Massive media event data analysis to assess world-wide political conflict and instability
Mining massive media data to gain insights into various
kinds of cultural trends, including political conflicts and instabilities,
are a primary goal of computational social science and the new
interdisciplinary
field, "culturomics". While the sheer size of media data makes
this task challenging, a greater hurdle is the nonstationarity of data,
manifested in several ways, which invalidates surge in media coverage as
a reliable indicator of political change. We demonstrate the use of advanced
statistical, information-theoretic, and random fractal methods to
analyze CAMEO-encoded political events data. In particular, we show
that on the country level, event distributions obey a Zipf-Mandelbrot law,
and interactions among countries follow an exponential law, indicating
that local or prioritized events dominate the political environment of a
country. Most importantly, we find that world-wide political instabilities,
such as the Arab Spring, are associated with breakdown or enhancement
of long-range correlations in political events.