System based on machine learning

[Shishirgano9]

The model studied is that of an AI that would assign a score to each dubious information, according to its risk of virality. By concentrating their editorial work on the rumors with the “highest” scores, journalists can deny them before they become viral.

In October 2018, MIT created a fake news rating (determined in particular by the number and type of publication ).

The French startup Neutral News distinguishes between an article croatia cell phone number list that reports factual information and fake news through a semantic analysis (words, lexical fields, turns of phrase). To do this, it uses Decodex . a database created by journalists from Le Monde that lists more than 5,000 questionable articles, tweets, YouTube videos and Facebook posts.

But AI is currently unable to differentiate between false information that is the result of an error and intentional disinformation.

For Tristan Mendès France, Associate Lecturer at Paris Diderot University and specialist in digital cultures, " There is no well-defined limit between fake news and a factual article. It is rather a continuum on which a cursor must be placed ."