In this repository we train a model to classify sentences from academic papers using AllenNLP and explain the predictions using LIME.
To train the model, run this:
python3 run.py train experiments/sentence_classifier_architecture.json --include-package sentence_classifier.dataset_readers --include-package sentence_classifier.models -s ./trained_model
And this is how we evaluate the model:
python3 run.py evaluate trained_model/model.tar.gz --include-package sentence_classifier.dataset_readers --include-package sentence_classifier.models --evaluation-data-file data/test/
The notebook has the code to explain predictions with LIME.