umcu / negation-detection

Negation detection in Dutch clinical text.
GNU General Public License v3.0
3 stars 0 forks source link

BERT-dataset loader #45

Open bramiozo opened 2 years ago

bramiozo commented 2 years ago

The TextDataset* functions from ner_training.py do not preserve the original index. To resolve this we pass along the id's along with the data and the labels: This ids vector is filled with "IDXX_begin_end".

This ids vector is passed to the evaluation function and added to the prediction output.