Post #2122092

Is there an effort to create a recommender algorithm for the Fediverse using TF-IDF weighted LSA or LDA exclusively trained on the public posts of people who opt-in? Those are old school LLMs that have decent performance on narrow tasks from the before times. They can't be used for gen AI crap but can be labelers or to recommend similar content. I imagine such a tool would be useful for creating feeds on the Fediverse to aid in discoverability.