ARDC - Annotate and Train Classifier for Multi-Label

[evamaxfield]

One of Tom's biggest requests is to try and classify awards by possible software created rather than just "software created." Specifically his categories were:

• analysis code (used once for a paper / project)
• infrastructure (larger projects, well maintained, well used, multiple components)
• prototype tools (new libraries, tools, engines, databases, etc.)
• algorithms (5 lines of code +)
• no software created

might need to go back and forth with Tom about the definitions for these prior to annotation

ultimately, this is likely valuable for us in general in other work. largely just takes time for me to annotate (do we want to double annotate or just me?)

cc @nniiicc iirc you previously linked me a paper about "definitions for different types of software" do you have that link handy somewhere?

[evamaxfield]

found tom's definitions: https://zenodo.org/record/6378082