Scientists from Novosibirsk State University (NSU), in collaboration with colleagues from Kazakhstan, have developed an algorithm that allows finding destructive information on the Web. The search is carried out in Russian and Kazakh.
The analysis of Kazakh words is carried out on the basis of the morphological features of the language. The algorithm analyzes words, determines their initial form or forms other forms on its basis. This approach is important for the semantic analysis of large texts or searches on the Web.
We analyzed news in the Kazakhstani segment of the Internet in Kazakh and Russian, collecting an extensive data set. In particular, we investigated destructive information and learned to identify it by a number of characteristics.
The scientist explained that one of the signs of destructive information is the call to action. The specialists also classified information according to the reliability of the data and the conclusions made by the authors of the publications.
The study involved a corpus of Kazakhstani media. It included more than 4 million publications from 36 sources. More than 2 million texts from Russian media were also used.
Destructive information was determined using machine learning and artificial intelligence methods based on criteria such as the absence of verifiable facts, politicization, call to action, negative tone and manipulativeness.
NSU scientists, together with employees of the Kazakh National Technical University named after K. I. Satpayev, continue research. They are developing an ontology in the field of natural language processing in three languages. It will allow training large language models
Read also on the topic:
GigaChat neural network will be taught the Udmurt language
At MTUCI, a neural network will be taught to read damaged barcodes
A neural network for creating films and TV series was developed in Moscow
