3 days agoArtificial Intelligence Solves False News Problem
Periskopi(6 years ago)
(6 years ago)This year, in early November, presidential elections are held in the US. Last year was marked by a pre-election campaign filled with false “reports that resulted in false news being discovered and eliminated more by social networks and other platforms. In the United States, especially in the election year and [...]
Last year was marked by a pre-election campaign filled with false “reports that resulted in false news being discovered and eliminated more by social networks and other platforms.
In the US, especially in the year of elections and around the world, the public is polarised on two political sides.
Often people on the Internet, perhaps unconsciously, shut down in their rooms following people and the media only from their political spectrum. This leads to further polarisation and radicalisation of attitudes, and it is increasingly difficult to weigh which news is biased and which provide more objective information.
The Bipartisan Press) tries to bring balanced stories and news to its portal and to be as transparent as it comes from its authors' political points.
To that end, they have trained an artificial intelligence system that is able to reread “text and determine how unilateral it is, and what political option is being supported. It only works in English and only on American occasions.
If you put a political text into their system, it will analyze it, and within seconds you give a result to prejudice between -1 (completely left, democratic) and 1 (completely right, Republican). The largest zero-designment will also mean a greater objective of the text.
The system is based on machine learning, Tensorflow and Pytorch libraries, trained in tens of thousands of texts, and is said to have 96 percent accuracy.
If you follow the American political scene, you can try the system yourself at this address: https://www.
