Who's Moderating Your Content Right Now?

Most user-generated content on the Web passes through some kind of swear filter. According to WebPurify CEO Jonathan Freger, the process is surprisingly democratic. Since 2007 WebPurify has been working behind the scenes to filter profanity for everyone from WordPress bloggers to Fortune 500 companies, as well as public media enterprise PBS. The company, which has offices in Los Angeles and New York, recently added image moderation and enterprise licensing to its list of services. We caught up with Freger last week to get his thoughts on the future of free speech on the Web.

“What [people who advocate free speech] don’t understand is that what they’re doing when they’re typing into someone else’s website is typing on their property,” Freger pointed out. Site administrators “are are self-moderating,” he said. “It’s not like a government entity that’s doing it.” Freger doesn’t foresee any government-mandated profanity filters on the Web other than the ones services like his provide to individual sites. “It’s too wide open right now,” he explained.

“I got started doing this not necessarily because I was an anti-profanity person,” Freger said, but because after 10 years of working in Web development, he found himself creating a variation of the same profanity filter for each of his clients who wanted some way to monitor what was being said on their sites. To save himself time Freger created an API that could be easily integrated into any web application. It got so much traction that he decided to form a company that focuses exclusively on filtering content.

“It’s subjective, what’s profane and what isn’t,” Freger said. There are religious sites that target anything even remotely offensive, while other clients, especially bloggers, are more concerned with racism. One university he worked with needed a way to filter tweets and hash tags before projecting comments from students onto a campus wall – a feature that Twitter itself didn’t provide.

There were other, less obvious uses for the filtering system. One of Freger’s first clients was a college student who wanted help with a game called Profane Game, the goal of which is to type in as many bad words as possible in 30 seconds or less ( Editor’s note: it’s not as easy as it looks). That student went on to work at Google, he added. Other people want to block content related to a competitor or negativity surrounding their own brands.

The biggest challenge for the company is modifying the algorithms for foreign languages. For that, Freger relies on consultants to help with the research, especially in languages with different character sets. He enjoys cutting and pasting the filtered phrases into Google Translate to see what they mean. Sometimes the words lose some of their potency outside of their cultural context. In the case of Arabic, “they often involve farm animals,” he said. “That speaks volumes to me.” WebPurify is currently available in English, Spanish and Arabic and Freger said he plans to add as many as five languages per year.

Now that the cloud has made the service easier to scale, the company has also extended the service to larger companies and hired an outside workforce to also scan images for obscenity. “No algorithm is going to be 100% in this space,” said Freger. But services like his “fill a grey area between people who can afford to monitor something full-time and people who can’t.”