Improve on Facebook, thanks to artificial intelligence, the detection of hate content.
97% of those removed were detected by automated systems before any user reported them.
In the previous quarter the percentage was 94%, 80.5% at the end of 2019, in 2017 it was 24%.
It is a data that emerges from the Periodic Report on the Application of Community Standards that shows how the company has dealt with this type of content.
Regarding two other sensitive topics, bullying and sexual harassment, detection thanks to AI went up on Facebook from 26% in the third quarter of 2020 to 49% in the fourth quarter, and on Instagram from 55% to 80%.
Another area that has made progress with AI is how systems work in multiple languages, even those widely spoken such as Spanish and Arabic: the amount of hate speech removed reached 26.9 million, compared to 22.1 in the previous quarter.
"We are slowly moving forward in replenishing our content review workforce globally - explains Guy Rosen, Facebook's VP of Integrity - although we expect this area to be affected by Covid-19 until a vaccine is widely available. available. Due to limited capacity, we prioritize the most harmful content, such as suicide and self-harm content, which is reviewed by our teams. "
For 2021, the company "intends to share more metrics on Instagram and add new policy categories on Facebook" and will continue "to optimize technologies and efforts to keep malicious content out of apps."
(HANDLE).