Fake News in the News: An Analysis of Partisan Coverage of the Fake News Phenomenon
Abstract
Since the 2016 U.S. election cycle, “fake news” (a term describing verifiably false and misleading news articles) has garnered increasing public attention. This work sheds insight onto this phenomenon by examining the way 10 popular partisan media sites discuss “fake news”. We use linguistic analysis techniques including Linguistic Inquiry and Word Count (LIWC), word embedding models, and supervised learning classifiers to analyze news stories containing the phrase “fake news” from left- and right-leaning news sites. Our results yield several insights, including that article text can be used to classify political affiliation with high accuracy, and that left-leaning sites focus on specific fake news stories and individuals involved, while right-leaning sites shift the focus to a narrative of mainstream media dishonesty more broadly.