Optical Character Recognition Error Correction Model

Patent metadata

Field	Value
Publication number	US-2021174109-A1
Application number	US-201916702693-A
Country	US
Kind code	A1
Filing date	Dec 4, 2019
Priority date	Dec 4, 2019
Publication date	Jun 10, 2021
Grant date	—

Abstract

Embodiments relate to an intelligent computer platform to create a document specific error correction model for amending OCR values. An image of a document is received and OCR is applied to the received image. Text is extracted from at least one static content field and the extracted text is compared to stored text from known static content. Responsive to a deviation identified in the comparison, a document specific error correction model is created and leveraged to correct OCR output. The model generates one or more variants for the dynamic content field associated with the static content field having the identified deviation. The generated variant(s) is subject to processing and one of the variants is selected as amended document content. A new document version is created from the amendment, the new document version including corrected OCR output.

First claim

What is claimed is:

1 . A computer system comprising:
a processing unit operatively coupled to memory; an artificial intelligence platform in communication with the processing unit, the platform including one or more tools to identify and amend one or more optical character recognition (OCR) values, including:
a document manager to receive an image of a document, wherein the document has at least one static content field and at least one dynamic content field;
an extraction manager, operatively coupled to the document manager, the extraction manager to apply OCR to the received image, extract text from the at least one static content field, and compare the extracted text to stored text from known static content;
a model manager, operatively coupled to the extraction manager, the model manager, responsive to a deviation identified in the comparison, to create a document specific error correction model, and leverage the error correction model to correct OCR output;
the model manager to employ the document specific error correction model to generate one or more variants for the dynamic content field associated with the static content field having the identified deviation;
the model manager to subject the one or more generated variants to processing, and select one of the one or more generated variants responsive to the processing; and
a director, operatively coupled to the model manager, to create a new document version from the selective amendment, the new document version including the selected one or more variants as corrected OCR output.

2 . The computer system of claim 1 , further comprising the extraction manager to utilize artificial intelligence to identify one or more updates to the known static content.

3 . The computer system of claim 2 , wherein the error correction model is trained using string data from the static content field to predict the one or more variants for string data for the dynamic content field.

4 . The computer system of claim 2 , further comprising the model manager to file the generated one or more variants based on a restricted character class assigned to the static content field.

5 . The computer system of claim 2 , wherein subjecting the one or more variants to processing further comprises the model manager to assess a score for each of the one or more variants, wherein the assessed score is responsive to domain based keyword matching associated with the static content field.

6 . The computer system of claim 1 , further comprising the director to selectively amend the static content field having the identified deviation, the selective amendment aligned with the selected variant.

7 . A computer program product to identify and amend one or more optical character recognition (OCR) values, the computer program product comprising a computer readable storage medium having program code embodied therewith, the program code executable by a processor to:
receive an image of a document, wherein the document has at least one static content field and at least one dynamic content field; apply optical character recognition (OCR) to the received image, extract text from the at least one static content field, and compare the extracted text to stored text from known static content; responsive to a deviation identified in the comparison, create a document specific error correction model, and leverage the error correction model to correct OCR output; the error correction model to generate one or more variants for the dynamic content field associated with the static content field having the identified deviation; subject the one or more generated variants to processing, and select one of the one or more generated variants responsive to the processing; selectively amend the static content field having the identified deviation, the selective amendment aligned with the selected variant; and create a new document version from the selective amendment, the new document version including the selected one or more variants as corrected OCR output.

8 . The computer program product of claim 7 , wherein the error correction model utilizes artificial intelligence (AI) to identify one or more updates to the static content field.

9 . The computer program product of claim 8 , wherein the error correction model is trained using string data from the static content field to predict the one or more variants for string data for the dynamic content field.

10 . The computer program product of claim 8 , further comprising program code to filter the generated one or more variants based on a restricted character class assigned to the static content field.

11 . The computer program product of claim 8 , wherein subjecting the one or more variants to processing further comprises program code to assess a score for each of the one or more variants, wherein the assessed score is responsive to domain based keyword matching associated with the static content field.

12 . The computer program product of claim 7 , further comprising program code to selectively amend the static content field having the identified deviation, the selective amendment aligned with the selected variant.

13 . A method comprising:
receiving an image of a document, wherein the document has at least one static content field and at least one dynamic content field; applying optical character recognition (OCR) to the received image, extracting text from the at least one static content field, and comparing the extracted text to stored text from known static content; responsive to a deviation identified in the comparison, creating a document specific error correction model, and leveraging the error correction model to correct OCR output; the error correction model generating one or more variants for the dynamic content field associated with the static content field having the identified deviation; subjecting the one or more generated variants to processing, and selecting one of the one or more generated variants responsive to the processing; selectively amending the static content field having the identified deviation, the selective amendment aligned with the selected variant; and creating a new document version from the selective amendment, the new document version including the selected one or more variants as corrected OCR output.

14 . The method of claim 13 , wherein the error correction model utilizes artificial intelligence (AI) to identify one or more updates to the static content field.

15 . The method of claim 14 , wherein the error correction model is trained using string data from the static content field to predict the one or more variants for string data for the dynamic content field.

16 . The method of claim 14 , further comprising filtering the generated one or more variants based on a restricted character class assigned to the static content field.

17 . The method of claim 14 , wherein subjecting the one or more variants to processing further comprises assessing a score for each of the one or more variants, wherein the assessed score is responsive to domain based keyword matching associated with the static content field.

18 . The method of claim 13 , further comprising selectively amending the static content field having the identified deviation, the selective amendment aligned with the selected variant.

Assignees

• IBM

Inventors

• Beller Charles E
• Katz Edward Graham
• Riendeau John A
• Phipps Christopher
• Thatcher Sean Thomas

Classifications

• G06V30/40

  Document-oriented image-based pattern recognition · CPC title

• G06V30/12 Primary

  Detection or correction of errors, e.g. by rescanning the pattern · CPC title

• G06V30/19173

  Classification techniques · CPC title

• G06V30/10

  Character recognition · CPC title

• G06V30/416

  Extracting the logical structure, e.g. chapters, sections or page numbers; Identifying elements of the document, e.g. authors · CPC title