METHOD AND COMPUTER SYSTEM FOR AUTOMATICALLY ANSWERING NATURAL LANGUAGE QUESTIONS

• US 20110153312A1
• Filed: 10/22/2008
• Published: 06/23/2011
• Est. Priority Date: 10/23/2007
• Status: Active Grant

• Current Assignee: Thomas Roberts
• Original Assignee: Thomas Roberts
• Inventors: Roberts, Thomas
• Granted Patent: US 8,538,744 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 computer system and method for automatically answering natural language questions. The system comprises an input to receive said natural language questions; a data store to record linked pairs of questions and corresponding answers; a matcher configured to compare a said received natural language question withsaid linked question and answer pairs and an output to transfer a said received natural language question to a researcher if no matches are found. The system may further comprise a system to link pairs of questions and corresponding answers into groups, to enable the generation of a prototypicalanswer for each group of pairs of questions and answers and to store said prototypicalanswers in said data store; wherein said matcher compares a said received naturallanguage question with a question in said data store having an associated prototypicalanswer and output said associated prototypical answer for said question in response to said matching. Alternatively, said matcher may be configured to output all linked question and answer pairs which matchsaid received natural language question. The system may be further adapted to distribute natural language questions to be answered to researchers by assigning unpopularity scores to each of said natural language questions.

Claims

1. A computer system for automatically answering natural language questions, the system comprising:

an input module to receive said natural language questions;

a data store to record linked pairs of questions and corresponding answers;

a system to link pairs of questions and corresponding answers into groups, to enable the generation of a prototypical answer that answers each question from said pairs of questions and answers in said group, and to store said prototypical answers in said data store;

a matcher configured to compare a said received natural language question with a question in said data store having an associated prototypical answer and outputs said associated prototypical answer for said question in response to said matching.

3. A computer system according to claim 1, wherein said system to link pairs of questions and corresponding answers is configured to construct a disconnected directed graph of said pairs of questions and corresponding answers.

4. A computer system as claimed in claim 1, wherein said system to link pairs of questions and corresponding answers includes a moderator user interface which is configured to present said groups of pairs of questions and corresponding answers to a moderator for review to generate said prototypical answer and which is configured to input said prototypical answer to said data store to store said prototypical answer in association with said group of pairs of questions and corresponding answers.

5. A computer system as claimed in claim 1, wherein said matcher is configured to output all linked question and answer pairs which match said received natural language question.

6. A computer system as claimed in claim 5, further comprising an input to receive a request to transmit a specific natural language question to a researcher and wherein said output transmits said specific natural language question to a researcher in response to said request.

7. A computer system according to any one of the preceding claims, further comprising a researcher user interface to present a said question to a researcher and to present an input box for a researcher to input an answer to the presented question.

8. A computer system according to claim 1, wherein said data store stores a record of a source used to generate each linked pair of question and answer and wherein said system may output said record with said linked pair of question and answer.

9. A computer system as claimed in claim 1, which is configured to determine whether a question received by the system is time dependent.

10. A computer system as claimed in claim 9, further comprising means to query a real-time data feed to generate an answer to a time dependent question.

11. A computer system as claimed in claim 9, wherein a natural language parsing system is used to determine whether a question is time dependent.

12. A computer system as claimed in claim 1, further adapted to distribute natural language questions to be answered to researchers, the system further comprising:

a server configured to assign unpopularity scores to each of said natural language questions;

a plurality of researcher work stations each having an researcher user interface, and a communications network connecting said server to said plurality of researcher work stations;

wherein each researcher user interface presents a plurality of said natural language questions to a researcher and wherein said server is configured to assign a value to each researcher based on the unpopularity score of each natural language questions selected by said researcher.

14. A computer system according to claim 1, wherein the server is configured to assign an initial unpopularity score of zero to each natural language question and ‘

subsequently increases the unpopularity score of each natural language question by tracking when a natural language question is presented to a researcher and not answered.

15. A computer system according to claim 14, wherein the unpopularity score of each natural language question is increased by determining its position in a list of work items presented to a researcher relative to a natural language questions selected by said researcher.

16. A computer system according to claim 12, wherein the server is configured to calculate said value of each researcher as a time-weighted combination of the unpopularity scores of each natural language question selected by that researcher over a particular time frame.

17. A computer system according to claim 16, wherein the server uses a modified time decay function to calculate said value as a time-weighted combination of the unpopularity scores of each work item.

2. (canceled)

13. (canceled)

18-69. -69. (canceled)