• B-LOC: 8
• B-MEDIA: 9
• B-MYTH: 10
• B-ORG: 11
• B-PER: 12
• B-PLANT: 13
• B-TIME: 14
• B-VEHI: 15
• I-ANIM: 16
• I-BIO: 17
• I-CEL: 18
• I-DIS: 19
• I-EVE: 20
• I-FOOD: 21
• I-INST: 22
• I-LOC: 23
• I-MEDIA: 24
• I-MYTH: 25
• I-ORG: 26
• I-PER: 27
• I-PLANT: 28
• I-TIME: 29
• I-VEHI: 30

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

• Loss: 0.0618
• Precision: 0.8965
• Recall: 0.8815
• F1: 0.8889
• Accuracy: 0.9810

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2

Comments

Any idea, help or report will always be welcome.

email: rhaymisoncristian@gmail.com