Contribution of ConTeXTMatch model

[warreveys]

Description

• I contributed the ConTeXTMatch model, a bi-encoder trained for skill-extraction, and a part of the following academic work: https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=11079977
• The more functional tests were deleted again. I however kept the basic tests about the compatibility of the implementation with the ModelInterface. This could maybe be made standard, to test all custom implementations. Let me know what you think!

Checklist

• [yes] Added new tests for new functionality
• [yes] Tested locally with example tasks
• [yes] Code follows project style guidelines
• [yes] Documentation updated
• [yes] No new warnings introduced

[Mattdl]

Great contribution Warre!

First 5-ish commits are still changes from the previous addition to evaluate multiple models.
Can you check where you based your new branch of? Is this the main?

• Ideally you would (1) go to your main and set this repo's main branch as upstream, so your local main is up-to-date with this one (2) rebase your local branch with the new main. (3) You'll have to force push the branch as you'll be overwriting the commit history.
• Alternatively for the second step, you can just merge the main of this repo instead of rebasing. This is to make sure the latest version is compatible

Second thing we'd like to check, to make sure the implementation of ContextMatch is correct, could we give it a shot on SkillExtract Tech (smallest)? Then we can see if it reproduces the original results in the paper.

[warreveys]

I have indeed branched my new addition from my previous one, as this was still the most up to date version. Both comments are very valid, and I'll make sure to (1) contribute in a standard way and (2) also check with the paper results.

[Mattdl]

Great functional test. LGTM