thinking about making a program that recommends videos that are on peertube. likes, titles, descriptions, and even closed caption can be collected locally and compared to the local users liked videos and watch time to rank videos the local user would like to watch.
later on there could be a server that people can willingly choose to send their data to build a better recommendation program.
I’m still in the research phase of this, but from my experience even implementing basic concepts would be effective
I know there is an anti algorithm crowd but this is opt in so this should be fine
You must log in or register to comment.
Kinda like https://quiblr.com/understanding_your_private_personalized_feed
But for Peertube instead of Lemmy? Sounds cool, I think Peertube needs it way more than Lemmy does
of cource i forget to log in to this account after posting. but so far Im building it to be more like a browser extension that collects your watch history and video meta data peertube video via api and locally suggests what video to watch next
Peertube, Loops and Mastodon all need something like that way more than Lemmy.
Nothing wrong if it includes a libre software license text file. Even better if it could be peer-to-peer.
yeah i hope people build on my work and improve it since i doubt ill even have this project working anytime soon.
i have hard time concieving of p2p system for this. is it like heres what this person likes so do “X”
P2P file system over a centralised server, keep users in control.
of okay. i guess if i get to the point where aggregating data from multiple users then yeah ill look into publishing the data somehow, most of the data will be pulled from peertube instances
I like the idea, even if it would be ‘just’ a small scale opt-in peertube search engine which would have better ranking algorythm with help of the data points you mention.
But I guess it would need to be opt-in for channels not for whole instances?
I have this list that I try to keep up to date: https://lemmy.wtf/post/15810205
