NEURAL NETWORK FOR INTERPRETING NATURAL LANGUAGE SENTENCES Russian patent published in 2019 - IPC G06F17/28 G06N3/02 

FIELD: computer equipment.

SUBSTANCE: invention relates to computer engineering and can be used for interpreting natural language sentences, including context-thematic machine translation. Device consists of layers of neurons, wherein first layer of neurons is intended for interpretation of values of separate words in sentence, intermediate layers of neurons are intended for interpretation of collocations, number of words in which corresponds to number of layer of neurons in neural network. Last layer of neurons is represented by an output neuron and is intended for interpreting a sentence based on words values interpreted on previous layers of neurons. Each neuron of the input and intermediate layers has a memory cell designed to store, respectively, the values of the word or word combination, wherein each such value corresponds to a binary feature, the encoding of which is described by the following rule: "1" is the main meaning of a word or phrase, "0" is an additional value of a word or phrase.

EFFECT: technical result is higher quality of interpreting sentences in natural language.

8 cl, 1 dwg