Answer Determination for Natural Language Questioning

• US 20150052113A1
• Filed: 09/08/2014
• Published: 02/19/2015
• Est. Priority Date: 11/30/2005
• Status: Abandoned Application

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property II LP (AT&T, Inc.)
• Inventors: Hakkani-Tur, Dilek Z., Tur, Gokhan, STENCHIKOVA, Svetlana
• Application Number:
  US14/479,765
• Publication Number:
  US 20150052113A1
• Time in Patent Office: Days
• US Class Current: 707/706
• CPC Class Codes: G06F 16/243 Natural language query formulation
• CPC Class Codes: G06F 16/3331 Query processing
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 40/35 Discourse or dialogue representation

Abstract

Open-domain question answering is the task of finding a concise answer to a natural language question using a large domain, such as the Internet. The use of a semantic role labeling approach to the extraction of the answers to an open domain factoid (Who/When/What/Where) natural language question that contains a predicate is described. Semantic role labeling identities predicates and semantic argument phrases in the natural language question and the candidate sentences. When searching for an answer to a natural language question, the missing argument in the question is matched using semantic parses of the candidate answers. Such a technique may improve the accuracy of a question answering system and may decrease the length of answers for enabling voice interface to a question answering system.

Claims

1. A method comprising:

receiving a natural language question from a user;

generating a search phrase based on the natural language question, the search phrase comprising a plurality of words in a non-stop order;

identifying a plurality of candidate sentences based on the search phrase, wherein each candidate sentence in the plurality of candidate sentences contains the plurality of words in the non-stop order;

evaluating the plurality of candidate sentences using a cascaded approach comprising a baseline approach, a semantic role labeler approach, and a combination of the baseline approach and the semantic role labeler approach, to yield a level of precision for each candidate sentence;

when the level of precision of a candidate sentence meets a threshold, generating a candidate answer by using a phrase sentence extraction approach to extract a portion from a side of the search phrase within the candidate sentence; and

providing the candidate answer in response to the natural language question.

2. The method of claim 1, wherein the candidate answer is one of a plurality of candidate answers, and further comprising ranking each candidate answer in the plurality of candidate answers by a confidence level.

3. The method of claim 2, wherein the candidate answer has a highest confidence level in the plurality of candidate answers.

4. The method of claim 1, wherein identifying of the plurality of candidate sentences further comprises:

scanning a plurality of documents for the search phrase.

5. The method of claim 1, wherein a conjunction of a sub-phrase occurs in a final step of the cascaded approach and comprises determining whether a predicate and a search argument are found within the natural language question.

6. The method of claim 5, wherein the search argument is classified as a location argument.

7. The method of claim 5, wherein the search argument is classified as a temporal argument.

8. A system comprising:

a processor; and

a computer-readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising;

receiving a natural language question from a user;

generating a search phrase based on the natural language question, the search phrase comprising a plurality of words in a non-stop order;

identifying a plurality of candidate sentences based on the search phrase, wherein each candidate sentence in the plurality of candidate sentences contains the plurality of words in the non-stop order;

evaluating the plurality of candidate sentences using a cascaded approach comprising a baseline approach, a semantic role labeler approach, and a combination of the baseline approach and the semantic role labeler approach, to yield a level of precision for each candidate sentence;

when the level of precision of a candidate sentence meets a threshold, generating a candidate answer by using a phrase sentence extraction approach to extract a portion from a side of the search phrase within the candidate sentence; and

providing the candidate answer in response to the natural language question.

9. The system of claim 8, wherein the candidate answer is one of a plurality of candidate answers;

and

the computer-readable storage medium has additional instructions stored which, when executed by the processor, result in operations comprising ranking each candidate answer in the plurality of candidate answers by a confidence level.

10. The system of claim 9, wherein the candidate answer has a highest confidence level in the plurality of candidate answers.

11. The system of claim 8, wherein identifying of the plurality of candidate sentences further comprises:

scanning a plurality of documents for the search phrase.

12. The system of claim 8, wherein a conjunction of a sub-phrase occurs in a final step of the cascaded approach and comprises determining whether a predicate and a search argument are found within the natural language question.

13. The system of claim 12, wherein the search argument is classified as a location argument.

14. The system of claim 12, wherein the search argument is classified as a temporal argument.

15. A computer-readable storage device having instructions stored which, when executed by a computing device, cause the computing device to perform operations comprising:

receiving a natural language question from a user;

generating a search phrase based on the natural language question, the search phrase comprising a plurality of words in a non-stop order;

identifying a plurality of candidate sentences based on the search phrase, wherein each candidate sentence in the plurality of candidate sentences contains the plurality of words in the non-stop order;

evaluating the plurality of candidate sentences using a cascaded approach comprising a baseline approach, a semantic role labeler approach, and a combination of the baseline approach and the semantic role labeler approach, to yield a level of precision for each candidate sentence;

when the level of precision of a candidate sentence meets a threshold, generating a candidate answer by using a phrase sentence extraction approach to extract a portion from a side of the search phrase within the candidate sentence; and

providing the candidate answer in response to the natural language question.

16. The computer-readable storage device of claim 15, wherein the candidate answer is one of a plurality of candidate answers;

and

having additional instructions stored which, when executed by the computing device, result in operations comprising ranking each candidate answer in the plurality of candidate answers by a confidence level.

17. The computer-readable storage device of claim 16, wherein the candidate answer has a highest confidence level in the plurality of candidate answers.

18. The computer-readable storage device of claim 15, wherein identifying of the plurality of candidate sentences further comprises:

scanning a plurality of documents for the search phrase.

19. The computer-readable storage device of claim 15, wherein a conjunction of a sub-phrase occurs in a final step of the cascaded approach and comprises determining whether a predicate and a search argument are found within the natural language question.

20. The computer-readable storage device of claim 19, wherein the search argument is classified as a location argument.