# US 2021/0342534 A1 · Sentence structure vectorization device, sentence structure vectorization method, and storage medium storing sentence structure vectorization program

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

- Number: US 2021/0342534 A1
- Document ID: US20210342534A1
- Status: Published application
- Kind: A1
- Application number: 17/378,543
- Filed: 2021-07-16
- Published: 2021-11-04
- Earliest claimed: 2019-01-31
- Assignees: Mitsubishi Electric Corporation
- Applicants: Mitsubishi Electric Corporation
- Inventors: Hiroyasu ITSUI
- Web page: https://patentgrep.com/patent/US20210342534A1

## CPC

- G06F 40/211: Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
- G06F 40/268: Morphological analysis
- G06F 16/3343: {using phonetics}
- G06F 16/3347: {using vector based model}
- G06F 16/353: into predefined classes
- G06F 40/205: Parsing
- G06F 40/279: Recognition of textual entities
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A sentence structure vectorization device includes processing circuitry to generate a plurality of morphemes by performing morphological analysis on an input sentence; to generate a dependence structure graph regarding the plurality of morphemes by performing dependency parsing on the plurality of morphemes; and to generate a sentence structure vector by extracting a plurality of pieces of partial structure information from the dependence structure graph and converting a morpheme string corresponding to the plurality of pieces of partial structure information into a numerical sequence.

## Claims

### Claim 1

A sentence structure vectorization device comprising:
processing circuitry
to generate a plurality of morphemes by performing morphological analysis on an input sentence;
to generate a dependence structure graph regarding the plurality of morphemes by performing dependency parsing on the plurality of morphemes; and
to generate a sentence structure vector by extracting a plurality of pieces of partial structure information from the dependence structure graph and converting a morpheme string corresponding to the plurality of pieces of partial structure information into a numerical sequence made up of a plurality of vector values, wherein
each of the plurality of vector values is the number of appearances of a morpheme structure number in each of a plurality of dimensions corresponding to the plurality of pieces of partial structure information,
the morpheme structure number is obtained based on morpheme structure information associating a dependency relationship between N morphemes where N is greater than or equal to 1, and a plurality of morpheme structure numbers, and
the number of the appearances is 1 in a case where the morpheme structure number in each of the plurality of dimensions exists in the morpheme structure information, and the number of the appearances is 0 or information on the number of the appearances is omitted in a case where the morpheme structure number in each of the plurality of dimensions does not exist in the morpheme structure information.

### Claim 2 (depends on 1)

The sentence structure vectorization device according to claim 1, wherein the processing circuitry
generates a plurality of pieces of morpheme information each including a pair consisting of a corresponding one of the plurality of morphemes and information indicating a dependency relationship obtained by the dependency parsing, and
extracts the plurality of pieces of partial structure information from the plurality of pieces of morpheme information.

### Claim 3 (depends on 1)

The sentence structure vectorization device according to claim 1, wherein
when N represents an integer greater than or equal to 3, the plurality of pieces of partial structure information includes structure information from 1-morpheme structure information regarding one morpheme to N-morpheme structure information regarding N morphemes, and
the processing circuitry generates the N-morpheme structure information by performing a process of merging (N−1)-morpheme structure information and 2-morpheme structure information together.

### Claim 4 (depends on 1)

The sentence structure vectorization device according to claim 1, further comprising a storage that stores the sentence structure vector.

### Claim 5 (depends on 1)

The sentence structure vectorization device according to claim 1, wherein the processing circuitry executes a process that uses the sentence structure vector.

### Claim 6 (depends on 5)

The sentence structure vectorization device according to claim 5, wherein the processing circuitry retrieves a document from retrieval object documents by using the sentence structure vector.

### Claim 7 (depends on 5)

The sentence structure vectorization device according to claim 5, wherein the processing circuitry executes classification of retrieval object documents by using the sentence structure vector.

### Claim 8

A sentence structure vectorization method executed by a computer, the method comprising:
generating a plurality of morphemes by performing morphological analysis on an input sentence;
generating a dependence structure graph regarding the plurality of morphemes by performing dependency parsing on the plurality of morphemes; and
generating a sentence structure vector by extracting a plurality of pieces of partial structure information from the dependence structure graph and converting a morpheme string corresponding to the plurality of pieces of partial structure information into a numerical sequence made up of a plurality of vector values, wherein
each of the plurality of vector values is the number of appearances of a morpheme structure number in each of a plurality of dimensions corresponding to the plurality of pieces of partial structure information,
the morpheme structure number is obtained based on morpheme structure information associating a dependency relationship between N morphemes where N is greater than or equal to 1, and a plurality of morpheme structure numbers, and
the number of the appearances is 1 in a case where the morpheme structure number in each of the plurality of dimensions exists in the morpheme structure information, and the number of the appearances is 0 or information on the number of the appearances is omitted in a case where the morpheme structure number in each of the plurality of dimensions does not exist in the morpheme structure information.

### Claim 9

A non-transitory computer-readable storage medium for storing a sentence structure vectorization program that causes a computer to execute
generating a plurality of morphemes by performing morphological analysis on an input sentence;
generating a dependence structure graph regarding the plurality of morphemes by performing dependency parsing on the plurality of morphemes; and
generating a sentence structure vector by extracting a plurality of pieces of partial structure information from the dependence structure graph and converting a morpheme string corresponding to the plurality of pieces of partial structure information into a numerical sequence made up of a plurality of vector values, wherein
each of the plurality of vector values is the number of appearances of a morpheme structure number in each of a plurality of dimensions corresponding to the plurality of pieces of partial structure information,
the morpheme structure number is obtained based on morpheme structure information associating a dependency relationship between N morphemes where N is greater than or equal to 1, and a plurality of morpheme structure numbers, and
the number of the appearances is 1 in a case where the morpheme structure number in each of the plurality of dimensions exists in the morpheme structure information, and the number of the appearances is 0 or information on the number of the appearances is omitted in a case where the morpheme structure number in each of the plurality of dimensions does not exist in the morpheme structure information.
