Lexical and Grammatical Features of Russian-Language Tweets in Comparison with Everyday Spoken Russian

This study examines the features of computer-mediated discourse, often perceived as neither purely written nor spoken. Twitter discourse serves as a case in point, reflecting attributes of both spoken and written language. The aim of the study is to discern how closely Russian-language Twitter discourse mirrors everyday spoken Russian. A dataset of 152,223 Russian-language tweets (over 2 million tokens) was examined and juxtaposed against transcripts from the ORD speech corpus, which captures 508 macro episodes of daily conversation, totaling just under 900,000 tokens. Both lexical and grammatical aspects of the tweets and the spoken episodes are analyzed. A detailed comparison of unigrams, discourse words, and pragmatic markers is undertaken, supplemented by a multidimensional analysis spanning 22 grammatical features. The findings indicate that while the lexical attributes of Russian Twitter discourse closely align with spoken Russian, its grammatical features differ. Notably, both the tweets and the speech episodes share a significant overlap in lemmas, discourse words and pragmatic markers. However, when viewed grammatically, the Twitter discourse diverges from spontaneous spoken language. These insights hold potential for refining computer-mediated discourse generation systems for the Russian language.