PsyNamic

A project to automatically extract relevant information from abstracts of clincal studies about psychedelic treatments of psychiatric conditions. This includes named entity recognition and single-/multilabel classification via fine-tuned BERT-based models.

This repository includes two submodule:

|-- PsyNamic-Webapp: the webapp written in dash

|-- PsyNamic-Prodigy: the dockerized prodigy setup used for annotating data

How to install

Either install this repository and the submodules in one go:

git clone --recurse-submodules git@github.com:Ineichen-Group/PsyNamic.git

or install the repos separaetly

git clone git@github.com:Ineichen-Group/PsyNamic.git
git clone git@github.com:Ineichen-Group/PsyNamic-Prodigy.git
git clone git@git@github.com:Ineichen-Group/PsyNamic-Webapp.git

Install web app

Checkout the README of PsyNamic-Webapp repo: https://github.com/Ineichen-Group/PsyNamic-Webapp/README.md

Install prodigy

Check out the Makefile.

How to deal with submodules

Keeping the reference up to date

• Work within the submodules, commit and push
• Update the references in the parent repository PsyNmiac

git submodule update --remote
git add PsyNamic-Prodigy PsyNamic-Webapp
git commit -m "Updated submodule references to the latest commit"
git push origin main

What is where?

• Anything related to the data and the annotation process is in the data folder
  • data/raw_data contains the raw data from the literature search with ASReview and David's relevant/irrelevant classification
  • data/prepared_data contains the annotated data, cleaned and ready for training
  • data/prodigy_inputs and data/prodigy_exports contains the unannotated prodigy input and the annotated prodigy output
  • data/iaa contains the inter-annotator agreement data
  • data/prediction_data contains the unannoated data for the prediction and the whole pulling newest research from PubMed pipeline
  • it also contains all scripts to process the data #TODO: adjust paths; I moved some scripts around
• models conains
  • the scripts for training and feeding data while training
  • performance evaluation and plots --> the actual trained models are on the server
• PsyNamic-Webapp contains the webapp for the living systematic review
• PsyNamic-Prodigy contains the prodigy setup for the annotation process
• test contains some not up-to-dat test cases for the data processing