Create function to call a prediction method for every example in DCC JSON file, currently only supporting MedCAT's biLSTM, but can be expanded to roberta & rule based approach.
Loop through EMC DCC and save an unique id for every entity, in format file_start_stop, such as GP0001_5_12. This makes it easy to compare between methods.
Create function to calculate scores
Evaluate MedCAT's biLSTM (use the predict_one function from MedCAT, so not optimized for performance)
Rename some biLSTM files / folder
Create an outputmodels folder in root dir, of which the contents are not tracked by git. Useful for resulting models/data without having the specify local paths.
Also added:
Save predictions as feather-file and add to results folder.
sandertan
commented
3 years ago
GP0001_5_12. This makes it easy to compare between methods.predict_onefunction from MedCAT, so not optimized for performance)outputmodelsfolder in root dir, of which the contents are not tracked by git. Useful for resulting models/data without having the specify local paths.Also added:
resultsfolder.