“We know from social psychological research that fake news and pseudoscientific texts circulate on the Internet with a reach ten times greater than real news,” T. Ineke Nemeth told Index. According to the head of the Department of General Linguistics at SZTE, the Covid-19 pandemic has clearly shown the damage that fake news can cause in small and large societies.
Recognize fake news
During the coronavirus pandemic, fake news has taken a measurable toll on human life by instilling distrust in public opinion with pseudoscientific arguments and false information. But we can also mention the economic damage. I am thinking here, for example, of the attacks on 5G system towers in Western Europe
Nemeth T. Enikő said. Due to the emergence of fake news and pseudoscientific opinions in many areas of life, their identification is inevitable from a social and economic point of view, as well as from the point of view of maintaining the credibility of modern science. The linguists and computer scientists of SZTE, in consortium with the University of Debrecen, have received 200 million forints from the Hungarian Academy of Sciences for research and mapping of this problem in the Science Program for the Hungarian language. The four-year project began last fall, but after only a few months has yielded interesting results.
It is important to explain how to identify fake news and pseudoscientific texts
– Explained by T. Enikő Németh. A large team of linguists researches the characteristics of fake news, on the basis of which these articles and texts are easily recognizable. For example, there is a good chance that we will be dealing with fake news if the text contains all words in capital letters, a lot of exclamation marks, expressions affecting feelings, mixed words and phrases, or a lot of exclamation marks.
Program learns
“One of our students has already created an application. The program is an artificial neural network with 110 million variables,” said Tibor Sendes. According to a teacher at the Institute of Informatics at SZTE, a neural network is a biologically inspired and very complex simulation. The point is that the machine learns, and the cells communicate Artificial neural systems are layered with each other through different functions.So far, Szeged IT professionals have loaded 5,000 fake news and 5,000 real news into the system so that the machine can learn.
Based on the teaching examples provided, the program attempts to establish general patterns that allow it to distinguish between positive and negative examples as accurately as possible.
One cannot say exactly how the program identifies news with misinformation either. Given the complexity of the model used, it is difficult to assess at present whether the program makes its decisions based on the exact characteristics of the texts examined – which can also be interpreted from Before humans — but at the same time, the ability to recognize fake news and pseudoscientific texts is in nearly 90 percent of the news used in the assessment, said Berend Gabor, assistant professor, head of the five-person development team. By preparing rules manually, based on their own expertise and experience.It is a vain task for them to read hundreds of articles in order to create rules of proper quality.
We are trying to create a device that makes decisions similar to humans, but the decision-making mechanism is often very different
Gábor Berend added it. The developers’ goal is to eventually be able to explain the device’s decision-making processes.
Focus on rubber products
Although the developers are still hard at work on the fake news find site, many people in healthcare are already interested in the program. “The National Institute of Pharmacy and Food Hygiene (OGYÉI) formally carries out this work using human resources and manual control. If they spot signs of abuse, they should act ex officio,” said Tibor Sendes, adding that OGYÉI will also develop a search bot that will monitor various health websites around the clock. .
The SZTE teacher indicated that they planned to write a short text syllabus, in which they would point out what is worth paying attention to and how we can identify fake news.
The software is currently available in Hungarian, but according to our experience so far, it can be translated into other languages as well.
– said Gabor Berend. According to the plans, the mobile version of the fake news detector will be available for download on both Android and iOS platforms.
(cover photo: index)
