METHOD OF CLASSIFIYING TEXTS RECEIVED AS RESULT OF SPEECH RECOGNITION Russian patent published in 2017 - IPC G06F17/27 

FIELD: physics.

SUBSTANCE: method of classifying the text includes the initial creation of a semantic dictionary in the form of a repository of semantic characteristics of words. Then the spoken language is recognized, and the text is received. In the result of the spoken language recognition, each word is selected in the text. A plurality of semantic characteristics in the semantic dictionary is found for each selected word. On the basis of the identified plurality of the semantic characteristics, the semantic consistency of, at least, one word sequence is determined to obtain a phrase. A plurality of phrases is selected from the text with a comparison of their semantic characteristics and the selection of the results of comparison of the dominant semantic characteristics. A plurality of phrases is converted into a plurality of key phrases containing a dominant semantic characteristic. A class is formed from the first received key phrases and their semantic characteristics.

EFFECT: increasing the accuracy of classifying the text files obtained as a result of recognizing speech in the telephone communication.

4 dwg, 1 tbl