[16:19:18] <isaacj>	 jaketangosierra: one option is to use inferred topics instead of the category-based ones. i have a model that does this well and if you need topics for lots of articles, I'm happy to help get you setup (https://meta.wikimedia.org/wiki/Machine_learning_models/Proposed/Language_agnostic_link-based_article_topic_model_card). if you do want to stick with the category-based approach, i know less about this, but a few potential pointers: 
[16:19:18] <isaacj>	 https://dlab.epfl.ch/people/west/pub/Piccardi-Catasta-Zia-West_SIGIR-18.pdf and https://mdpi-res.com/d_attachment/computers/computers-08-00060/article_deploy/computers-08-00060-v2.pdf. generally, i think folks use the main categories as top nodes (https://en.wikipedia.org/wiki/Category:Main_topic_classifications) and then assign top categories in a proportional way based on which top-level nodes they go to. so if an article has two 
[16:19:18] <isaacj>	 categories and one leads to History and STEM and one leads to History, then the article would be 66% History and 33% STEM.
[17:08:47] <jaketangosierra>	 isaacj: these are all great pointers! We might go down the path of inferred topics, depending on how things go!
[17:19:17] <isaacj>	 Good luck! Don't hesitate to reach back out