Optical character recognition system having context analyzer

• US 6,577,755 B1
• Filed: 09/26/1997
• Issued: 06/10/2003
• Est. Priority Date: 10/18/1994
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Lorie, Raymond Amand
• Primary Examiner(s): Patel, Jayanti K.
• Application Number:
  US08/938,044
• Time in Patent Office: 2,083 Days
• Field of Search: 382/155-161, 382/181, 382/185-190, 382/224-231, 382/317, 382/321, 382/312, 382/140, 364/822, 704/1-10, 704/231, 704/232, 704/259, 704/257, 704/258, 704/9, 706/1-12, 706/15, 706/16, 706/20, 706/25, 707/9
• US Class Current: 382/140
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/274 Syntactic or semantic context, e.g. balancing

Abstract

An optical character recognition (OCR) system is provided, in which syntactical and semantic rules, provided along with an input image to be scanned and applicable to the contents of the scanned image, are used in connection with the results of the OCR scan to identify the scanned characters. As a result, the recognition rate and confidence are enhanced. By providing the checking based on syntactical and semantic rules within the OCR system, application programs which would receive and use the OCR results are freed from the added burden of having to perform their own syntactical and/or semantic checking on the OCR results the application programs receive from the OCR system.

Claims

1. An optical character recognition (OCR) system comprising:

means for producing a scan of an input image of text to be recognized;

a context analyzer coupled to receive the scan, for checking the scan for consistency with a predetermined text content constraint, the predetermined text content constraint including a syntactical constraint and a semantic constraint;

user input means for accepting user-selected text content constraints, including a character-based syntactical constraint and a character-based semantic constraint, the user input means including a document specification language which serves as a user interface to allow the user to enter the user selected text content constraints;

syntax means for checking the preliminary scan for consistency with the character-based syntactical constraint; and

semantics means for checking the preliminary scan for consistency with the character-based semantic constraint.

2. An OCR system as recited in claim 1, wherein the semantics means is operative responsive to completion of operation of the systax means.

3. An OCR system as recited in claim 1, wherein:

the syntax means includes;

(i) means for receiving a document description, programmed in the document specification language, pertaining to the text to be recognized, and

(ii) means for compiling the document description to produce a context structure; and

the context analyzer includes means for checking the preliminary scan for consistency with the context structure.

4. An OCR system as recited in claim 1, wherein:

the semantics means includes a library of semantic routines; and

the context analyzer includes means for checking the preliminary scan for consistency with the semantic routines.

5. An OCR system as recited in claim 4, further comprising means for facilitating modification of the library of semantic routines by a user of the OCR system.

6. An OCR system as recited in claim 1, wherein the document specification language includes instructions for defining at least one of:

(i) a field within the text to be recognized,

(ii) a character type for characters occurring within the field,

(iii) an alphabet, and

(iv) a representation of a sequence of characters in terms of types of the characters of the sequence.

7. An OCR system as recited in claim 1, further comprising an object oriented buffer coupled to the context analyzer for passing values of variables.

8. An OCR system as recited in claim 1, wherein:

the OCR system further comprises a dictionary containing a set of valid text items; and

the context analyzer includes means for performing a fuzzy search through the dictionary to identify best matching values, among the text items.

9. An OCR system as recited in claim 1, wherein:

the OCR system includes a plurality of dictionaries containing sets of valid text items for respective fields of the text to be recognized; and

the context analyzer includes;

(i) means for performing respective fuzzy searches through the dictionaries to identify best matching values, among the text items, for respective fields of the text to be recognized, and

(ii) means for comparing the best matching values of the respective fuzzy searches to identify a best combination of the best matching values.

10. An OCR system as recited in claim 1, further comprising:

a recognition engine coupled to the context analyzer for performing an initial character recognition procedure on the image; and

an object oriented buffer coupled to the recognition engine for receiving and storing results of the initial character recognition procedure in a predetermined structure, for storing results of the context analyzer, and for providing updatable data to the semantic means.

11. The OCR system in claim 1, wherein said text comprises numerical and alphabetic words, each including at least one character, said character-based syntactical constraint and said character-based semantic constraint comprising formatting and logical constraints on said characters.

12. The OCR system in claim 1, wherein said text comprises numerical and alphabetic words, each including at least one character, said character-based syntactical constraint and said character-based semantic constraint comprising input field controls of said characters.

13. The OCR system in claim 1, wherein a logical operation of said context analyzer is uneffected by said user input means.

14. A method for performing optical character recognition (OCR) on an image to recognize text in the image, the method comprising the steps of:

receiving, as a user input, syntax and semantic definitions of an expected content of the field of the image, the user input being given in terms of user-selected text content constraints, including a character-based syntactical constraint and a character-based semantic constraint, the step of receiving including receiving the user-selected text content constraints expressed in a user-programmable document specification language which serves as a user interface to allow the user to enter the user-selected text content constraints;

using the character-based syntactical constraint to determine character types that are relevant to the syntax definitions of the field;

operating a recognition engine on the image to produce character hypotheses of a content of the image, and probability values for the character hypotheses;

converting the character hypotheses into a character type hypotheses;

using the character-based semantic constraint to enumerate possible models for the image content based on the character type hypotheses and on the probability values;

replacing the character type hypotheses with character values to produce a set of solutions; and

selecting one of the solutions as the recognized text.

15. The method in claim 14, wherein said text comprises numerical and alphabetic words, each including at least one character, said character-based syntactical constraint and said character-based semantic constraint comprising formatting and logical constraints on said characters.

16. The method in claim 14, wherein said text comprises numerical and alphabetic words, each including at least one character, said character-based syntactical constraint and said character-based semantic constraint comprising input field controls of said characters.

17. A computer program product, for use with a processing system, for directing the processing system to perform optical character recognition (OCR) on an image to recognize text in the image, the computer program product comprising:

a recording medium;

means, recorded on the recording medium, for directing the processing system to receive, as a user input, syntax and semantic definitions of an expected content of the field of the image, the user input being given in terms of user-selected text content constraints, including a character-based syntactical constraint and a character-based semantic constraint, the means for directing to receive including a document specification language which serves as a user interface to allow the user to enter the user-selected text content constraints;

means, recorded on the recording medium, for directing the processing system to use the character-based syntactical constraint to determine character types that are relevant to the syntax definitions of the field;

means, recorded on the recording medium, for directing the processing system to operate a recognition engine on the image to produce character hypotheses of a content of the image, and probability values for the character hypotheses;

means, recorded on the recording medium, for directing the processing system to convert the character hypotheses into a character type hypotheses;

means, recorded on the recording medium, for directing the processing system to use the character-based semantic constraint to enumerate, possible models for the image content based on the character type hypotheses and on the probability values;

means, recorded on the recording medium, for directing the processing system to replace the character type hypotheses with character values to produce a set of solutions; and

means, recorded on the recording medium, for directing the processing system to select one of the solutions as the recognized text.

18. The computer program product in claim 17, wherein said text comprises numerical and alphabetic words, each including at least one character, said character-based syntactical constraint and said character-based semantic constraint comprising formatting and logical constraints on said characters.

19. The computer program product in claim 17, wherein said text comprises numerical and alphabetic words, each including at least one character, said character-based syntactical constraint and said character-based semantic constraint comprising input field controls of said characters.