AI-powered tools help filter online hate targeting athletes – Technologist

Just a few more balls to serve. The men’s and women’s finals are approaching and, on Sunday, June 9, they will bring the 123^rd edition of the French Open tennis tournament to a close. Far from the Center Court and its hushed bourgeois atmosphere, another match is being played out, more devious and violent, more obscene and limitless: social media moderation. In this aggressive virtual space, French company Bodyguard is in charge, for the second year running, of managing the protection of the accounts – X, Instagram, Facebook and TikTok – of Roland-Garros and over 20 players involved in the tournament.

Athletes are under no obligation to entrust their accounts to this company. “It’s a service,” stressed Stéphane Morel, head of the French Tennis Federation (FFT), which organizes the French Open. This service is mentioned in the welcome booklet that each player receives before the start of the competition, which shows how much it is deemed important.

The premise of Bodyguard is simple: The technology, powered by artificial intelligence, is capable of removing “in real time,” according to the company (more exactly in a few milliseconds), hostile messages in 45 languages and also emojis considered insulting or discriminating (in the shape of a banana to insult a Black athlete, for example). These posts or comments are then masked and visible only to the author.

“In recent years, we’ve become aware of the importance of the mental health of our players,” explained Morel. “Quite pragmatically, it’s in our interest for our [athletes] to be in top physical and mental shape to deliver the best performances. We could have told everyone to manage their own platforms, but we feel it’s a moral obligation. We’re in the business of protection.”

‘More brutal for the smaller players’

Since the start of the tournament on May 26, Bodyguard has recorded and analyzed just over 70,000 messages sent to the pages of the FFT, Roland-Garros and the 20 or so players who place their trust in it. Just over 5% of these messages have been removed for being violent, abusive, racist or misogynistic – there are over 50 categories.

Poland’s Iga Swiatek – who competes in the final against Italy’s Jasmine Paolini on Saturday, June 8 – is one of the protected athletes. In almost two weeks, the world number 1 received 8,000 posts on her X and Instagram accounts: 33% were positive but some 2.4% were removed as hate comments. Insults, sometimes of a pornographic nature, were sent in French, English and her native language. “She won’t see them,” boasted Charles Cohen, founder and president of Bodyguard.

You have 54.51% of this article left to read. The rest is for subscribers only.