ABSTRACT:

In an effort to help government officials anticipate significant events such as political unrest, disease outbreaks, or natural disasters, the US government's Intelligence Advanced Research Projects Agency is launching a mass dataset mining effort, hoping to develop technologies that can mine disparate sources such as blogs, search engine results, Internet traffic, webcams, and many others. Researchers in the natural and social sciences have long been doing similar work, however, which might serve to show the current limitations of computational linguistics, especially in trying to discern, on the fly, events that could have significant policy implications.