this is interesting. though i keen to know how do you decide the categories?
It’s a mix of a classification model + LLM for when none of our topics are a high probability fit.
interesting! and this data, will it be open sourced or paid kn demand?
For now anyone can access the results here as a guest: https://search.nostr.wine/app/discover#/view/dc3500b0-cc04-11ed-b988-2561bfd60652 Long term our plan is to allow filter.nostr.wine subscribers to create custom feeds using our topics/language/sentiment labels through our relay. We can also offer the classifier as an API (that will not be free) and classification results through an API (perhaps a freemium model). All of that is TBD. This is my weekend proof of concept.