Krishna Bharat, Principal Scientist at Google and creator of Google News

There are currently 41 editions of Google News in 18 languages. Founder Krishna Bharat began this shuttle service to news and networked information on the Web based on the idea that since Google had millions of global users anyway, why not create a newspaper that would learn about what people are reading in every language, every country, every newspaper, and every story? Google News does not own or create anything, they merely aim to make what others write discoverable.

Bharat spoke at a Symposium on Computation and Journalism at Georgia Tech on Feb. 22, 2008. “We believe in fair and unbiased ranking,” he said. Google news crawlers gather stories in “clusters” based on title, snippet from description, source, timestamp, URL, and links to “other perspectives.” This YouTube video, In Conversation: Google’s principal scientist-Krishna Bharat, explains his thinking behind developing Google News algorithms.

When a big story breaks Google News gathers and tracks a story’s evolution by weighing aggregated editorial interest by some fairly subjective metrics: originality, freshness, the quality or expertise of the source, and the localness of the source.

The audience was comprised of many of the journalists who wrote the stories that Google News gathers up in an untouched-by-human-hands, automated fashion. They wanted to talk about whether gathering news by computation was really unbiased, and further, was it fair to use other people’s work to build a Google News empire.

I was there to talk about the ideas, technology and workflow behind Beyond Penguins and Polar Bears. Since most NSDL content creators are interested in having their content become more discoverable in searches at Google, I asked Bharat what technologies or methods he would suggest to make that happen.

He reiterated that new, frequently updated, and well-written content with lots of links to other perspectives would be found and ranked highly by Google News, but he did not offer specific methods. These suggestions are not new, but worth keeping in mind as we seek to maximize the semantic effects of interconnected search optimizations strategies.