topic modeling on authen posts

@karebu "another man kink" have i posted about beign a cuck

@authen

@amalthea @authen im vconfused what youre asking

@authen so cool how it finds associations between these and puts them in the same area

@karebu can you publish this somewhere is there a link to this

@authen

@amalthea @authen this isn’t a bluesky thing i queried my instance database for the text of every authen post and used bertopic on it

@amalthea @authen oh yeah i just downloaded a bluesky dataset off hugging face and did the same thing with it

@amalthea @authen i was trying to figure out how the trending algorithm on twitter/bluesky works and found out about topic modeling and i think it works by measuring the frequency of topics over a specific time period

@amalthea @authen it uses embedding llms which learn the association between text/image/audio like the words rain, umbrella, puddle would all have similar numerical representations to each other and even an actual image of what those words represent google started using this in 2018 on google search which is why it can understand human language

@karebu @authen