Kalev Leetaru wants to forecast political crises, like a meteorologist does the weather.

The assistant director of text and digital media analytics at the University of Illinois Institute for Computing in Humanities, Arts and Social Sciences, Leetaru said his supercomputer project can't predict uprisings and downfalls.

But as with "a 70 percent chance of rain," it can suggest there's a direction to politics based on media coverage and retroactively did so using 100 million news articles to forecast the so-called Arab Spring and the location of Osama bin Laden.

"We have reached a critical point in society where there's so much digital data — in the Google Books digitalization, for instance — that we can begin to predict patterns such as displeasure with a foreign leader," he said.

The articles in his database date back to 1945.

"One collection from 1979 to the present was created by the British government. They sample everything on radio and TV, take a sample of the news each day, and created an archive," Leetaru said.

He said the database has 10 billion people and 1,000 trillion relationships. It was fed into a University of Tennessee SGI Altix supercomputer, known as Nautilus. The machine's 1024 Intel cores have a total processing power of 8.2 teraflops (8.2 trillion floating point operations per second).

All that data shows public opinion about leaders, governments and rebels.

"It's like shining a flashlight in a dark room. Patterns emerged about Egypt as the tone of comments changed and darkened. Twitter has predictive value for the stock market," he said.

So far, Leetaru's work hasn't actually forecast any upheavals.

"This was done after the fact using only the data was available at time," he said.

"But can we get insights? We can see conflict and cooperation," he said. "When data mentions France, how often does it also mention Spain? What you end up with is this big network of the world's relationships."

For example, Nautilus generated graphs for different countries involved in the Arab Spring, the popular uprisings in such countries as Tunisia, Egypt and Libya.

"In each case, the aggregated results of thousands of news stories showed a notable dip in sentiment ahead of time — both inside the country, and as reported from outside," he said in a recent paper called "Culturomics 2.0: Forecasting Large-Scale Human Behavior Using Global News Media Tone in Time and Space."