NEWSGAC is a research project which aims at transparent automatic classification of genres of newspaper articles. The project is a cooperation between the University of Groningen, the Amsterdam Center for Mathematics and Computer Science and the Netherlands eScience Center.
In the project, we developed an online platform for applying machine learning models to text data, with the opportunity to closely analyze the performance of the models. This repository contains the code of this platform.
In order to run the platform at your computer, you need to have docker available on your system. Then execute the following commands in a command line environment (instructions for Linux):
git clone https://github.com/newsgac/platform.gitcd platformdocker build . -t "newsgac/newsgac"export $(egrep -v '^#' .env.default | xargs)docker stack deploy -c docker-compose.yml -c docker-compose.dev.yml newsgacdev
When these commands have successfully completed, the platform will be available as a web server on the address: http://YOUR-IP-ADDRESS:5050
Steps 1, 2 and 3 need to be executed only once for installing the system. Both step 4 and step 5 are required each time when you start the system.
Optional steps:
- For adaption to local environment: edit file .env.default or create your own version
- For Jupyter notebook support:
docker build . -f jupyter/Dockerfile -t "newsgac/jupyterhub" - During production:
docker build ./nginx -t newsgac/nginx - installation instructions for usage of a kubernetes cluster (CLARIAH)
Stopping the system:
docker service rm newsgacdev_database newsgacdev_frog newsgacdev_frogworker newsgacdev_redis newsgacdev_web newsgacdev_worker
Note that it takes a few seconds to completely stop all parts of the system.
You might want to run flask outside of Docker (because it is e.g. easier to attach a debugger).
- Follow
Setup Instructions for DOCKERinstructions so that all services are online (Mongo, Redis, FROG, celery workers). - Make sure the flask docker container is DOWN:
docker service rm newsgacdev_web
- Set up a virtual environment (python 3.7) and install the requirements:
pip install -r requirements.txt
- Setup the correct environmental variables (
.env.local) e.g. by running
export $(cat .env.local | xargs)
- To run from command line, navigate to
platform/and run:
PYTHONPATH=. python newsgac/app.py
- The local web server will be running on
http://localhost:5050.
Typically tasks are executed by celery workers. If you want to debug a task you can do one of two things:
- Run a celery worker in debug mode
- Make sure
CELERY_EAGER=True(or unset). This will cause celery to run tasks in the main thread instead of offloading it to workers.
docker run --name=mongo -it --rm -d mongo
docker run \
--name=newsgactest \
-it \
--network=container:mongo \
--mount type=bind,src="$(pwd)"/newsgac,destination=/newsgac/newsgac \
--entrypoint=sh \
newsgac/newsgac -c pytest --cov=newsgac --cov-report=xml
docker stop newsgactest mongo
- Setup local (virtual) environment as when running flask locally
- Load the test env vars:
export $(cat .env.test | xargs)
- Make sure the database, Frog and redis are running (e.g.
docker stack deploy -c docker-compose.yml -c docker-compose.dev.yml newsgacdev - Load env variables, then run tests using
pytest .
E.g. to create a user:
- Start console using docker (or from you local environment using
python):
docker exec -it newsgac_dev web python
- Import database & user model
from newsgac import database
from newsgac.users.models import User
- Create new user
u = User(email='[email protected]', password='testtest', name='Test', surname='User')
u.save()
- You can now login from the frontend as this user.
docker stack ps newsgacdevdocker service ps newsgacdev_workerdocker service inspect newsgacdev_workerdocker service logs newsgacdev_worker
A. Bilgin, E. Tjong Kim Sang, K. Smeenk, L. Hollink, J. van Ossenbruggen, F. Harbers and M. Broersma, Utilizing a Transparency-driven Environment toward Trusted Automatic Genre Classification: A Case Study in Journalism History (2018)
@inproceedings{bilgin2018utilizing,
title={Utilizing a Transparency-driven Environment toward Trusted Automatic Genre Classification: A Case Study in Journalism History},
author={Bilgin, Aysenur and Tjong Kim Sang, Erik and Smeenk, Kim and Hollink, Laura and van Ossenbruggen, Jacco and Harbers, Frank and Broersma, Marcel},
booktitle={2018 IEEE 14th International Conference on e-Science (e-Science)},
pages={486--496},
year={2018},
organization={IEEE}
}
K. Smeenk, A. Bilgin, T. Klaver, E. Tjong Kim Sang, L. Hollink, J. van Ossenbruggen, F. Harbers and M. Broersma, Grounding Paradigmatic Shifts In Newspaper Reporting In Big Data. Analysing Journalism History By Using Transparent Automatic Genre Classification (2019)
@inproceedings{smeenk2019dh,
author = "Kim Smeenk and Aysenur Bilgin and Tom Klaver and Erik Tjong Kim Sang and Laura Hollink and Jacco van Ossenbruggen and Frank Harbers and Marcel Broersma",
title = "{Grounding Paradigmatic Shifts In Newspaper Reporting In Big Data. Analysing Journalism History By Using Transparent Automatic Genre Classification}",
booktitle = "{Digital Humanities Conference 2019 (DH2019)}",
publisher = "{Utrecht, The Netherlands}",
year = "2019"
}
T. Klaver, E. Tjong Kim Sang, A. Bilgin, K. Smeenk, L. Hollink, J. van Ossenbruggen, F. Harbers and M. Broersma, Introducing a transparency-driven platform for creating, comparing and explaining machine learning pipelines (2019)
@inproceedings{klaver2019ictopen,
author = "Tom Klaver and Erik Tjong Kim Sang and Aysenur Bilgin and Kim Smeenk and Laura Hollink and Jacco van Ossenbruggen and Frank Harbers and Marcel Broersma",
title = "{Introducing a transparency-driven platform forcreating, comparing and explaining machinelearning pipelines}",
booktitle = "{ICT-Open}",
publisher = "{Hilversum, The Netherlands}",
year = "2019",
note = "(demo presentation abstract)"
}
- Aysenur Bilgin ([email protected])
- Erik Tjong Kim Sang ([email protected])
- Tom Klaver ([email protected])