Research Lab WS1920, Sticky Fake News: A study on diffusion and sentiment. On the context of Forschungspraktikum, Online Political Polarization, content analysis for detecting fake statements group. University of Koblenz-Landau.
We perform a study on the differential diffusion of fake political and gossip news stories distributed on Twitter as well as opinion mining on the social context of these news pieces. The data comprises FakeNewsNet, a versatile dataset containing fake and true news pieces. We find that fake political stories diffuse farther and significantly faster than fake gossip stories. However, when it comes to the total number of users involved (including followers) fake gossip stories seem to have higher numbers than fake political stories. We also found that the general sentiment of direct Tweets for political news is more negative whereas for gossip news it was neutral. This changed when examining the replies to these Tweets where political news had an overall neutral sentiment, whereas gossip news was related with a more positive sentiment. We interpret this as a possible enjoyment of fake gossip stories.
A complete explanation may be found in our report document sheet below: Paper
The dataset cannot be published due to Twitter's policy rules and news publisher copyrights. Link to the git repository of the dataset utilized here