i was thinking about ways to distill the feed down to only topics a user would like to see and i came across an open source machine learning model by facebook research. the use case would work by having the user define a set of topics they are interested in, and, given a post, the model would assign probabilities to each of the topics. the post need not contain the keywords explicitly. the post would make it through to the users customized feed if it passes a certain threshold for the likelihood the post is relevant to the user’s topic of interest.

model reference and examples: https://huggingface.co/facebook/bart-large-mnli?candidateLabels=urgent%2C+not+urgent%2C+phone%2C+tablet%2C+computer&multiClass=false&text=I+have+a+problem+with+my+iphone+that+needs+to+be+resolved+asap%21%21

was thinking about building a simple client to test this idea