Question answering system, data search method, and computer program

• US 7,844,598 B2
• Filed: 09/22/2005
• Issued: 11/30/2010
• Est. Priority Date: 03/14/2005
• Status: Expired due to Fees

• 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
• Primary Examiner(s): Jalil; Neveen Abel
• Assistant Examiner(s): Chbouki; Tarek
• Application Number:
  US11/231,845
• Publication Number:
  US 20060204945A1
• Time in Patent Office: 1,895 Days
• Field of Search: 707/713
• US Class Current: 707/713
• CPC Class Codes: G09B 7/02 of the type wherein the student is expected to construct an answer to the question which is presented or wherein the machine gives an answer to the question presented by a student

Abstract

A question answering system includes a question dividing unit, a question answering unit, and an answer determination unit. The question dividing unit applies a predetermined question dividing rule to an input question, to divide the input question into sub-questions. The question answering unit acquires answer candidates corresponding to the sub-questions generated by the question dividing unit, together with certainty values serving as reliability indexes. The answer determination unit selects an answer to the input question from among the answer candidates acquired by the question answering unit.

Claims

1. A question answering system having a processor, comprising:

a question dividing unit that applies a predetermined question dividing rule to an input question, to divide the input question into sub-questions;

a question answering unit that acquires a plurality of answer candidates corresponding to each sub-question generated by the question dividing unit, and acquires certainty values of each answer candidate corresponding to each sub-question generated by the question dividing unit, the certainty values of the answer candidates serving as reliability indexes; and

an answer determination unit that identifies common answer candidates provided by the question answering unit, calculates a sum of the certainly values corresponding to each common answer candidate, and selects one answer candidate having the largest sum of the certainty values from among the plurality of answer candidates acquired for the sub-questions, as an answer to the input question, wherein

the processor controls the question dividing unit, the question answering unit, and the answer determination unit, and

the question dividing unit references the answer candidates acquired by the question answering unit, and further divides the input question if no answer candidate having a certainty value exceeding a given threshold value is found.

2. The question answering system according to claim 1, further comprising a syntactic and semantic analysis unit that execute a syntactic and semantic analysis process on the input question,

wherein the question dividing unit applies the predetermined question dividing rule to a result of the syntactic and semantic analysis of the input question generated by the syntactic and semantic analysis unit.

3. The question answering system according to claim 1, wherein the answer determination unit selects an answer candidate having the largest sum of the certainty values as the answer to the input question from among answer candidates commonly appearing in the answer candidates acquired for the sub-questions.

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

the question dividing unit applies the predetermined question dividing rule to the result of the syntactic and semantic analysis of the input question to divide the sub-questions into a plurality of search patterns, and

the question answering unit executes a search process based on the plurality of search patterns.

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

a unit that replaces an element of another sub-question with an answer candidate acquired for one sub-question of the sub-questions generated by the question dividing unit, to generate a transformational sub-question, the unit acquiring an answer to the transformational sub-question.

6. A question answering system according to claim 1, wherein the certainty values are based on consistency with a question type.

7. A data search method comprising:

applying a predetermined question dividing rule to an input question, to divide the input question into sub-questions;

acquiring a plurality of answer candidates corresponding to each sub-question generated by the question dividing rule, and acquires certainty values of each answer candidate corresponding to each sub-question generated by the question dividing rule, the certainty values of the answer candidates serving as reliability indexes; and

identifying common answer candidates, calculating a sum of the certainty values corresponding to each common answer candidate, and selecting one answer candidate having the largest sum of the certainty values from among the plurality of answer candidates acquired for the sub-questions, as an answer to the input question, wherein

the acquiring comprises referencing the answer candidates acquired, and further dividing the input question if no answer candidate having a certainty value exceeding a given threshold value is found, and

a processor implements the data search method.

8. The data search method according to claim 7, further comprising:

executing a syntactic and semantic analysis process on an input question;

wherein the predetermined question dividing rule is applied to a result of the syntactic and semantic analysis of the input question generated by the syntactic and semantic analysis step.

9. The method according to claim 7, wherein:

the applying comprises applying the predetermined question dividing rule to the result of the syntactic and semantic analysis of the input question to divide the input question into a plurality of search patterns, and

the acquiring comprises executing a search process based on the plurality of search patterns.

10. The method according to claim 7, further comprising:

replacing an element of another sub-question with an answer candidate acquired for one sub-question of the sub-questions, to generate a transformational sub-question; and

acquiring an answer to the transformational sub-question.

11. The method according to claim 7, wherein the selecting comprises selecting an answer candidate having the largest sum of the certainty values as the answer to the input question from among answer candidates commonly appearing in the answer candidates acquired for the sub-questions.

12. A data search method according to claim 7, wherein the certainty values are based on consistency with a question type.

13. A computer-readable storage medium storing a computer program stored in a computer readable storage medium, the computer program causing a computer to execute a data search process comprising:

applying a predetermined question dividing rule to an input question, to divide the input question into a plurality of sub-questions;

acquiring a plurality of answer candidates corresponding to each sub-question generated by the question dividing rule, and acquiring certainty values of each answer candidate corresponding to each sub-question generated by the question dividing rule, the certainty values of the answer candidates serving as reliability indexes; and

identifying common answer candidates, calculating a sum of the certainty values corresponding to each common answer candidate, and selecting one answer candidate having the largest sum of the certainty values with respect to respective sub-questions from among the plurality of answer candidates acquired for the sub-questions, as an answer to the input question, wherein

the acquiring comprises referencing the answer candidates acquired, and further dividing the input question short if no answer candidate having a certainty value exceeding a given threshold value is found.

14. A computer-readable storage medium according to claim 13, wherein the certainty values are based on consistency with a question type.