Introduction

SWAN is a web-based system for natural language annotation. Its key features are:

• support of discourse annotation (i.e., annotation of complete texts) by allowing to scroll through the entire document
• focus on usability for limited tag sets (quick selection)
• a graph visualization box showing the discourse structure of the document
• performs fast with large documents

If you want to learn more, read our paper and check out the GitHub Wiki for more information on functionality and setup!

Screenshots

Admin and project manager view

Annotator view

Demo

If you want to try out (a slightly older version of) SWAN yourself, check out our Web Demo!

Roadmap

• Add support for crowdsourcing services like Amazon Mechanical Turk
• Add support for cross-document annotation (links between multiple documents)
• General performance and editor improvements
• For ongoing development and suggestions, see our issues page

Contact

We are happy to hear from you! If you love SWAN, send us an email. If you find some bugs or if you have suggestions for improvements, please open an issue on this GitHub site!

People involved with SWAN:

• Julia Dembowski (Saarland University, Department of Computational Linguistics)
• Stefan Grünewald (Saarland University, Department of Computational Linguistics)
• Annemarie Friedrich (Saarland University, Department of Computational Linguistics)
• Simon Ostermann (Saarland University, Department of Computational Linguistics)

People that have worked on SWAN in the past:

• Nicklas Linz (German Research Center for Artificial Intelligence (DFKI GmbH))
• Rafael Theis (Saarland University, Department of Computer Science)
• Janna Herrmann (Saarland University, Department of Computer Science)
• Timo Gühring (Saarland University, Department of Computer Science)