صفر is a user on octodon.social. You can follow them or interact with them if you have an account anywhere in the fediverse.
If you don't, you can sign up here.
haha i am mildly excited for the sentiment analysis plugin for MRF
i'm going to modify my emojify MRF plugin to insert the right kind of emoji depending on if the message is positive or negative
i'm going to modify my emojify MRF plugin to insert the right kind of emoji depending on if the message is positive or negative
@kaniini what algorithm/nlp library is it based on?
@lichen
the AFINN-165 sentiment analysis algorithm for microblogs
the particular implementation is https://hex.pm/packages/veritaserum
i guess vertiaserum is a truth serum used in the harry potter books
the AFINN-165 sentiment analysis algorithm for microblogs
the particular implementation is https://hex.pm/packages/veritaserum
i guess vertiaserum is a truth serum used in the harry potter books
@kaniini interesting I'll try to dig up some benchmarks to see how it performs, thanks for the link.
in my experience sentiment analysis is hard as hell, I totally agree in seeing this just as an aid tool for mods and admins and not used to automatize any practical action
@lichen right now i believe it is just experimental. i doubt the actual model will be based on AFINN-165 at the end of it.
@lichen but none the less, it is cool to see people doing this type of research and using pleroma as the platform to do it. tools and techniques to assist moderators is a major area in the fediverse where research is very useful :)
@kaniini sure research is always useful, and it's worth exploring techniques to assist moderators, with the caveat, in this case, that even the best sentiment analysis would fail to recognize some types of trolling behavior, and some kinds of indirect subtle harassment
@lichen yes, absolutely, it is not perfect. hince human component to review anything it flags :)
