Semi Supervised NLP for Fine-Grained Medical Report Classification

This folder contains the most useful notebooks to train/eval/run the semi-supervised model.

• quick_training is functionally the same as detailed_training, but without the complex documentation and methods defined in the notebook. It pulls the methods from the architecture folder, which has scripts of predefined functions.
• interpret allows for model evaluation, through a variety of metrics, saliency scores of words in test strings, and encoding generation tools.
• inference runs the given models on any given data, outputting a folder of cohort files, with relevant model predictions.