Alibaba Tests New Technology That Detects Pornographic Content

Alibaba’s new porn-busting service converts speech to text and checks it against a database of licentious lexemes.

A livestreamer broadcasts from her bedroom in Beijing, Feb. 10, 2015. Jason Lee/Reuters

What’s man-made, fluent in a dozen or so languages, and has a knack for identifying libidinous moans?

Introducing the Ali A.I. Voice Anti-Garbage Service, an innovation from tech giant Alibaba that uses voice-recognition technology to identify pornography.

The tool can detect pornographic content — which is illegal in China, if vaguely defined by law — spoken in numerous languages and dialects, including English, Japanese, Russian, Cantonese, Sichuanese, and Hunanese. The salacious sounds have nowhere to hide, as audio content can be monitored both on local files and during real-time playback.

The testing phase for the new service began on Aug. 20, and it is slated to be launched in September, according to state-run newspaper Science and Technology Daily. The program will be the latest addition to Alibaba’s arsenal of porn-busting technologies used to identify sexual — or “yellow,” to use the Chinese term — content in videos, images, and text.

An unnamed representative from Alibaba’s security department told Science and Technology Daily that the program works by first converting audio into text, which it then checks against a database of tens of thousands of disallowed terms. Beyond key words and phrases, the program also scans for sounds that may be suspect. Just like a child, the source said, the technology needs to be “trained” to differentiate sounds — for example, by processing volumes of Cantonese TV dramas.

“You can think of it as a robot or a child,” said the Alibaba employee. “It needs to be fed, trained, and taught until it acquires more capabilities.” Continue to read the full article here.


-This is original content by Sixth Tone and has been republished with permission.