Classifiers

The resource contains trained models, the pre-processed dataset, and all train logs of our results in the NLBSE2023 Tool Competition. In addition, the files are zipped separately for each classifier.

all_train_log.tar.gz: all training logs

final_final_dataset.tar.gz: the preprocessed dataset

unzip the zip file:

1.  checkpoint-best-ppl: the best model on the validation dataset
2.  checkpoint-last: the last model in the training. We used it to test the testing dataset to calculate the precision, recall, fscore recorded in the paper.
3. the dev.gold is the ground truth of the validation dataset; the dev.output is the ground truth of the validation dataset in the trainning stage. 
4. the test_0.gold is the ground truth of the validation dataset; the test_0.output is the ground truth of the validation dataset in the testing stage.
5. the test_1.gold is the ground truth of the testing dataset; the test_1.output is the prediction of our classifier.