I appreciate you throwing some data for a few keywords together in your comment, but you'll excuse me if I lean toward trusting the published paper instead.
FWIW, paper is downloadable here (at least, for now): http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2769645
They did not compare it to the news cycle, or consider that people are reading different pages about the same subject (AQ in Iraq for example being superseded by ISIS).
Sorry if I sound harsh but this study was terribly executed.
Doing it properly would require a lot more resources than Wikipedia daily page view dumps. I think the NY Times would be well placed to do a study on this using Wikipedia page views, their page views and their mentions of topics overtime.
Such as the April 2013 Boston Marathon bombings. Or the rise of ISIL and Boko Haram in early 2013.
I still think the study was well executed, but the instant effect seems to disappear when you remove the 6-7 months before mid june 2013.
The long-term effect they claim to observe is more shaky. (I think one can find "statistically significant" groupings of say, Harry Potter articles, that show a similar peak and decline around the Snowden revelations.)
Can we find more flaws in this paper?