A project for political education based on simple machine learning applied to texts of political manifestos, annotated by the political scientists of the Manifesto Project.

The idea is to use the high-quality (but relatively low volume) manifesto project data annotated by human experts in order to train a text-classification model that can be used to extrapolate the experts' annotations to larger text corpora such as news articles. The hope is to support political education.

This code is partially based on an earlier project, which learned a similar text classification model on speeches in the German Parliament.

A preliminary demo can be found here.

Install virualenv(-wrapper). In the folder containing the directory cloned from github then type:

mkvirtualenv -a fipi fipi

Go to the web/ folder and install the dependencies with

pip install -r requirements.txt

Start the webserver with

python api.py

Open a browser window and navigate to localhost:5000.

Install Docker and start it. In the project root folder then build the docker image and start it with:

docker-compose up

Open a browser window and navigate to [IP-of-docker-container]:5000.

Install EB CLI

pip install awsebcli

Create and deploy app, then open it

eb init
eb create
eb open