search

facebookresearch / Clinical-Trial-Parser

Library for converting clinical trial eligibility criteria to a machine-readable format.
Apache License 2.0
163 stars 58 forks source link

No parsing of nominal criteria? #14

Closed kevinwu23 closed 3 years ago

kevinwu23 commented 3 years ago

The README describes nominal criteria as the most common, and yet in the test example, there appear to be none. Is there an issue with the way I am running the script or is this expected behavior?

salkola commented 3 years ago

You are probably doing nothing wrong. cfg_parse.sh yields ordinal and numerical criteria relations. ie_parse.sh extracts nominal variable names. Please see the general section of the architecture document.

kevinwu23 commented 3 years ago

Thanks for the help. I see that one of the TODO is "Post process IE relations and merge with CGF results" What general steps do you have in mind in terms of post-processing IE relations? I'm probably going to write something for myself, and can try to push something if this hasn't been done yet.

salkola commented 3 years ago

I look forward to your contributions.

Post processing includes removing duplicate criteria relations and making sure that the extracted relations are valid and consistent with each other.

Depending on your need, a good place to start might be implementing a negation module for the IE parser.