ravenscroftj/candc_coresc repository overview

⁠C&C Tools with CoreSC Super Tagger

This docker container contains a very niche set of tools based on the (now defunct) C&C NLP tools package.

C&C tools is a CCG tagger that was originally developed by James Curran and Stephen Clark at the University of Sydney. In 2016 the original C&C tools project tracker went down and never came back up. The package is now widely considered to be abandoned⁠.

Our Python-based Sapienta⁠ implementation that enriches biomedical scientific papers with CoreSC scientific discourse information relies on a legacy version of C&C Tools that is no longer available elsewhere.

I have provided the 'gubbins' that we use to host C&C tools under the MIT open source license. I have included a tarball of C&C source code which comes with its own License which you should follow if possible (although it is difficult to say with abandonware how applicable this license is).

⁠About the docker image

The docker image is based on ubuntu:19.04 which is not the most slimline base image but since C&C relies on glibc which Alpine linux does not support out of the box⁠ it was the logical choice for a base image.

As with the original C&C tools we run a SOAP server on port 9004 which is exposed by the container.

The application is stored in /app/candc-1.00/ and the models are hosted in /app/models - a bash script called run_server.sh which lives in /app/candc-1.00/bin/ is used to launch the soap server and configure the models.

⁠About the models

We use C&C tools with a custom super tagger for CoreSC (called super_coresc) as well as the original biomedical models( pos_bio and super_bio)

There is a tar ball containing all of the models that we use in the main directory of this project.