3 ms·
Text classification with Transformers is hard. Humans are expensive. My guess is also that the link got flagged because of who shared it, but that's just a gue
by svaha1728 5y ago
Text classification with Transformers is hard. Humans are expensive.
My guess is also that the link got flagged because of who shared it, but that's just a guess.
- _-david-_ 5y agoInformation can be accurate or fake based on who said it?
- svaha1728 5y agoI'm guessing the heuristics behind their text classification model, not if that's a valid assumption outside of the Facebook metaverse. My guess would be that you could also figure out word salad to throw at their False Information model that would create false positives.
- Roboprog 5y agoSure. Orange man rushed out a shot. It must be bad. Gray man said get the shot. It’s good. I’m not saying that kind of “thinking” is valid, but it’s apparently widely accepted. (I got the shot, but I can see why some people would have considerations)
- kevinmgranger 5y agoThe context of a page can provide misleading implications. But "False / Fake" isn't the word I'd use for that, to be fair.
- specialist 5y agoSurely government sites in their host countries can be whitelisted?
- beezischillin 5y agoI might be wrong on this but I think their fact checking partners just go all around on Facebook and fact check specific links and claims, I don’t think it’s flagged automatically and paired with a fact check, or at least the few I looked at were directly referencing the source for the content that was shared.