Skip to content

Releasing NLP Components as Stand alone Packages

Mark Sammons edited this page Oct 11, 2016 · 1 revision

The intent is to provide non-expert users with a single download that contains everything you need to run the relevant NLP component -- for example, to process a set of plain text files and generate a set of annotated text files.

Generate a zip that contains the models, dependencies, and NLP component jars, plus simple command line scripts to run a basic test (allowing the user to verify that the package works) and to process text files. Naturally, the documentation must cover the use of these scripts, in addition to programmatic use of the component (which will most likely involve telling the user to use Maven and managing dependencies that way).

There is a basic release.sh script in ner that should work as-is, or with minimal changes: https://github.com/IllinoisCogComp/illinois-cogcomp-nlp/blob/master/ner/scripts/release.sh

There should be a test run from the command line that annotates a specific text file and compares the output to a reference output file. For an example, look at the ner version: https://github.com/IllinoisCogComp/illinois-cogcomp-nlp/blob/master/ner/scripts/test-linux.sh

Clone this wiki locally