Our technology

Our solution is based on a neural network classifier that helps determine whether particular information is fake or real. The Kaggle Fake and Real News Dataset databases were chosen to train the neural network classifier. Our database has also been expanded using modern articles about Ukraine in English sources. A recurrent neural network, the Long short-term memory (LSTM) type was chosen as a classifier. These types of neural networks are designed to build text classifiers. LSTM is a multilayer neural network that contains 11 million parameters. We took 35,000 articles and labels from the database to train our neural network, with information about which class they belong to, fake news or true. The neural network showed a classification accuracy of 98% on the training sample.