METHOD FOR PROCESSING NATURAL LANGUAGE QUESTIONS AND APPARATUS THEREOF

• US 20100299139A1
• Filed: 04/23/2010
• Published: 11/25/2010
• Est. Priority Date: 04/23/2009
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: MA, Li, Ferrucci, David Angelo, Welty, Christopher, Wang, Chen, Pan, Yue, Zhang, Lei, Qiu, Zhao Ming
• Granted Patent: US 8,301,438 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems

Abstract

A method and an apparatus for selecting an answer to a natural language question. The method includes: detecting a named entity in the natural language question; extracting information related to an answer from the natural language question; searching in linked data according to the detected named entity; generating a candidate answer according to a search result; parsing the candidate answer according to the information related to the answer; and obtaining a value of a feature of the candidate answer; and evaluating each candidate answer by synthesizing the value of the feature of the candidate answer.

Claims

1. A computer implemented method for selecting an answer to a natural language question, comprising:

detecting a named entity in the natural language question;

extracting information related to an answer from the natural language question;

searching in linked data according to the detected named entity;

generating at least one candidate answer according to a search result;

parsing the candidate answer according to the information related to the answer;

obtaining a value of a feature of the candidate answer; and

evaluating each candidate answer by synthesizing the value of the feature of the candidate answer.

2. The method according to claim 1, wherein the step of searching in linked data according to the detected named entity comprises:

searching for a Uniform Resource Identifier (URI), matching to the named entity in the linked data based on similarity; and

searching with spreading activation, for a URI linked to the URI matching to the named entity by using linking relationship between URIs; and

generating the candidate answer according to the linked URI.

3. The method according to claim 2, wherein the candidate answer is a label contained in the URI.

4. The method according to claim 3, wherein the feature of the candidate answer comprises at least the information related to the answer, and the number of directly linked URIs associated with the candidate answer.

5. The method according to claim 4, wherein the information related to the answer comprises at least a type of the answer.

6. The method according to claim 5, wherein the information related to the answer further comprises time verification information related to the answer extracted from the natural language question, and the feature of the candidate answer further comprises linking information between URIs matching to the named entity.

7. The method according to claim 1, further comprising merging candidate answers retrieved from different linked data according to the feature of the candidate answers.

8. The method according to claim 1, further comprising:

performing machine learning according to the feature of the candidate answer to train a scoring model; and

computing a score for each candidate answer according to the scoring model while evaluating each candidate answer.

9. The method according to claim 1, wherein the linked data is Resource Description Framework data.

10. The method according to claim 1, wherein the linked data is obtained by mapping Micro-format data.

21. A computer-readable storage medium tangibly embodying computer-executable program instructions which, when executed, cause a computer to perform a method for selecting an answer to a natural language question according to claim 1.

11. An apparatus for selecting an answer to a natural language question, comprising:

a question parsing module configured to detect a named entity in the natural language question and extract information related to an answer from the natural language question;

a candidate answer generating module configured to search in linked data according to the detected named entity and generate a candidate answer according to a search result;

a feature value generating module configured to parse the candidate answer according to the information related to the answer and obtain a value of a feature of the candidate answer; and

a candidate answer evaluating module configured to evaluate each candidate answer by synthesizing the value of the feature of the candidate answer.

12. The apparatus according to claim 11, wherein the candidate answer generating module comprises:

a matching unit, configured to search for a Uniform Resource Identifier, URI, matching to the named entity in the linked data based on similarity;

a spreading activation unit, configured to search with spreading activation for a URI linked to the URI matching to the named entity by using linking relationship between URIs; and

a candidate generating unit, configured to generate the candidate answer according to the linked URI.

13. The apparatus according to claim 12, wherein the candidate generating unit uses a label contained in the URI as the candidate answer.

14. The apparatus according to claim 13, wherein the feature of the candidate answer comprises at least the information related to the answer, and the number of directly linked URIs associated with the candidate answer.

15. The apparatus according to claim 14, wherein the information related to the answer comprises at least a type of the answer.

16. The apparatus according to claim 15, wherein:

the information related to the answer further comprises time verification information related to the answer extracted from the natural language question; and

the feature of the candidate answer further comprises linking information between URIs matching to the named entity.

17. The apparatus according to claim 11, further comprising a merging module configured to merge candidate answers retrieved from different linked data according to the feature of the candidate answers.

18. The apparatus according to claim 11, further comprising a training module configured to perform machine learning according to the feature of the candidate answer to train a scoring model;

and

wherein the candidate answer evaluating module computes a score for each candidate answer according to the scoring model while evaluating each candidate answer.

19. The apparatus according to claim 11, wherein the linked data is Resource Description Framework data.

20. The apparatus according to claim 11, wherein the linked data is obtained by mapping Micro-format data.

22. A computer-readable storage medium tangibly embodying computer-executable program instructions which, when executed, cause a computer to be configured to function as the apparatus for selecting an answer to a natural language question according to claim 11.