The article deals with the questions of digitization, markup and software support in solving the problem of attribution of texts. Effective solution of the actual problems of analyzing textual works, including searching for borrowings and determining authorship, requires combining both the work of a linguistic expert and the use of various mathematical methods on large amounts of data. Currently, one of the actual tasks is the analysis of 500 non-attributive publicistic articles from the magazines "Time", "Epoch" and the weekly "Citizen" (1861—1874). It is known that F. M. Dostoevsky edited and headed these magazines; therefore, research on the belonging of these works to his pen has been conducted for a long time. The work shows that modern information systems for attribution support do not allow to carry out such studies, so the staff of Petrozavodsk State University are working on the modernization of the "SMALT" software complex for implementing the text attribution task tool and the "Folklore" information system for automated construction and analysis of graph-theoretic models of texts, which will bring the research to a new level. The paper describing the architecture of the "SMALT", the modular structure of "Folklore" and a mathematical model for defining grammatical features in case of homonymy.