rst-workbench

The rst-workbench lets you easily install and run existing RST parsers usind Docker containers. It provides a convenient web app for running multiple RST parsers concurrently and allows you to visualize and post-edit the trees produced by the parsers.

Here's a video demonstrating its usage.
You can also try our demo server at https://rst-workbench.arne.cl/.

For installation instructions and usage examples, see below.

Supported Parsers

Currently, we support the following RST parsers that we provide as Docker containers with a REST API and pre-trained on RST-DT:

• HILDA (Hernault et al. 2010) [our repo] [paper]

  • HILDA itself is only available from the Prendinger lab, so we can't provide automated builds but only the build scripts here.

• Feng and Hirst (2014) [our repo] [paper]

• DPLP (Ji and Eisenstein 2014) [our repo] [paper]

• Heilman and Sagae (2015) [our repo] [paper]

• CODRA (Joty et al. 2015) [our repo] [paper]

• StageDP (Wang et al. 2017) [our repo] [paper]

  • the parser requires EDU-segmented input, so we combine it with the NeuralEDUSeg discourse segmenter (Wang et al. 2018)

Installation

Please install Docker and docker-compose first, then clone this repo including its submodules.

git clone --recurse-submodules https://github.com/rst-workbench/rst-workbench

Usage on your local machine

To run the rst-workbench on your local machine, simply run
DOMAIN=localhost docker compose up --build --force-recreate in the folder into which you cloned the repo.

On the first run, this will take a long time and download the Docker containers of all supported RST parsers.
On all subsequent runs, this will be much quicker and simply start the Docker containers from your hard drive.

Once it has started, go to http://localhost:8000/ in your browser to use the web interface.

By setting set the DOMAIN to localhost, the rst-workbench will be only available on the machine where you run this command. The parameters --build and --force-recreate should not be needed on normal runs, but you will need them whenever you change the configuration.

Usage on a server

The repo contains everything necessary to run our demo server at https://rst-workbench.arne.cl/ . We usually start it like this:

DOMAIN=rst-workbench.arne.cl docker compose -f docker-compose-server.yml up --build --force-recreate

To make it work on your domain, you will need to change the domains and email variables in the file nginx-certbot/init-letsencrypt.sh.

Citation

If you use rst-workbench, rst-converter-service or any of the dockerized RST parsers, please consider citing our paper:

@inproceedings{neumann-2021-rst-workbench,
    title = "Using and comparing {R}hetorical {S}tructure {T}heory parsers with rst-workbench",
    author = "Neumann, Arne",
    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations",
    month = apr,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://www.aclweb.org/anthology/2021.eacl-demos.1",
    pages = "1--6",
}