POS-Tagger Portuguese

We fine-tuned the BERTimbau model with the MacMorpho corpus for the Post-Tagger task, with 10 epochs, achieving a general F1-Score of 0.9826.

Metrics:

              Precision  Recall  F1    Suport
accuracy                         0.98  33729
macro avg     0.96       0.95    0.95  33729
weighted avg  0.98       0.98    0.98  33729

F1:  0.9826 Accuracy:  0.9826

Parameters:

nclasses = 27
nepochs = 30
batch_size = 32
batch_status = 32
learning_rate = 1e-5
early_stop = 3
max_length = 200

Tags:

Questions?

Please, post a Github issue on the NLP Portuguese POS-Tagger.