Update biLSTM to work with latest MetaCAT version from MedCAT's master branch. Zeljko is planning to release this version on PyPI very soon.
Update the biLSTM method we used to predict the negation of a single DCC sample. The previous predict_one() method from MetaCAT didn't survive the latest refactoring, so I created a new method based on Zeljko's suggestions.
Move predicting on complete DCC from 01_general_evaluation.ipynb to 05_evaluate_bilstm_on_complete_dcc.ipynb, so that the general notebook is now really general and does not require MedCAT
Generate new results on complete DCC
Improved file organization
Lots of changes in this PR, but I think the only thing really relevant for @lcreteig @bramiozo @marijnschraagen to look at is the updated 01_general_evaluation.ipynb
Next: Run updated pipeline on folds and gather results.
predict_one()
method from MetaCAT didn't survive the latest refactoring, so I created a new method based on Zeljko's suggestions.01_general_evaluation.ipynb
to05_evaluate_bilstm_on_complete_dcc.ipynb
, so that the general notebook is now really general and does not require MedCATLots of changes in this PR, but I think the only thing really relevant for @lcreteig @bramiozo @marijnschraagen to look at is the updated
01_general_evaluation.ipynb
Next: Run updated pipeline on folds and gather results.