Artificial Intelligence Recognizes Pike News

A new study published on 21 August in Nature attempts to answer the baffling question of how online hate groups manage to thrive on social media platforms.

ynet logo
December 19, 2019

This article was translated from Hebrew

The social network is struggling hard with hate publicity and lies. According to Facebook data, most violent publications are currently blocked: 99 percent of child abuse posts, 96 percent of publications that include nudity and 97 percent of graphic violence publications. However, in the field of Pike News, success is much lower and Facebook reports a low rate of only 52 percent of false publications, which are identified and removed.

A recent paper presented at an academic conference on Natural Language Processing (NLP) presented two artificial intelligence approaches that can identify satire and distinguish it from Pike News under the guise of satire. The authors of the article, including Or Levy, CEO of, and researchers at George Washington University and Amazon's AWS, sought to present a different approach from previous research, where they tried to locate keywords that appear in satirical text. We trained the artificial intelligence on existing texts and let it learn for itself what the characteristics of satire are.

Read more in ynet