A top-secret catalogue of profanities has been amassed by Facebook after it trawled billions of posts in a bid to stamp out hate speech online.
The social media giant is reported to have compiled a list of slurs and swear words in almost every language which it will never publish to prevent offenders from “gaming the system” by misspelling the offending words.
A Facebook source, who confirmed the existence of the list to a national newspaper this week, said a combination of artificial intelligence and 15,000 human “reviewers” now refer to the words to enforce the site’s “community standards”.
With more than 2.3 billion users worldwide, the site has long been criticised for struggling to moderate the level of inappropriate content uploaded every day.
Facebook’s chief technology officer Mike Schroepfer told a conference earlier this year of the constant battle against offensive material, which can also come in the form of memes, images and slang.
He said: “We build a new technique, we deploy it, people work hard to try to figure out ways around this.”
The site has also developed a system called Rosetta to scan text embedded in images to analyse whether the meaning conveyed by the text alongside the picture is offensive.
Rosetta is understood to be scanning more than one billion images every day.
Hate speech is currently defined as a “direct attack on people based on protected characteristic” like race, ethnicity and sexual orientation.
In the first quarter of the year, the site took action on four million pieces of content, a rise from 2.5million for the same period last year.
