Sentence model statistics as means of corpus studies of attitudes (based on the material of online discussions)

Abstract

The paper considers the problem of choosing the parameters of linguostatistical studies when comparing corpora of different semantics. The proposition is to be proven that the set of such parameters should include the models of sentences (structural diagrams) as the basic units of thought expression. It sets the task of programming the sentence structural patterns which are presented in the academic grammars of the Russian language as a complete closed list. A brief overview of works on linguostatistics is offered from the point of view of the choice of text analysis units. We examine the key theses of the psychological school of L. S. Vygotsky — A. N. Leontiev on the structure of consciousness elements to distinguish the components of attitude — affective, cognitive and behavioral. Based on the material of online discussions, three corpora (collections) of statements have been compiled, each embodies one of those components. The belonging of statements to a particular collection was determined by three expert psychologists (if only their assessment completely coincided). The linguistic analysis includes 1360 utterances — 269 affective, 859 cognitive and 232 behavioral. Then, 7 sentence models were selected and programmed which semantics coincides with the content of any of the attitude components. This consistency ensured the interpretability of the work results. The hypothesis on the effectiveness of a comparative study of corpora using the “sentence model” parameter was verified. Qualitative analysis of the material showed that collections are effectively differentiated by semantic variants of polysemous structural patterns. Quantitative analysis, taking into account only the most abstract semantics of the sentence model at this stage of work, made it possible to differentiate the collections of behavioral and affective types, but did not solve the problem for the collection of cognitive-type utterances. Further improvement of the research methodology includes the analysis of structural and semantic variants of polysemous sentence models.