Researchers from the Federal University of São Carlos and the University of São Paulo have developed a tool to identify fake news which can be used on the web or WhatsApp. The project is called Fake News Detector and was financed by the National Council of Technological and Scientific Development (CNPq). It has a 90 percent accuracy in detecting false information, according to creators. Opinion articles, for instance, may be incorrectly flagged as fake news.
To create the tool, researchers used data from 3,600 fake articles and 3,600 genuine ones. A code learned to identify the main characteristics and patterns of each group. This analysis showed that fake news tends to have more spelling errors, shorter sentences, a more diverse vocabulary, and intense use of nouns, pronouns, and adverbs. And, of course, falsehoods rely heavily on adjectives.
From that information, the detector makes its diagnosis over whether the news can be trusted, or if it is fraudulent. Among the preferred subjects of fake news are politics (58 percent), TV and...
Search
