April 4, 2016 by Canadian Press
In a world where U.S. and UK homeowners can buy cyber bullying endorsements, a British Columbia tech firm has developed technology to keep online trolls under the bridge.
Community Sift, based in Kelowna, has built digital armour for social media and gaming companies trying to protect their virtual worlds. The chat filter and moderation tool examines real-time website commentary, chat room conversations and banter between game players.
“We’re not just talking about four-letter words,” said CEO Chris Priebe, a senior programmer and security specialist. “We want to get rid of bullying across the entire Internet.”
The firm’s technology advances a global campaign against digital abuse in part spurred by the 2012 suicide of Amanda Todd, a teenager from Port Coquitlam, B.C., who was victimized by online sexual exploitation.
“The Amanda Todds of the world, we want to prevent that,” said Karen Olsson, the firm’s chief operating officer. “We want to be part of the solution.”
Based on the firm’s analysis of four billion messages sorted daily, less than one percent of social users behave badly yet they’re causing the bulk of harm. Offensive material is classified into categories such as bullying, sexting, racism and bomb threats.
The firm has catalogued more than one million phrases used frequently by trolls, for example, “u r so ugly,” Priebe said.
The technology takes context into account when identifying toxic behaviour. It combines machine learning and human verification by employing artificial intelligence and 30 language specialists. Priebe said online users are shielded from cyberbullies like anti-virus software protects computers.
“We’re looking for social viruses that are causing social destruction of social products and social lives.”
About 30 global clients are already using Community Sift. The flexible technology is tailored to client specifications, such as modifying content filters to be age appropriate.
An internal database query by the firm estimated it has protected at least 34 million users over a recent two-week period in its U.S. data centre alone.
The technology sifts the posts to emphasize positive comments from the 40 percent of online users who are normally well-behaved to derail the attacks.
“They’re going to say, ‘You’re beautiful, you’re wonderful, you’re helpful,”’ Priebe said. “Now she’ll have two voices inside her head and she can build the ability to handle all this bullying.”
The firm builds reputations for users participating online, and detects when someone crosses into a high-risk threshold. Consequences may include limiting identified trolls to certain queues where a moderator can decide if the content is inflammatory, silencing them automatically or banning them outright.
“We always joke you can put them in the basement with all the other trolls and let them harass themselves,” Olsson said.
This story was originally published by Canadian Insurance Top Broker.