 | rstWeb - Browser Annotation of Rhetorical Structure Theory |
rstWeb is an open source, browser based annotation tool for discourse analyses in Rhetorical Structure Theory and its enhanced version, eRST. It is meant to support collaborative, online annotation projects using just a web browser, without the need to install software for annotators.
New: Version 4 is now out, supporting enhanched Rhetorical Structure Theore (eRST)! For more information, see our paper.
Do RST and eRST: Online — Collaborative — Browser Based!
For a server installation, rstWeb needs Python > 2.6.X for its backend (versions at least up to Python 3.11 have been extensively tested), which accesses a SQLite database. The client runs in JavaScript and only needs a browser. User management and configurations use Michael Foord's logintools, and the front end uses the excellent jsPlumb, jQuery and Font Awesome, all of which are packaged with rstWeb and require no installation. To take screenshots of analyses automatically, it is optionally possible to install Selenium and PhantomJS, which automate saving scaled screenshots that fit your analysis size at high quality. This project's source code is available from Github - feel free to fork, contribute, get in touch and let me know what you're working on.
Features
- Annotators only need a browser
- Single mode structure-editor - no switching between linking, adding spans, multinucs and unlinking
- Import .rs3 files from RSTTool or plain text
- Support for multiple annotated versions of each document
- Download automatically scaled screenshots of your analyses
- Option to annotate discourse relation signals such as connectives / discourse markers
- Add tree-breaking, non-projective/multiple parentage eRST secondary edges
- Admin interface to manage users, assignments, and groups of documents in projects
New: Color coded discourse relation signal annotation and secondary edges!
Highlighted warnings for annotation scheme violations
Secondary edges (in blue) with color-coded eRST signals
Licensing and availability
rstWeb is open source, freely available and comes with absolutely no warranty under the MIT License. You can always find the latest release on Github:
Online demo
You can try out rstWeb at https://gucorpling.org/rstweb/. Just use the demo login credentials supplied on the login page (note that saving your work does not work for the demo login).
Example documents
You can get a lot of good examples of RST annotations online from the RST Website. At Georgetown University we are also building a corpus that includes RST annotation: the GUM corpus. You can download the RST annotations for GUM here:
Citing
If you're using rstWeb for a project or article, please cite the following paper:
- Gessler, Luke, Yang Liu and Amir Zeldes (2019) A Discourse Signal Annotation System for RST Trees. In: Proceedings of Discourse Relation Treebanking and Parsing (DISRPT 2019). Minneapolis, MN, 56-61.
Documentation
The rstWeb User Guide gives basic information on installing and using the software:
If you have questions about installing or using the interface, please contact me. If you have feature requests or find bugs, please use the issue tracker on the Github repository.
© Amir Zeldes 2015-2024