Proposal for auto filtering bots and other unwanted posters.
I am soliciting ideas for a Chrome extension which automatically parses a thread’s posts, uploads that information to my analysis system, and automatically returns an ID block list. The chrome extension then automatically applies the block list to hide unwanted posts. This is all based on learned post characteristics such as sentence structure, informational entropy, and more.
Would you find such a Chrome extension for Zig Forums/4chan helpful?
Slight tangent, and maybe unnecessary: it is also possible to build a list of verified known good users with the same system, as tripcodes could be automatically randomly generated and tracked by the upstream platform, so all other users including hostile agents would not see the chain. There is a drawback however: this does introduce a privacy single point of failure which would be identical to reusing a trip code everywhere. So maybe this would be useful as an opt-in service. This still allows anonymous posting, as the system would be the only one tracking the chain of random IDs, but since your trail is marked and noted by other users of the automatic blocking tool, you would not be randomly accidentally filtered by the aforementioned system. You would be either fully filtered or fully unblocked.
Agreed, the properties and quantity of red text would be a great feature to track for auto blocking
Cameron Perez
Filenames Red Text Text
I don't want to get into parsing images at this time, but it remains a future possibility.
David Thompson
Don't get hung up on the specific browser. The point is to find out if there are any specific techniques you would like to see implemented for automatic blocking of shitposters, JIDF, etc.
No Chrome. Only Javascript so we can paste it into [Options].
Filters have been proposed before but they usually lead to type II errors (false positives) not type I (false negatives).
Levi Sanchez
although JS is insecure so idk
maybe a decentralized database with shill attack vectors? kinda mirroring their attack on us?
Samuel Watson
Filtering sages alone would cut down immensely on unwanted posts.
Jack Price
Fuck, didn't know that was even an option. Hell if anything it makes my life easier.
I agree, there needs to be some sort of sensitivity (true positive rate) vs specificity (true negative rate) reporting.
Nigger, JS isn't inherently "insecure." Read the fucking source code, I would not minify it. Plus the only code that needs to run is a client side parser that triggers an AJAX request to the black box filtering API, then parses its output for reporting or decision making (blocking).
Jace Morales
Idiotic bait, even for a jew.
Bentley Smith
I thought about this method but making the black box public would defeat its purpose, as it would be possible to train a generative adversarial model to reverse engineer it. I can mitigate that by throttling.
Jeremiah Watson
It would be possible to make a sage-specific filter and subscribe to just that one if you believe it works. I'm not judging how you hang yourself.
Kayden Moore
Maybe salt the data? Throw in a bunch of garbage from old books? Idk sry need to go wish I could help more this evening. Remember to use fuzzy logic if possible friend.
Robert Bennett
"I want a hugbox where I'm not exposed to anyone who disagrees with me"
It doesn't have to be an autofilter. You can obtain additional information next to each post such as the historical information entropy of a particular poster, and have those metrics reported next to each post with a little bit of Javascript manipulation.
Might be possible to report a histogram of trigger words for each poster as well, indicating what words were used to elicit replies from other posters.
just make an opt-in mod system. Whoever you subscribe to is allowed to minimize posts in your own view automatically. If you don't like the guy you are subscribed to, switch filters.
