Question answering system, data search method, and computer program

• US 20060206472A1
• Filed: 09/22/2005
• Published: 09/14/2006
• Est. Priority Date: 03/14/2005
• Status: Active Grant

• Current Assignee: Fujifilm Business Innovation Corp. (Fujifilm Holdings Corporation)
• Original Assignee: Fuji Xerox Company Limited (Xerox Holdings Corp.)
• Inventors: Masuichi, Hiroshi, Yoshimura, Hiroki, Sugihara, Daigo, Ohkuma, Tomoko
• Granted Patent: US 7,461,047 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 16/3338 Query expansion
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 40/30 Semantic analysis

Abstract

A question answering system includes a similar-question generation unit, a first search unit, a second search unit, and an answer selection unit. The similar-question generation unit selects a feature word from an input question to generate a transformational question in which the feature word is replaced with another word. The first search unit executes a first search process based on the transformational question. The second search unit selects a word included in a first search result acquired as an extension word and executes a second search process with using the selected extension word. The answer selection unit analyzes a correspondence relation among extension words included in a second search result, determines the feature word and a constructive part of an answer equivalent word in the second search result based on an analysis result, and selects a word determined to be an answer.

Claims

1. A question answering system comprising:

a similar-question generation unit that selects a feature word from an input question to generate a transformational question in which the feature word is replaced with another word;

a first search unit that executes a first search process based on the transformational question generated by the similar-question generation unit;

a second search unit that selects a word included in a first search result acquired by executing the first search process as an extension word and executes a second search process with using the selected extension word; and

an answer selection unit that analyzes a correspondence relation among extension words included in a second search result acquired by executing the second search process, determines the feature word and a constructive part of an answer equivalent word in the second search result based on an analysis result, and selects a word determined to be an answer.

2. The question answering system according to claim 1, wherein the answer selection unit analyzes the correspondence relation among the extension words based on HTML format data, which is acquired as the second search result acquired, determines the feature word and the constructive part of the answer equivalent word in the second search result based on the analysis result, and selects the word determined to be the answer.

3. The question answering system according to claim 1 wherein:

the similar-question generation unit generates a transformational question pattern 1, which is a similar question to the input question, without changing the feature word making up the input question,

the first search unit executes a search process based on the transformational question pattern 1, and

if no answer is acquired as a result of the search process based on the transformational question pattern 1,

the similar-question generation unit generates as the transformational question a transformational question pattern 2 in which the feature word making up the input question is replaced with a same-group word belonging to the same group in a thesaurus as the feature word;

the first search unit executes the first search process based on the transformational question pattern 2; and

the second search unit selects the word included in the first search result acquired by executing the first search process as the extension word and executes the second search process with using the selected extension word.

4. The question answering system according to claim 1, wherein:

the second search unit selects (a) a same-group word, which is selected from words included in the first search result and belongs to the same group as the feature word, and (b) an answer equivalent word in the transformational question as the extension words, and

the second search unit executes the second search process with using the selected extension words.

5. The question answering system according to claim 1, further comprising:

a first storage unit that stores question patterns; and

a second storage unit that stores transformational patterns of the question patterns stored in the first storage unit, wherein;

the similar-question generation unit compares the input question with the question patterns stored in the first storage unit to select a question pattern corresponding to the input question, and

the similar-question generation unit generates the transformational question corresponding to the selected question pattern, based on the transformational patterns stored in the second storage unit.

6. The question answering system according to claim 1, further comprising:

a third storage unit that stores group classification information of feature words, which are acquired from input questions as proper nouns, wherein;

the similar-question generation unit generates the transformational question in which the feature word making up the input question with a same-group word belonging to the same group as the feature word, with using the classification information stored in the third storage unit.

7. The question answering system according to claim 1, wherein the similar-question generation unit generates the transformational question in which the feature word is replaced with the other word, based on a predetermined rule.

8. The question answering system according to claim 1 further comprising:

a syntactic and semantic analysis unit that executes syntactic and semantic analysis on the input question, wherein;

the similar-question generation unit generates a question pattern based on a analysis result acquired by executing the syntactic and semantic analysis, and

the first search unit executes a search process based on the question pattern generated by the similar-question generation unit.

9. A data search method comprising:

selecting a feature word from an input question;

generating a transformational question in which the feature word is replaced with another word;

executing a first search process based on the transformational question generated;

selecting a word included in a first search result acquired by executing the first search process as an extension word;

executing a second search process with using the selected extension word;

analyzing a correspondence relation among extension words included in a second search result acquired by executing the second search process;

determining the feature word and a constructive part of an answer equivalent word in the second search result based on an analysis result; and

selecting a word determined to be an answer.

10. The data search method according to claim 9, wherein the analyzing comprises analyzing the correspondence relation among the extension words based on HTML format data, which is acquired as the second search result acquired.

11. The data search method according to claim 9, further comprising:

generating a transformational question pattern 1, which is a similar question to the input question, without changing the feature word making up the input question,

executing a search process based on the transformational question pattern 1, and

if no answer is acquired as a result of the search process based on the transformational question pattern 1,

generating as the transformational question a transformational question pattern 2 in which the feature word making up the input question is replaced with a same-group word belonging to the same group as the feature word;

executing the first search process based on the transformational question pattern 2;

selecting the word included in the first search result acquired by executing the first search process as the extension word; and

executing the second search process with using the selected extension word.

12. The data search method according to claim 9, wherein the selecting of the word included in the first search result as the extension word comprises selecting (a) a same-group word, which is selected from words included in the first search result and belongs to the same group as the feature word, and (b) an answer equivalent word in the transformational question as the extension words.

13. The data search method according to claim 9 wherein:

the generating of the transformational question comprises;

comparing the input question with question patterns stored in a first storage unit to select a question pattern corresponding to the input question, and

generating the transformational question corresponding to the selected question pattern, based on transformational patterns stored in a second storage unit.

14. The data search method according to claim 9, wherein the generating of the transformational question comprises generating the transformation question in which the feature word making up the input question with a same-group word belonging to the same group as the feature word, with using classification information stored in a third storage unit.

15. The data search method according to claim 9, wherein the generating of the transformational question comprises generating the transformation question in which the feature word is replaced with the other word, based on a predetermined rule

16. The data search method according to claim 9, further comprising:

executing syntactic and semantic analysis on the input question;

generating a question pattern based on a analysis result acquired by executing the syntactic and semantic analysis; and

executing a search process based on the question pattern generated.

17. A computer program stored in a computer readable medium, the computer program causing a computer to execute a data search process comprising:

selecting a feature word from an input question;

generating a transformational question in which the feature word is replaced with another word;

executing a first search process based on the transformational question generated;

selecting a word included in a first search result acquired by executing the first search process as an extension word;

executing a second search process with using the selected extension word;

analyzing a correspondence relation among extension words included in a second search result acquired by executing the second search process;

determining the feature word and a constructive part of an answer equivalent word in the second search result based on an analysis result; and

selecting a word determined to be an answer.