In this folder, you can find training scripts of BiLSTM-CRF and Turkish BERT, and BERT-CRF models for Named Entity Recognition task.
- flair==0.8.0.post1
- torch==1.7.1
Two datasets should be downloaded for training. Download data from the v1.0 release of mukayese-datasets.
Then extract the files using unzip:
unzip milliyet-ner.zip
unzip wikiann.zip
After downloading milliyet-ner and WikiANN datasets, you can run the training script. For instance to run the training of BERT:
bash bert/run.sh wikiann milliyet-ner