vabatista/geological-ner

This model is a Brazilian Portuguese Named Entity Recognition (NER), based on neuralmind/bert-base-portuguese-cased base model and specialized in Geological concepts. It was trained for 3 epochs using the dataset from this paper.

You can find the notebook used to train the model here. Trainer output was:

To use this model, run into a pipeline:

## run the prediction
txt = YOUR_TEXT
classifier = pipeline("ner", model=model, tokenizer=tokenizer, aggregation_strategy='simple')
entities = classifier(txt)

## display in a fancy way
dict_ents = {
    'text': txt,
    'ents': [{'start': ent['start'], 'end': ent['end'], 'label': ent['entity_group']} for ent in entities],
    'title': None

}

displacy.render(dict_ents, manual=True, style="ent")