The Method of Relational-Situational Analysis of Text in Psychological Research

Abstract

Some artificial intelligence methods for reliable identification of personality traits in speech can be useful for upgrade of psycho-diagnostics. Such tools should be developed by psychologists, linguists and computer scientists jointly, and therefore a preliminary special methodology and organization are required. The described in our paper original tool for analyzing written text was created at the Artificial Intelligence Research Institute of National Institute for Research in Computer Science and Control of RAS under the guidance of Dr. of Physical and Mathematical Sciences G.S. Osipov in collaboration with the members of Mental Health Research Center, the Institute of the Russian Language of the Russian Academy of Sciences and the Perm State University. G.S. Osipov proposed a method of relational-situational analysis of text (RSA) based on the ideas of the communicative grammar of the Russian language by G.A. Zolotova, and this approach is implemented in the tool. Analysis of the text using the tool “RSA Machine” turns it into a set of predicate-role structures that are associated with the specifics of the author's worldview. The RSA machine identifies 127 semantic features and semantic roles, and includes a set of psycholinguistic indicators and data of specially created groups of thematic lexis, extracts a total of 197 indicators from texts. The empirical studies have been conducted and have shown some singularities of texts written by people with certain psychological characteristics (according to psychodiagnostic methods and expert opinion), as well as with a psychiatric status (schizophrenia and clinical depression), and these results can be utilized for diagnosis and monitoring. By using the data that the RSA machine receives for further machine learning, it can adapt and develop according with specific research tasks.