The article is aimed at comparing the lists of keywords, obtained using two different reference corpora of the same size (1 million words). The target corpus is a lemma-tized collection of texts of internal documents such as «The informed voluntary consent to medical intervention)) (38,5 thousand words). The reference corpora are the off-line disambiguated subcorpus of the RNC (ruscorpora.ru) and the off-line disambiguated subcorpus of the GICR (webcorpora.ru). The study showed that the «tops» of two lists of positive keywords are very similar, and the keywords themselves indicate primarily «aboutness» of the texts.