Machine-Assisted Script Curation (MASC)

The MASC schema curation tool is currently being developed for the LESTAT project.

Requirements

Python must be installed. It currently runs on Python 3.7, but newer versions might work as well.

NodeJS v14 and npm v6 must be installed. Tools like nodenv or nvm can be used for installation. For alternative methods, refer to the NodeSource blog posts Installing Node.js Tutorial: Using nvm (macOS and Ubuntu) or Installing Node.js Tutorial: Windows (Windows) for instructions.

Installation

Use the provided Makefile to install MASC:

make install

To configure deployment-specific settings, create pycurator/.env and add settings specified in pycurator/common/config.py.

Usage

The application has two main components, and they must be run simultaneously to use the application:

• To run the back end, navigate to pycurator/flask_backend and run bash start_gunicorn.sh. The server will be available at http://localhost:5000/. Usage of the Flask server is possible, but due to a bug in sentence-transformers, the application might crash when run with Flask instead of Gunicorn.
• To run the front end, navigate to angular-frontend and run npx ng serve. The application will be hosted at http://localhost:4200/, which is viewable in a modern web browser. The application will automatically reload if any source files are changed.

While these instructions are sufficient for a local deployment, they should not be used on an actual server. It is up to the user to determine the proper server configuration for themselves.

GPT-2 component

The GPT-2 component currently must be run manually.

It is recommended to use the batch_run.py script to do so. It automatically finds all schemas without a JSON file and runs a Slurm job for each. It doesn't use anything but the standard library, so the virtual environment is not necessary.

batch_run.py must be run from pycurator/gpt2_component to allow paths to work properly. Use the command PYTHONPATH=../../ python -m pycurator.gpt2_component.batch_run.

Do not start a new run until all previous jobs are finished. batch_run.py only checks whether there is output when deciding which schemas to run, so it can't tell if there is a currently running job for a schema.

Publications

For more information about the project, see the related paper:

Ciosici, Manuel, et al. "Machine-Assisted Script Curation." Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies: Demonstrations, Association for Computational Linguistics, 2021, pp. 8–17. ACLWeb, https://www.aclweb.org/anthology/2021.naacl-demos.2.

License

MIT