A First Attempt to Detect Misinformation in Russia-Ukraine War News through Text Similarity

The paper focuses on misinformation detection in established global news outlets' texts covering significant and well-known events of the Russian-Ukraine war. We created the RUWA dataset and applied unsupervised ML approaches as the first dimension of misinformation detection. We consider several different aspects of semantic similarity identification of the articles from various regions in order to confirm the hypothesis that if the news covering the same event from the outlets of various regions over the world are similar enough it means they reflect each other or, instead, if they are completely divergent it means some of them are likely not trustworthy.