Sorry, news anchors - you might soon have to share your job with avatars. A virtual news technology is turning heads by quickly creating news stories and commentary, no humans required.
Designed by the Intelligent Information Laboratory at Northwestern University, News At Seven (newsatseven.com) is an automated system combining 3-D avatars, images, video, opinion and generated speech. The website collects news stories from the Web, edits them automatically and formats the content for artificial anchors.
News At Seven doesn't just amass stories; it also pulls relevant blog posts and images to complement the editorial. So far, the beta version broadcasts entertainment and gossip news only. But the project is in its infancy, says Kristian Hammond, the computer-science professor quarterbacking the site.
Once users are able to enter keywords, watch out, Mansbridge. Hammond says the final product will deliver a customized newscast based on user-input phrases.
Type in "Afghanistan" and the system scours RSS feeds across the Web, pulls appropriate editorial, edits them and slips the words into text-to-speech software. Then virtual hipster anchors begin broadcasting the stories.
"It will be a complete news site," says Hammond. The technology intrigued Wired Magazine enough to be featured at last month's Wired NextFest exhibit in Chicago.
The system assigns opposing opinions to two anchors, who exchange the back-and-forth banter CNN viewers have come to expect. For comedic effect, Hammond has also introduced a talking fish who rants about the assigned story.
"This op-ed reporter finds snippets of opinion across the Web relating to the topic, parses them and compiles something coherent," says Hammond.
Animation is done via Flash, and the audio is created by an off-the-shelf text-to-speech tool. To find video, News At Seven searches popular sites like YouTube. It can compile a story within 15 seconds.
But how can technology scouring a blog, say, identify a sarcastic phrase? Won't that screw up story compilation?
"Sarcasm is a killer," Hammond admits. "But our system tries to find what people really like or hate about a subject, looking for keywords like ‘hate.'"
Hammond hopes the site will attract brands looking for "rich media advertising." A system like his could potentially integrate news stories with a social-network-like experience, pulling images of bands and posting CD release details when discussing North By Northeast, for example.
Hammond is tight-lipped about how News At Seven will generate revenue, but he is looking for partners to "expand the dynamics of the system."
There will be a wait for a completely personalized version. Hammond predicts it will take six months to release that innovation, which he hopes will revolutionize the future of broadcast news.
"News At Seven was inevitable," he declares. "If I hadn't worked on it someone else would have."