Statement of the problem. The paper discusses multiple deviations of science texts in the English language written by Russian novice writers from the norms and expectations of the international discourse community. The purpose of the article is to propose a concept of utilizing computer linguistics tools of automated text analysis for pedagogical purposes to identify real needs of students in science writing in English. The methodology is based on the contrastive discourse-analysis of two corpora: students’ research paper manuscripts and published research papers by international expert writers. Both corpora belong to one engineering discipline and were compiled specifically for the research purposes. The contrastive corpus analysis was conducted by means of a computer linguistics tool Gramulator. Research results. The quantitative and qualitative analysis of differential bigrams in Gramulator demonstrated that the writing by novice Russian authors significantly differs from the expert writing. The considered bigrams indicated lower lexical diversity; insufficient skills in using grammar that is absent in the Russian language; tendency to underuse predicative clauses and hedging devices; tendency to overuse words and phrases typical of Russian scientific style, as well as referencing and specifying words and phrases. Conclusions. The proposed concept is aimed to make the academic writing process more student-centered. The identification of common linguistic problems in the novice researchers’ writing may raise students’ awareness of the differences in the rhetorical choice in the two languages, improve strategic skills in making the proper choice when writing their own texts, and, thus, approach to the norms and expectations of the target discourse community.