QUESTION ANSWERING SYSTEM AND METHOD

• US 20150324456A1
• Filed: 01/22/2015
• Published: 11/12/2015
• Est. Priority Date: 05/08/2014
• Status: Active Grant

• Current Assignee: Viralpick Company Limited
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: KIM, Young Rae, LEE, Hyung Jik, MOON, Jin Young, BAE, Chang Seok, KIM, Hyun Ki
• Granted Patent: US 10,025,849 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3331 Query processing
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/93 Document management systems
• CPC Class Codes: G06N 5/00 Computing arrangements using knowledge-based models
• CPC Class Codes: G06N 5/04 Inference or reasoning models

Abstract

Provided is a question answering system with respect to a natural language question and a method thereof. The question answering system includes a candidate answer generating unit configured to extract a document mapped to an input natural language question, and generate candidate answers with respect to the natural language question from the extracted document, a text entailment recognizing unit configured to generate a text entailment recognition result representing a degree of association between multiple evidence sentences including the generated candidate answers and the natural language question, a list generating unit configured to generate a candidate answer list including the multiple evidence sentences in high association degree order on the basis of the text entailment recognition result, and an output unit configured to output the generated candidate answer list as a search result with respect to the natural language question.

Claims

1. A question answering system comprising:

an input unit configured to receive a natural language question in at least one of a sound form and a text form;

a candidate answer generating unit configured to extract a document mapped to the input natural language question, and generate candidate answers with respect to the natural language question from the extracted document;

a text entailment recognizing unit configured to generate a text entailment recognition result representing a degree of association between multiple evidence sentences including the generated candidate answers and the natural language question;

a list generating unit configured to generate a candidate answer list including the multiple evidence sentences in high association degree order according to the text entailment recognition result; and

an output unit configured to output the generated candidate answer list as a search result with respect to the natural language question.

2. The question answering system of claim 1, wherein the text entailment recognizing unit comprises:

a word entailment recognizing unit configured to recognize word entailment between the multiple evidence sentences and the natural language question by using at least one of a distance between lexical networks and word similarity, stored previously;

a sentence entailment recognizing unit configured to recognize sentence entailment between the multiple evidence sentences and the natural language question on the basis of similarity of a syntax structure and pattern similarity; and

a context entailment recognizing unit configured to recognize context entailment between the multiple evidence sentences and the natural language question by using a case frame.

3. The question answering system of claim 1, wherein the list generating unit comprises:

a basic form list generating unit configured to generate a word type basic form candidate answer list by combining the candidate answers and the multiple evidence sentences on the basis of similarity of words;

a definition form list generating unit configured to generate a sentence type definition form candidate answer list by combining the candidate answers and the multiple evidence sentences on the basis of at least any one of the similarity of a syntax structure, the pattern similarity, and the case frame; and

a parade form list generating unit configured to generate a word parade type parade form candidate answer list by combining the candidate answers and the multiple evidence sentences.

4. The question answering system of claim 3, wherein the definition form list generating unit statistically and semantically extracts common order of the text lists through entailment recognition of text lists and combines the extracted text lists.

5. The question answering system of claim 1, wherein the output unit outputs an evidence sentence corresponding to a candidate answer selected by a user from the candidate answer list.

6. The question answering system of claim 1, wherein the number of candidate answer list is determined according to a size of a display screen.

7. A question answering method comprising:

generating candidate answers with respect to a natural language question;

generating a text entailment recognition result representing a degree of association between multiple evidence sentences including the generated candidate answers and the natural language question;

generating a candidate answer list including the multiple evidence sentences in high association degree order according to the text entailment recognition result; and

outputting the generated candidate answer list as a search result with respect to the natural language question.

8. The question answering method of claim 7, wherein the generating of candidate answers with respect to the natural language question comprises:

receiving the natural language question;

performing entailment recognition on the input natural language question;

extending a meaning of the natural language question through the entailment recognition result;

extracting a document mapped to the extended natural language question; and

generating candidate answers with respect to the natural language question from the extracted document.

9. The question answering method of claim 7, wherein the generating of an entailment recognition result comprises:

recognizing word entailment between the multiple evidence sentences and the natural language question by using at least one of a distance between lexical networks and word similarity;

recognizing sentence entailment between the multiple evidence sentences and the natural language question on the basis of at least one of similarity of a syntax structure and pattern similarity; and

recognizing context entailment between the multiple evidence sentences and the natural language question on the basis of a case frame.

10. The question answering method of claim 7, wherein the generating of a candidate answer list comprises:

generating a word type basic form candidate answer list by combining the candidate answers and the multiple evidence sentences on the basis of similarity of words;

generating a sentence type definition form candidate answer list by combining the candidate answers and the multiple evidence sentences on the basis of at least any one of the similarity of a syntax structure, the pattern similarity, and the case frame; and

generating a word parade type parade form candidate answer list by combining the candidate answers and the multiple evidence sentences.

11. The question answering method of claim 10, wherein the generating of a definition form candidate answer list comprises, when a definition form candidate answer list in which the generated candidate answers are in a text list form is generated, statistically extracting common order of the text lists through entailment recognition of the text lists and displaying the extracted text list.

12. The question answering method of claim 7, wherein the outputting of a generated candidate answer list as a search result with respect to the natural language question comprises, when a certain answer is selected from the candidate answer list, outputting an evidence sentence corresponding to the selected certain answer.

13. The question answering method of claim 8, wherein the outputting of a generated candidate answer list as a search result with respect to the natural language question comprises adjusting the number of candidate answer list according to a size of a display screen.