Citation Hunt

Citation Hunt is a simple tool for finding unsourced statements on Wikipedia in different languages. It is hosted at

This repository contains the full server and client code. The scripts/ directory contains all the scripts used for processing Wikipedia dumps. Hopefully they will be illustrative and reusable for similar applications.

I want to help!

That's great! There are many ways you can help. Please take a look at for guidelines and instructions.

Running in Toolforge

There are three major components to Citation Hunt and they are each set up in slightly different ways in Toolforge:

All of them run on Kubernetes. The Kubernetes configuration is in the k8s directory.

After logging in to, run the following commands to create the directory structure and enter the virtualenv:

$ mkdir www/python/
$ webservice --backend=kubernetes python3.7 shell
$ python3 -m venv www/python/venv/
$ . www/python/venv/bin/activate

Now, clone this repository, point uwsgi to it and install the dependencies:

$ git clone
$ ln -s ../../citationhunt www/python/src
$ pip install -r citationhunt/requirements.txt

and start the webservice:

$ webservice --backend=kubernetes python3.7 start

Then, generate the Cron jobs for Kubernetes:

$ (cd citationhunt; ./ | kubectl apply -f -)
$ kubectl get cronjobs # verify it

See scripts/ for more information about those jobs.

Finally, use k8s/compute_fixed_snippets.yaml to launch scripts/ to detect snippets that get fixed:

$ kubectl create --validate=true -f k8s/compute_fixed_snippets.yaml