I think it would be a good start, but would still require some human vetting. We played with some simple NLP solutions (trying to run TF-IDF on the keywords/des

I think it would be a good start, but would still require some human vetting. We played with some simple NLP solutions (trying to run TF-IDF on the keywords/description/readme of packages to determine the category), the results weren't great honestly, so we opted of curating everything manually for now.