POS-Tagger Bio Portuguese

We fine-tuned the BioBERTpt(all) model with the MacMorpho corpus for the Post-Tagger task, with 10 epochs, achieving a general F1-Score of 0.9818.

Metrics:

              Precision  Recall  F1    Suport
accuracy                         0.98  38320
macro avg     0.95       0.94    0.94  38320
weighted avg  0.98       0.98    0.98  38320

F1:  0.9818 Accuracy:  0.9818

Parameters:

nclasses = 27
nepochs_total = 30
nepochs_stop = 12 (stop in 12th because early stop)
batch_size = 32
batch_status = 32
learning_rate = 1e-5
early_stop = 3
max_length = 200

Acknowledgements

This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) - Finance Code 001.

Citation

coming soon

Questions?

Please, post a Github issue on the NLP Portuguese Chunking.