US university coding future of news

Click to follow
The Independent Online

Personalized newscasts culled from the Web and presented by digital avatars. Baseball stories written by computers using raw data.

Television anchors and sports reporters may not need to start looking for other jobs just yet but students and professors at Northwestern University are working to make this futuristic vision of news a reality.

The Internet-based customized newscast has been cooking in the Intelligent Information Lab (InfoLab) at Northwestern's McCormick School of Engineering and Applied Science in Illinois for over three years.

The project, known as "News at Seven," uses a software program to create a "virtual news show" presented by animated anchors who speak - somewhat robotically - using text-to-speech technology.

The program crawls news stories and blogs on the Web for information about a particular topic, matches it up with pictures and video and generates a script.

Viewers can choose from a selection of avatars and customize a newscast to their particular interests.

InfoLab director Kristian Hammond, a professor of electrical engineering and computer science at Northwestern, said the voices used by the digital avatars were the biggest "limiting factor" to "News at Seven" for the moment.

"The information-gathering is more compelling than the presentation," Hammond told AFP.

"The nature of computer-generated voices is that they're not actually very expressive right now," he said. "That limits the system considerably."

Hammond said the computer-generated story project known as "Stats Monkey" was probably further along.

"Stats Monkey" uses a computer to write a story about a baseball game from its box score - the play-by-play and historical data. It also generates a headline and includes a picture to illustrate the action.

"Stats Monkey," a joint project between Northwestern's Medill School of Journalism and the McCormick engineering school, uses statistical models to figure out what the "news" is in the story and is able to pick out key plays and players.

The "Stats Monkey" stories, which carry the byline "The Machine" and are generated in just a few seconds, match up favorably with those written in haste by flesh-and-blood reporters, Hammond said.

"We generated a story for every game," he said. "We found it was a little bit better."

Besides just providing a recap of the game, "Stats Monkey" can also plug in quotes from individuals involved in the stories.

The Northwestern professor said plans were in the works to try out "Stats Monkey" on swimming and even business stories such as company quarterly or annual earnings reports.

"Anything where you have raw numerical information," he said.

Hammond said both projects show potential, "but we think there's more of a future for 'Stats Monkey' than the "News at Seven' right now."

Hammond said the projects had evoked interest from a number of media organizations which he could not identify at this time.

"We're pretty much unique in this," he said.

Comments