THEMATIC MODELS WITH A PRIORI TONALITY PARAMETERS BASED ON DISTRIBUTED REPRESENTATIONS

摘要

FIELD: data processing.;SUBSTANCE: invention relates to means for thematic modelling with a priori tone parameters based on distributed representations. Text document is inserted into a thematic model and a presentation for each word in the text document is determined by the thematic model, wherein the representations are word vectors in the semantic space. Assessing presentations using a priori tone parameters to determine a theme corresponding to said text document, wherein the topic model comprises a priori tonality parameters, trained based on representations distributed using a regularizer, which sets the same tonality to words having similar word vectors, and wherein each a priori tonality parameter is the same for words having similar word vectors.;EFFECT: technical result consists in detecting a greater number of aspect-oriented tonal words and further improved classification.;8 cl, 5 dwg, 9 tbl