kgweber-sag / predications

University of Minnesota

Program for Clinical AI (P4AI) in the Center for Learning Health System Sciences

NLP Task - Analyst Position

Kate Weber

[e-mail]

Project Description

This repo contains training code for an algorithm to indicate whether a specific predication identified by SemRep is in fact contained within its source sentence, as well as source materials from the University.

If you wish to load the model and run it, the evaluation code in the second half of the main notebook does this.

Contents

• main_notebook.ipynb: The main content in this repo, contains narrative of the development and the actual training and evaluation code
• data_prep.ipynb: A short notebook containing some data exploration and preprocessing
• requirements.txt: A list of the packages used in this project
• Task_Description.md: The original task description
• README.md: This file
• data/: contains the original csv of the data as well as a pickle file of the data after preprocessing
• models/: contains the trained model (uploaded with Git Large File Storage)
• extras/: holding space for working-idea scratch notebooks and a plot included in the main notebook