Open edengby opened 1 year ago
Looks like the hospitals lexicon isn't used by default. I would suggest to add it to the PhiIdentifier class similarly to how diseases and medications are handled. First, add it to the imports here: https://github.com/8400TheHealthNetwork/HebSafeHarbor/blob/4a0ba72e0f0ab5e8421f7b45637c463decaddbe3/hebsafeharbor/identifier/phi_identifier.py#L30
Then, instantiate it similar to this: https://github.com/8400TheHealthNetwork/HebSafeHarbor/blob/4a0ba72e0f0ab5e8421f7b45637c463decaddbe3/hebsafeharbor/identifier/phi_identifier.py#L163
Hope this helps
Thank you for the quick response. If I'm not mistaken, the diseases and medications are not anonymized by default. What else should be changed? Example of run (took from lexicons files):
גדעון לבנה הגיע היום לבית החולים שערי צדק עם תלונות על כאבים בחזה קיבל אנטיטריפסין והרספטין בנוסף אובחן עם רגל שטוחה ללא היסטוריה קודמת
I haven't tested it but adding it to the ner_signals
should be adding it to the detection pipeline. Perhaps the confidence score is too low and causing some results to be removed?
If I want to add values to the HOSPITALS, how it should be done? (hebsafeharbor/lexicons/hospital_lexicon.py) I noticed that the code is not referenced. How come? I tried running the docker-compose-development build and up but it did not change the list to recognize the new values.