Verification and correction method and system for optical character recognition

• US 5,933,531 A
• Filed: 08/23/1996
• Issued: 08/03/1999
• Est. Priority Date: 08/23/1996
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Lorie, Raymond Amand
• Primary Examiner(s): Shalwala, Bipin H.
• Assistant Examiner(s): Mariam, Daniel G.
• Application Number:
  US08/697,380
• Time in Patent Office: 1,075 Days
• Field of Search: 382/138, 382/161, 382/229, 382/231, 382/311, 382/178, 345/338, 345/468, 395/183.14, 395/183.22, 395/568
• US Class Current: 382/229
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/127 with the intervention of an operator
• CPC Class Codes: G06V 30/268 Lexical context

Abstract

An optical character recognition method and system are provided, employing context analysis and operator input, alternatively and in combination, on the same batch of documents. After automatic character recognition, the context analyzer processes the fields that are good enough to expect resolution. This will accept as many fields as possible without any operator intervention. For some other fields, the process uses operator input to certify the character-level OCR result of, or to enter, a certain percentage of the characters, so that context analysis may accept some of the remaining fields. If the context analyzer successfully identifies a small set of very close hypotheses, the process asks the operator to certify one or two characters to resolve the ambiguity between the hypotheses. For the fields that are still not resolved, the fields and the hypotheses are shown to the operator for acceptance, correction, or entry.

Claims

1. A method for correcting and verifying results of an optical character recognition process comprising:

a) performing said optical character recognition on an image including characters to be recognized;

b) performing an automatic context analysis on said image;

c) evaluating results obtained by said automatic context analysis, to identify characters requiring further processing;

d) performing a user-assisted verification and correction of less than all of said characters identified as requiring further processing, thereby producing a partially recognized image, wherein said user-assisted verification and correction is omitted if a predetermined level of confidence of recognition of said image is reached by said evaluating of said results; and

e) repeating steps b-d on said partially recognized image until said predetermined level of confidence of recognition of said image is reached by said evaluating of said results.

2. A method as recited in claim 1, wherein the step of performing an automatic context analysis includes identifying characters which satisfy a first predetermined threshold of confidence that the characters are correctly recognized.

3. A method as recited in claim 2, wherein the step of performing an automatic context analysis further includes performing a dictionary lookup on characters which satisfy the first predetermined threshold of confidence.

4. A method as recited in claim 2, wherein the step of performing an automatic context analysis further includes performing a fuzzy search based on the characters which satisfy the first predetermined threshold of confidence.

5. A method as recited in claim 2, wherein the step of performing an automatic context analysis further includes performing a user-assisted verification and correction of characters which satisfy a second predetermined threshold of confidence, so that results satisfy the first predetermined threshold of confidence.

6. A method as recited in claim 5, further comprising the step, executed prior to the step of performing a user-assisted verification and correction of characters which satisfy a second predetermined threshold, of determining that confusion is possible between two possible solutions, both of which may be correct results of verification and correction of the characters which satisfy the second predetermined threshold of confidence.

7. A method as recited in claim 1, wherein the step of performing an automatic context analysis includes identifying characters which satisfy a first predetermined threshold of confidence that the characters are correctly recognized.

8. A method as recited in claim 7, wherein the step of performing a user-assisted verification and correction further includes performing a dictionary lookup on characters which satisfy the first predetermined threshold of confidence.

9. A method as recited in claim 7, wherein the step of performing a user-assisted verification and correction further includes performing a fuzzy search based on the characters which satisfy the first predetermined threshold of confidence.

10. A method as recited in claim 7, wherein the step of performing a user-assisted verification and correction further includes performing a user-assisted verification and correction of characters which satisfy a second predetermined threshold of confidence, so that results satisfy the first predetermined threshold of confidence.

11. A method as recited in claim 10, further comprising the step, executed prior to the step of performing a user-assisted verification and correction, of determining that confusion is possible between two possible solutions, both of which may be correct results of verification and correction of the characters which satisfy the second predetermined threshold of confidence.

12. A method as recited in claim 1, wherein the step of repeating includes temporarily storing intermediate results, between iterations, in a workspace storage medium.

13. A method as recited in claim 1, wherein the step of evaluating the results of the automatic context analysis includes:

identifying distinct fields within a form; and

marking as ACCEPTED those of the fields which satisfy a confidence threshold.

14. A method as recited in claim 1, wherein the step of performing a user-assisted verification and correction includes:

identifying a field whose results are uncertain; and

identifying a subset of characters of the identified field whose identification would resolve the ambiguity.

15. A method as recited in claim 14, wherein the step of identifying a subset of characters includes:

identifying two possible valid values for the field which have characters in common; and

identifying the subset as those characters which the two possible valid values do not have in common.

16. A method as recited in claim 1, wherein the step of performing a user-assisted verification and correction includes:

presenting the user with a selection of possible choices for characters within a field; and

allowing the user the options of;

(i) ratifying a choice selected as the best choice by the system,

(ii) selecting one of the selection of choices other than the choice selected by the system as the best, and

(iii) entering a choice different from those within the selection.

49. A method as recited in claim 1, wherein said repeating step alternates said user assisted verification and said automatic context analysis on a given field of said image.

50. A method as recited in claim 1, wherein said step of performing a user-assisted verification comprises a step of requiring user input on critical characters of said characters.

17. A system for correcting and verifying results of an optical character recognition process comprising:

a) means for performing said optical character recognition on an image including characters to be recognized;

b) means for performing an automatic context analysis on said image;

c) means for evaluating results obtained by said automatic context analysis, to identify characters requiring further processing;

d) means for performing a user-assisted verification and correction of less than all of said characters identified as requiring further processing, thereby producing a partially recognized image, wherein said user-assisted verification and correction is omitted if a predetermined level of confidence of recognition said image is reached by said evaluating of said results; and

e) means for repeating processing performed by means b-d on said partially recognized image until said predetermined level of confidence of recognition of said image is reached by said evaluating of said results.

18. A system as recited in claim 17, wherein the means for performing an automatic context analysis includes means for identifying characters which satisfy a first predetermined threshold of confidence that the characters are correctly recognized.

19. A system as recited in claim 18, wherein the means for performing an automatic context analysis further includes means for performing a dictionary lookup on characters which satisfy the first predetermined threshold of confidence.

20. A system as recited in claim 18, wherein the means for performing an automatic context analysis further includes means for performing a fuzzy search based on the characters which satisfy the first predetermined threshold of confidence.

21. A system as recited in claim 18, wherein the means for performing an automatic context analysis further includes means for performing a user-assisted verification and correction of characters which satisfy a second predetermined threshold of confidence, so that results satisfy the first predetermined threshold of confidence.

22. A system as recited in claim 21, further comprising means, operable prior to operation of the means for performing a user-assisted verification and correction of characters which satisfy a second predetermined threshold, for determining that confusion is possible between two possible solutions, both of which may be correct results of verification and correction of the characters which satisfy the second predetermined threshold of confidence.

23. A system as recited in claim 17, wherein the means for performing an automatic context analysis includes means for identifying characters which satisfy a first predetermined threshold of confidence that the characters are correctly recognized.

24. A system as recited in claim 23, wherein the means for performing a user-assisted verification and correction further includes means for performing a dictionary lookup on characters which satisfy the first predetermined threshold of confidence.

25. A system as recited in claim 23, wherein the means for performing a user-assisted verification and correction further includes means for performing a fuzzy search based on the characters which satisfy the first predetermined threshold of confidence.

26. A system as recited in claim 23, wherein the means for performing a user-assisted verification and correction further includes means for performing a user-assisted verification and correction of characters which satisfy a second predetermined threshold of confidence, so that results satisfy the first predetermined threshold of confidence.

27. A system as recited in claim 26, further comprising means, operable prior to operation of the means for performing a user-assisted verification and correction, for determining that confusion is possible between two possible solutions, both of which may be correct results of verification and correction of the characters which satisfy the second predetermined threshold of confidence.

28. A system as recited in claim 17, wherein the means for repeating includes means for temporarily storing intermediate results, between iterations, in a workspace storage medium.

29. A system as recited in claim 17, wherein the means for evaluating the results of the automatic context analysis includes:

means for identifying distinct fields within a form; and

means for marking as ACCEPTED those of the fields which satisfy a confidence threshold.

30. A system as recited in claim 17, wherein the means for performing a user-assisted verification and correction includes:

means for identifying a field whose results are uncertain; and

means for identifying a subset of characters of the identified field whose identification would resolve the ambiguity.

31. A system as recited in claim 30, wherein the means for identifying a subset of characters includes:

means for identifying two possible valid values for the field which have characters in common; and

means for identifying the subset as those characters which the two possible valid values do not have in common.

32. A system as recited in claim 17, wherein the means for performing a user-assisted verification and correction includes:

means for presenting the user with a selection of possible choices for characters within a field; and

means for allowing the user the options of;

(i) ratifying a choice selected as the best choice by the system,

(ii) selecting one of the selection of choices other than the choice selected by the system as the best, and

(iii) entering a choice different from those within the selection.

51. A system as recited in claim 17, wherein said repeating means alternates said user assisted verification and said automatic context analysis on a given field of said image.

52. A system as recited in claim 17, wherein said means for performing a user-assisted verification comprises means for requiring user input on critical characters of said characters.

33. A computer program product, for use with a computer system, for causing the computer system to correct and verify results of an optical character recognition process, the computer program product comprising:

a) a computer-readable medium;

b) means, provided on the computer-readable medium, for directing the computer system to perform said optical character recognition on an image including characters to be recognized;

c) means, provided on the computer-readable medium, for directing the computer system to perform an automatic context analysis on said image;

d) means, provided on the computer-readable medium, for directing the computer system to evaluate results obtained by said automatic context analysis performing means, to identify characters requiring further processing;

e) means, provided on the computer-readable medium, for directing the computer system to perform a user-assisted verification and correction of less than all of said characters identified as requiring further processing, thereby producing a partially recognized image, wherein said user-assisted verification and correction means is bypassed if a predetermined level of confidence of recognition of said image is reached by said evaluating means; and

f) means, provided on the computer-readable medium, for directing the computer system to repeat processing performed by means c-e on said partially recognized image until said predetermined level of confidence of recognition of said image is reached by said evaluating means.

34. A computer program product as recited in claim 33, wherein the means for directing to perform an automatic context analysis includes means, provided on the computer-readable medium, for directing the computer system to identify characters which satisfy a first predetermined threshold of confidence that the characters are correctly recognized.

35. A computer program product as recited in claim 34, wherein the means for directing to perform an automatic context analysis further includes means, provided on the computer-readable medium, for directing the computer system to perform a dictionary lookup on characters which satisfy the first predetermined threshold of confidence.

36. A computer program product as recited in claim 34, wherein the means for directing to perform an automatic context analysis further includes means, provided on the computer-readable medium, for directing the computer system to perform a fuzzy search based on the characters which satisfy the first predetermined threshold of confidence.

37. A computer program product as recited in claim 34, wherein the means for directing to perform an automatic context analysis further includes means, provided on the computer-readable medium, for directing the computer system to perform a user-assisted verification and correction of characters which satisfy a second predetermined threshold of confidence, so that results satisfy the first predetermined threshold of confidence.

38. A computer program product as recited in claim 37, further comprising means, provided on the computer-readable medium, operable prior to operation of the means for directing to perform a user-assisted verification and correction of characters which satisfy a second predetermined threshold, for directing the computer system to determine that confusion is possible between two possible solutions, both of which may be correct results of verification and correction of the characters which satisfy the second predetermined threshold of confidence.

39. A computer program product as recited in claim 33, wherein the means for directing to perform an automatic context analysis includes means, provided on the computer-readable medium, for directing the computer system to identify characters which satisfy a first predetermined threshold of confidence that the characters are correctly recognized.

40. A computer program product as recited in claim 39, wherein the means for directing to perform a user-assisted verification and correction further includes means, provided on the computer-readable medium, for directing the computer system to perform a dictionary lookup on characters which satisfy the first predetermined threshold of confidence.

41. A computer program product as recited in claim 39, wherein the means for directing to perform a user-assisted verification and correction further includes means, provided on the computer-readable medium, for directing the computer system to perform a fuzzy search based on the characters which satisfy the first predetermined threshold of confidence.

42. A computer program product as recited in claim 39, wherein the means for directing to perform a user-assisted verification and correction further includes means, provided on the computer-readable medium, for directing the computer system to perform a user-assisted verification and correction of characters which satisfy a second predetermined threshold of confidence, so that results satisfy the first predetermined threshold of confidence.

43. A computer program product as recited in claim 42, further comprising means, provided on the computer-readable medium, operable prior to operation of the means for directing to perform a user-assisted verification and correction, for directing the computer system to determine that confusion is possible between two possible solutions, both of which may be correct results of verification and correction of the characters which satisfy the second predetermined threshold of confidence.

44. A computer program product as recited in claim 33, wherein the means for directing to repeat includes means, provided on the computer-readable medium, for directing the computer system to temporarily store intermediate results, between iterations, in a workspace storage medium.

45. A computer program product as recited in claim 33, wherein means for directing to evaluate the results of the automatic context analysis includes:

means, provided on the computer-readable medium, for directing the computer system to identify distinct fields within a form; and

means, provided on the computer-readable medium, for directing the computer system to mark as ACCEPTED those of the fields which satisfy a confidence threshold.

46. A computer program product as recited in claim 33, wherein the means for directing to perform a user-assisted verification and correction includes:

means, provided on the computer-readable medium, for directing the computer system to identify a field whose results are uncertain; and

means, provided on the computer-readable medium, for directing the computer system to identify a subset of characters of the identified field whose identification would resolve the ambiguity.

47. A computer program product as recited in claim 46, wherein the means for directing to identify a subset of characters includes:

means, provided on the computer-readable medium, for directing the computer system to identify two possible valid values for the field which have characters in common; and

means, provided on the computer-readable medium, for directing the computer system to identify the subset as those characters which the two possible valid values do not have in common.

48. A computer program product as recited in claim 33, wherein the means for directing to perform a user-assisted verification and correction includes:

means, provided on the computer-readable medium, for directing the computer system to present the user with a selection of possible choices for characters within a field; and

means, provided on the computer-readable medium, for directing the computer system to allow the user the options of;

(i) ratifying a choice selected as the best choice by the system,

(ii) selecting one of the selection of choices other than the choice selected by the system as the best, and

(iii) entering a choice different from those within the selection.

53. A computer program product as recited in claim 33, wherein said means for directing the computer system to repeat steps alternates said user assisted verification and said automatic context analysis on a given field of said image.

54. A computer program product as recited in claim 33, wherein said means for directing the computer system to perform a user-assisted verification and correction comprises means, provided on the computer-readable medium, for requiring user input on critical characters of said characters.