Scientists from Microsoft Research and the Technion Institute of Technology in Israel have developeda software that predicts future events based on the headlines of the public news in the last 22 yearsin the newspaperNew York Times, as well as information published in different online sources such as Wikipedia or FreeBase.
Eric Horvitz, co-director of Microsoft Research, explains that to develop this software each chain of events is associated “to a domain D of relevant words.” “For example, the challenge of predicting future deaths would consider terms such as ‘death’, ‘dying’ and other related ones; and to predict future epidemics, mentions of ‘cholera’, ‘malaria’ or ‘dysentery’ would be taken into account”, he clarifies . From the detailed analysis of the news, andgeneralizing certain repeating sequences of eventsand that are reflected in the information published between 1986 and 2008, scientists show that it is possible to “guess” certain future events.
“With this work we demonstrate thepredictive power of mining data from thousands of news stories“, Horvitz adds. The level of success achieved with this new software ranges between 70% and 90%, according to the authors in their article, where they assure that the real challenge will be to identify more patterns by monitoring increasingly abundant sources of information. Yestablish probabilistic associations that alert us to possible threats such as epidemics, social unrest and deaths from different causes. After all, Horvitz recalls, Mark Twain already put it clearly enough: “The past doesn’t repeat itself, but it does rhyme.”
