CORRECTION OF MISSPELLINGS IN QA SYSTEM

• US 20200134010A1
• Filed: 10/26/2018
• Published: 04/30/2020
• Est. Priority Date: 10/26/2018
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Boxwell, Stephen A., Frost, Keith G., Vernier, Stanley J., Brake, Kyle M.
• Granted Patent: US 10,803,242 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 40/20 Natural language analysis semantic analysis of natural language G06F40/30
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/153 using recognition of characters or words
• CPC Class Codes: G06V 30/268 Lexical context
• CPC Class Codes: G06V 30/40 Document-oriented image-based pattern recognition
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Embodiments provide a computer implemented method for identifying and correcting a misspelling in a question answering (QA) system, wherein the QA system is coupled to a document corpus, and the document corpus includes a plurality of documents related to a particular domain. The method includes the following steps: receiving an input question and a plurality of passages, wherein the plurality of passages are extracted from the document corpus by the QA system; providing at least one alternate form for each token extracted from the input question and the plurality of passages; identifying at least one misspelled token; and scoring at least one alternate form of each identified misspelled token.

Claims

1. A computer implemented method for identifying and correcting a misspelling in a question answering (QA) system, wherein the QA system is coupled to a document corpus, and the document corpus includes a plurality of documents related to a particular domain, the method comprising:

receiving, by a processor coupled to one or more user devices, an input question and a plurality of passages, wherein the plurality of passages are extracted from the document corpus by the QA system;

providing, by the processor, at least one alternate form for each token extracted from the input question and the plurality of passages;

identifying, by the processor, at least one misspelled token from the input question and the plurality of passages; and

scoring, by the processor, at least one alternate form of each identified misspelled token.

2. The method as recited in claim 1, further comprising:

providing, by the processor, a modified Levenshtein distance value for each alternate form, wherein the modified Levenshtein distance value is between 0 and 1; and

identifying, by the processor, at least one token having the modified Levenshtein distance value more than a predetermined threshold value.

3. The method as recited in claim 2, wherein the input question is submitted to the QA system through a speech to text process, and the modified Levenshtein distance value of each token of the input question is provided based on a first substitution confusion matrix associated with misspellings generated during the speech to text process.

4. The method as recited in claim 3, wherein in the first substitution confusion matrix, a first group of characters are substituted with a single character, or a second group of characters.

5. The method as recited in claim 2, wherein the plurality of passages of the document corpus are provided through an optical character recognition process, and the modified Levenshtein distance value of each token of the plurality of passages is provided based on a second substitution confusion matrix associated with misspellings generated during the optical character recognition process.

6. The method as recited in claim 5, wherein in the second substitution confusion matrix, a first group of characters are substituted with a single character, or a second group of characters.

7. The method as recited in claim 3, wherein the plurality of passages of the document corpus is provided through an optical character recognition process, and the modified Levenshtein distance value of each token of the plurality of passages is provided based on a second substitution confusion matrix associated with misspellings generated during the optical character recognition process.

8. The method as recited in claim 2, wherein the at least one alternate form includes a synonym and a corrected spelling of the at least one misspelled token.

9. A computer program product for identifying and correcting a misspelling in a question answering (QA) system, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

receive, by a processor coupled to one or more user devices, an input question and a plurality of passages, wherein the plurality of passages are extracted from the document corpus by the QA system;

provide, by the processor, at least one alternate form for each token extracted from the input question and the plurality of passages;

provide, by the processor, a modified Levenshtein distance value for each alternate form, wherein the modified Levenshtein distance value is between 0 and 1;

identify, by the processor, at least one misspelled token and at least one token having the modified Levenshtein distance value more than a predetermined threshold value from the input question and the plurality of passages; and

score, by the processor, at least one alternate form of each identified misspelled token.

10. The computer program product as recited in claim 9, wherein the input question is submitted to the QA system through a speech to text process, and the modified Levenshtein distance value of each token of the input question is provided based on a first substitution confusion matrix associated with misspellings generated during the speech to text process.

11. The computer program product as recited in claim 10, wherein in the first substitution confusion matrix, a first group of characters are substituted with a single character, or a second group of characters.

12. The computer program product as recited in claim 9, wherein the plurality of passages of the document corpus are provided through an optical character recognition process, and the modified Levenshtein distance value of each token of the plurality of passages is provided based on a second substitution confusion matrix associated with misspellings generated during the optical character recognition process.

13. The computer program product as recited in claim 12, wherein in the second substitution confusion matrix, a first group of characters are substituted with a single character, or a second group of characters.

14. The computer program product as recited in claim 10, wherein the plurality of passages of the document corpus are provided through an optical character recognition process, and the modified Levenshtein distance value is provided based on a second substitution confusion matrix associated with misspellings generated during the optical character recognition process.

15. The computer program product as recited in claim 9, wherein the at least one alternate form includes a synonym and a corrected spelling of the at least one misspelled token.

16. A system for identifying and correcting a misspelling in a question answering (QA) system, comprising:

a processor configured to;

receive, by a processor coupled to one or more user devices, an input question and a plurality of passages, wherein the plurality of passages are extracted from the document corpus by the QA system;

provide, by the processor, at least one alternate form for each token extracted from the input question and the plurality of passages;

provide, by the processor, a modified Levenshtein distance value for each alternate form, wherein the modified Levenshtein distance value is between 0 and 1;

identify, by the processor, at least one misspelled token and at least one token having the modified Levenshtein distance value more than a predetermined threshold value from the input question and the plurality of passages; and

score, by the processor, at least one alternate form of each identified misspelled token.

17. The system as recited in claim 16, wherein the input question is submitted to the QA system through a speech to text process, and the modified Levenshtein distance value of each token of the input question is provided based on a first substitution confusion matrix associated with misspellings generated during the speech to text process.

18. The system as recited in claim 17, wherein the plurality of passages of the document corpus are provided through an optical character recognition process, and the modified Levenshtein distance value of each token of the plurality of passages is provided based on a second substitution confusion matrix associated with misspellings generated during the optical character recognition process.

19. The system as recited in claim 18, wherein in the first substitution confusion matrix and the second substitution confusion matrix, a first group of characters are substituted with a single character, or a second group of characters.

20. The system as recited in claim 18, wherein the at least one alternate form includes a synonym and a corrected spelling of the at least one misspelled token.