Error correction using fact repositories

• US 8,560,300 B2
• Filed: 09/09/2009
• Issued: 10/15/2013
• Est. Priority Date: 09/09/2009
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ferrucci, David A., Gondek, David C., Zadrozny, Wlodek W.
• Primary Examiner(s): Harper, Vincent P
• Application Number:
  US12/556,090
• Publication Number:
  US 20110060584A1
• Time in Patent Office: 1,497 Days
• Field of Search: 704/9, 704/8, 715/210, 715/217
• US Class Current: 704/9
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation

Abstract

The disclosed system and method apply stores of factual information to correct errors in digital text, for example, generated from OCR, speech and/or handwriting recognition devices, and other automatic recognition devices. A text produced by OCR, speech recognition, handwriting recognition, and others may be processed to extract discussed facts. Databases of facts are searched based on information in the text. After comparing facts asserted in the text with the factual data from the databases, suggested corrections of the text are produced.

Claims

1. A method for correcting error in digital text using fact repositories, comprising:

searching computer data repository storing information for one or more relevant facts associated with input digital text;

extracting one or more facts postulated in the input digital text;

comparing said one or more facts postulated in the input digital text with said one or more relevant facts;

analyzing one or more differences between said one or more facts postulated in the input digital text and said one or more relevant facts resulting from the comparing step; and

transforming the input digital text into corrected digital text by correcting the input digital text using evidence resulting from the analyzing step,

wherein the analyzing further comprises producing a score identifying a degree of inconsistency between said one or more facts postulated in the input digital text and said one or more relevant facts, and the transforming further comprises highlighting the corrected digital text in a manner to indicate a certainty level of correction.

2. The method of claim 1, wherein the digital text is generated by one or more of OCR, speech recognition device, handwriting recognition device, or combinations thereof.

3. The method of claim 1, wherein the digital text include metadata.

4. The method of claim 1, wherein the evidence is generated using machine learning.

5. The method of claim 1, wherein said one or more relevant facts associated with input digital text are represented as rich data structures.

6. The method of claim 1, wherein the comparing step further includes constructing a mapping between elements of said one or more facts postulated in the input digital text and said one or more relevant facts.

7. The method of claim 1, wherein the analyzing step includes using a model trained over a training corpus of facts.

8. A system for error correction using fact repositories, comprising:

one or more fact repositories stored on one or more storage devices; and

a correction module executable on a computer processor, the correction module operable to search the fact repositories for one or more relevant facts associated with input digital text, the correction module further operable to extract one or more facts postulated in the input digital text and determine one or more differences between said one or more facts postulated in the input digital text and said one or more relevant facts retrieved from the fact repositories, the correction module further operable to transform the input digital text into corrected digital text using evidence resulting from the determined differences,

wherein the correction module is further operable to produce a score identifying a degree of inconsistency between said one or more facts postulated in the input digital text and said one or more relevant facts, and further operable to highlight the corrected digital text in a manner to indicate a certainty level of correction.

9. The system of claim 8, further including an output device for outputting the corrected digital text.

10. The system of claim 9, wherein the output device includes a display device, a storage device, or network communication interface device, or combinations thereof.

11. The system of claim 8, wherein the one or more fact repositories include one or more databases storing information.

12. The system of claim 8, wherein the one or more fact repositories are accessed over the Internet.

13. The system of claim 8, wherein the input digital text is generated from one or more of OCR, speech recognition device, or handwriting recognition device, or combinations thereof.

14. The system of claim 8, further including an interface connected to one or more automatic recognition devices that generate the input digital text, the interface operable to receive the input digital text.

15. The system of claim 8, wherein the correction module includes:

a querying module operable to query one or more repositories for relevant facts associated with elements of the input digital text;

a facts extraction module operable to extract postulated facts in the input digital text;

a comparison module operable to determine differences between the postulated facts and the relevant facts; and

a correction model operable transform the input digital text into corrected digital text based on the differences between the postulated facts and the relevant facts.

16. The system of claim 15, wherein the correction model includes a machine learning module.

17. The system of claim 15, wherein the correction model includes an evidence gathering module.

18. A program storage memory device readable by a machine, tangibly embodying a program of instructions executable by the machine to perform a method of modeling and evaluating impact on service components, comprising:

correcting error in digital text using fact repositories, comprising;

searching computer data repository storing facts for one or more relevant facts associated with input digital text;

extracting one or more facts postulated in the input digital text;

comparing said one or more facts postulated in the input digital text with said one or more relevant facts;

analyzing one or more differences between said one or more facts postulated in the input digital text and said one or more relevant facts resulting from the comparing step; and

transforming the input digital text into corrected digital text by correcting the input digital text using evidence resulting from the analyzing step,

wherein the analyzing further comprises producing a score identifying a degree of inconsistency between said one or more facts postulated in the input digital text and said one or more relevant facts, and the transforming further comprises highlighting the corrected digital text in a manner to indicate a certainty level of correction.

19. The program storage memory device of claim 18, wherein the digital text is generated by OCR, speech recognition device, handwriting recognition device, or combinations thereof.

20. The program storage memory device of claim 18, wherein the digital text include metadata.