# US 10,503,828 B2 · System and method for answering natural language question

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

- Number: US 10,503,828 B2
- Document ID: US10503828B2
- Status: Granted
- Kind: B2
- Application number: 14/737,130
- Filed: 2015-06-11
- Published: 2016-05-19
- Granted: 2019-12-10
- Earliest claimed: 2014-11-19
- Assignees: Electronics and Telecommunications Research Institute
- Applicants: Electronics and Telecommunications Research Institute
- Inventors: Yong Jin Bae; Hyun Ki Kim; Pum Mo Ryu; Hyo Jung Oh; Chung Hee Lee; Soo Jong Lim; Joon Ho Lim; Myung Gil Jang; Mi Ran Choi; Jeong Heo
- Web page: https://patentgrep.com/patent/US10503828B2

## CPC

- G06F 40/211: Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
- G06F 16/3329: Natural language query formulation
- G06F 16/3332: Query translation
- G06F 16/36: Creation of semantic tools, e.g. ontology or thesauri
- G06F 40/30: Semantic analysis
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Provided are a system and method for answering a natural language question which show improved information retrieval performance. The system includes an index unit configured to analyze text of previously stored irregular documents and classify and index the irregular documents according to meanings of sentences or paragraphs, a database configured to receive and store the irregular documents indexed according to the meanings and transmitted from the index unit, a retrieval unit configured to extract an index word by semantically analyzing an input question and search the database for documents related to the extracted index word, and a provision unit configured to generate a correct answer to the question by analyzing the documents searched by the retrieval unit and provide the search results and the correct answer.

## Claims

### Claim 1

A system for answering a natural language question, the system comprising:
one or more processors that process computer executable program code embodied in non-transitory computer readable storage media,
wherein when the program code is executed by computer the program code causes the processor to perform the steps of:
analyzing text of irregular documents stored in a storage unit, classifying and indexing the irregular documents according to meanings of sentences or paragraphs, and stores in a database the irregular documents indexed according to the meanings by the index unit, wherein the sentences or the paragraphs in the irregular documents are classified based on structural information or results of understanding a natural language;
extracting an index word by semantically analyzing an input question, and searches the database for documents related to the extracted index word; and
generating a correct answer to the question by analyzing the documents searched by the retrieval unit, and provides results of the search and the generated correct answer,
wherein analyzing the text of the stored irregular documents comprises analyzing by a document analyzer the text of the stored irregular documents;
wherein the classifying comprises receiving by a semantic analyzer text analyzed by the document analyzer and classifying the meanings of the received text in units of sentences or paragraphs by a semantic classifier; and
wherein the indexing comprises indexing by a document indexer sentences or paragraphs classified by the semantic classifier according to the meanings in units of morphemes, entity names, phrases, syntax structures, semantic structures, sentence structures, “subject-verb” structures, “object-verb” structures, and “subject-verb-object” structures.

### Claim 2 (depends on 1)

The system of claim 1, wherein the database comprises a plurality of index databases classified according to indices.

### Claim 3 (depends on 2)

The system of claim 2, wherein the processor searches for documents in an index database corresponding to the index word among the plurality of index databases.

### Claim 4 (depends on 2)

The system of claim 2, wherein the processor extracts a plurality of index words by analyzing the question.

### Claim 5 (depends on 4)

The system of claim 4, wherein the processor searches for documents in respective index databases corresponding to the plurality of index words.

### Claim 6 (depends on 1)

The system of claim 1, wherein the processor performs morpheme analysis, lexical analysis, syntax analysis, and sentence structure analysis on the text.

### Claim 7 (depends on 1)

The system of claim 1, wherein the processor classifies the meanings by extracting sentence features and generating patterns or by using a machine learning technique.

### Claim 8 (depends on 1)

The system of claim 1, wherein the processor receives the question from an outside of the system;
analyzes the question input through the question input portion;
receives the question analyzed by the question analyzer, classifying a meaning, and extracts the index word; and
searches the database for documents related to the index word extracted by the question classifier.

### Claim 9 (depends on 1)

The system of claim 1, wherein, when the database is searched with a plurality of index words, the processor provides results of the search according to weights previously given to the respective index words.

### Claim 10

A method of answering a natural language question, the method comprising:
processing computer executable program code embodied in non-transitory computer readable storage media by one or more processors, the processing comprising:
analyzing previously stored irregular documents, and classifying and indexing the irregular documents according to meanings of sentences or paragraphs in the irregular documents, wherein the sentences or the paragraphs in the irregular documents are classified based on structural information or results of understanding a natural language;
transmitting the irregular documents indexed according to the meanings to a database and storing the indexed irregular documents in the database;
when a question is input, analyzing the question to extract an index word and searching the database for documents related to the extracted index word; and
analyzing the searched documents to generate a correct answer to the question, and providing results of the search and the generated correct answer,
wherein the classifying and indexing the irregular documents comprises:
analyzing by a document analyzer text of the previously stored irregular documents;
receiving by a semantic classifier the analyzed text and classifying the meanings of the received text in units of sentences or paragraphs; and
indexing by a document indexer the sentences or the paragraphs classified according to the meanings in units of morphemes, entity names, phrases, syntax structures, semantic structures, “subject-verb” structures, “object-verb” structures, and “subject-verb-object” structures.

### Claim 11 (depends on 10)

The method of claim 10, wherein the sentences or the paragraphs in the irregular documents are classified based on structural information or results of understanding a natural language.

### Claim 12 (depends on 10)

The method of claim 10, wherein the classifying and indexing the irregular documents comprises classifying one sentence or paragraph in the documents into two or more meanings.

### Claim 13 (depends on 10)

The method of claim 10, wherein the analyzing the text of the previously stored irregular documents comprises morpheme analysis, lexical analysis, syntax analysis, and sentence structure analysis on the text.

### Claim 14 (depends on 10)

The method of claim 10, wherein the searching the database for documents related to the extracted index word comprises searching for documents in an index database corresponding to the index word among a plurality of index databases classified according to indices.

### Claim 15 (depends on 10)

The method of claim 10, wherein the classifying the meanings of the received text in units of sentences or paragraphs comprises classifying the meanings by extracting sentence features and generating patterns or using a machine learning technique.

### Claim 16 (depends on 10)

The method of claim 10, wherein the providing the results of the search and the generated correct answer comprises, when the database is searched with a plurality of index words, comprises providing results of the search according to weights previously given to the respective index words.