# US 2021/0334470 A1 · Neural network for interpreting sentences of a natural language

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0334470 A1
- Document ID: US20210334470A1
- Status: Published application
- Kind: A1
- Application number: 17/294,550
- Filed: 2019-11-14
- Published: 2021-10-28
- Earliest claimed: 2018-11-19
- Assignees: Obshchestvo S Ogranichennoj Otvetsvetvennostu "Invek"
- Applicants: Obshchestvo S Ogranichennoj Otvetsvetvennostu "Invek"
- Inventors: Mikhail Nikolaevich MOLIN; Georgij Mikhajlovich PERSKIJ
- Web page: https://patentgrep.com/patent/US20210334470A1

## CPC

- G06F 40/30: Semantic analysis
- G06F 40/284: Lexical analysis, e.g. tokenisation or collocates
- G06F 40/289: Phrasal analysis, e.g. finite state techniques or chunking
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
- G06N 3/0499: Feedforward networks
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- G06N 3/042: Knowledge-based neural networks; Logical representations of neural networks
- G06N 5/02: Knowledge representation; Symbolic representation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

The invention relates to computer systems and is intended for interpreting sentences in a natural language. A neural network consists of layers of neurons. A first neuron layer is intended for interpreting the meanings of individual words in a sentence. Intermediate neuron layers are intended for interpreting phrases in a sentence, in which the number of words in a phrase corresponds to the number of the neuron layer. The last neuron layer is represented by an output neuron and is intended for interpreting a sentence on the basis of the lexical meanings interpreted in the preceding neuron layers. Each neuron of the input and intermediate layers comprises a memory cell intended for storing the meanings of a word or a phrase. Moreover, in order to improve the quality of interpretation, each such meaning is assigned a binary feature, the coding of which is described according to the rule whereby “1” is the primary meaning of a word or phrase and “0” is the secondary meaning of the word or word group.

## Claims

### Claim 1

A neural network for the interpretation of natural language sentences, consisting of layers of neurons, wherein the first neuron layer is configured to interpret the meanings of individual words of a sentence, wherein the intermediate neuron layers are configured to interpret sentence phrases, the number of words in which corresponds to the number of a neuron layer in the neural network, and wherein the last layer of neurons is represented by an output neuron and configured to interpret the sentence based on the interpreted lexical meanings in the previous neuron layers, wherein each neuron of the input and intermediate layers contains a memory cell configured to store the meanings of a word or a phrase, respectively, wherein each such meaning corresponds to a binary feature, the coding of which is described by the following rule: “1” represents the basic meaning of the word or phrase, “0” represents an additional meaning of the word or phrase.

### Claim 2 (depends on 1)

The neural network according to claim 1, wherein the criterion for determining the binary sign of the meaning of a word or a phrase is the greatest statistical indicator of their use in large amounts of information.

### Claim 3 (depends on 1)

The neural network according to claim 1, wherein the criterion for determining a binary feature of the meaning of a word or a phrase represents an assessment made by an expert, when training the neural network taking into account the context of the use of the word or phrase.

### Claim 4 (depends on 1)

The neural network according to claim 1, wherein each neuron of the input layer and the intermediate layers corresponds to a target function of assigning a basic or additional meaning of a word or a phrase for the user and transferring the basic meaning to the output neuron.

### Claim 5 (depends on 4)

The neural network according to claim 4, wherein the execution of the target function of the input or intermediate layer neuron is performed upon activating such neuron.

### Claim 6 (depends on 5)

The neural network according to claim 5, wherein the condition for activating the neuron of the input or intermediate layers is to match the input word or phrase, respectively, with a word or a phrase, the meanings of which are recorded in the neuron memory cell.

### Claim 7 (depends on 1)

The neural network according to claim 1, wherein the target function of the output neuron is to add the meanings of words and phrases, the meanings of which are interpreted in the input and intermediate neuron layers.

### Claim 8 (depends on 1)

The neural network according to claim 1, which is implemented on a user-programmable gate array.
