README.md

CollabKG: A Learnable Human-Machine-Cooperative Information Extraction Toolkit for (Event) Knowledge Graph Construction

• CollabKG is an open-source IE annotation toolkit that unifies NER, RE, and EE tasks, integrates KG and EKG, and supports both English and Chinese languages.

• CollabKG combines automatic and manual labeling to build a learnable human-machine cooperative system. In particular, humans benefit from machines and meanwhile, manual labeling provides a reference for machines to update during annotation.

• Additionally, CollabKG is designed with many other appealing features (customization, training-free, propagation, etc) that enhance productivity, power, and user-friendliness. We holistically compare our toolkit with other existing tools on these features.

• CollabKG Extensive human studies suggest that CollabKG can significantly improve the effectiveness and efficiency of manual annotation, as well as reduce variance.

  🖥 Try out CollabKG online

  🖹 CollabKG paper

  🎥 CollabKG systems demonstration video

  📌 Overview of how to use CollabKG

  📌 Frequently Asked Questions (FAQ)

  📨 Feel free to reach out if you have any questions by emailing [email protected]

Getting started

CollabKG can be built using Docker. Before doing so please add a secure token to the TOKEN_SECRET field in /server/.env for user password hashing and salting. After this, in the repository root directory, execute:

$ make run

or alternatively:

$ docker-compose -f docker-compose.yml up

Issues, Bugs and Feedback

If you come across any issues, bugs or have any general feedback please feel free to reach out (email: [email protected]). Alternatively, feel free to raise an issue, or better yet, make a pull request 🙂.

Known Issues/Bugs

Future features

Acknowledges

Thanks to the QuickGraph team for their support.

Attribution

Please cite our [paper] if you find it useful in your research: