The automatic analysis and detection of fake news are of high research value and great practical significance.By building a corpus of true and fake news,the differences of Russian fake news at the level of word characteristics can be investigated.The results show that fake news tends to use multiple words,short words,simple words,and monosyllabic words to convey information,and the text difficulty is relatively low.The word frequency distribution conforms to Zipf's law,but the goodness of fit between real and fake news is not quite different.The frequency of nouns,numbers and proper nouns that convey conclusive factual information in fake news is lower than that in real news.The use of binary collocations is more than that of multiple collocations,and the economic characteristics are more significant.The language tends to be simplified and is more inclined to express ideas and intentions than to state existing facts.The 5 overall features,13 diversity features,11 distribution features,5 collocation features,and 17 lexical feature indicators in fake news are statistically different from those in real news.The use of the above indicators for the automatic clustering of real and fake news works well.
NETL
NSTL
万方数据
