US 9,053,350 B1 · Efficient identification and correction of optical character recognition errors through learning in a multi-engine environment

Abstract

OCR errors are identified and corrected through learning. An error probability estimator is trained using ground truths to learn error probability estimation. Multiple OCR engines process a text image, and convert it into texts. The error probability estimator compares the outcomes of the multiple OCR engines for mismatches, and determines an error probability for each of the mismatches. If the error probability of a mismatch exceeds an error probability threshold, a suspect is generated and grouped together with similar suspects in a cluster. A question for the cluster is generated and rendered to a human operator for answering. The answer from the human operator is then applied to all suspects in the cluster to correct OCR errors in the resulting text. The answer is also used to further train the error probability estimator.

Claims20

Dependent claims

A method of correcting Optical Character Recognition (OCR) errors, comprising:

converting a graphical representation of an original segment of text from a document into a first segment of text using a first OCR engine;
converting the graphical representation of the original segment of text into a second segment of text using a second OCR engine;
estimating an error probability for the first segment of text by matching the first segment of text with the second segment of text;
responsive to the error probability exceeding a predetermined threshold value, generating a suspect interpretation, for the first segment of text, the suspect interpretation associated with a suspected OCR error;
grouping the suspect interpretation with other similar suspect interpretations to form a cluster of suspect interpretations that are each associated with one or more occurrences of the suspected OCR error in the document;
generating a representative question for the cluster;
determining a correct answer for the representative question;
and
generating a corrected segment of text based on the correct answer, the corrected segment of text matching the original segment of text.

6 more elements

A non-transitory computer-readable storage medium containing executable computer program code for performing a method of correcting Optical Character Recognition (OCR) errors comprising:

converting a graphical representation of an original segment of text from a document into a first segment of text using a first OCR engine;
converting the graphical representation of the original segment of text into a second segment of text using a second OCR engine;
estimating an error probability for the first segment of text by matching the first segment of text with the second segment of text;
responsive to the error probability exceeding a predetermined threshold value, generating a suspect interpretation, for the first segment of text, the suspect interpretation associated with a suspected OCR error;
grouping the suspect interpretation with other similar suspect interpretations to form a cluster of suspect interpretations that are each associated with one or more occurrences of the suspected OCR error in the document;
generating a representative question for the cluster;
determining a correct answer for the representative question;
and
generating a corrected segment of text based on the correct answer, the corrected segment of text matching the original segment of text.

6 more elements

A computer system for correcting Optical Character Recognition (OCR) errors, the computer system comprising:

a non-transitory computer-readable storage medium storing executable computer program instructions for correcting Optical Character Recognition (OCR) errors, the computer program instructions comprising instructions for:
converting a graphical representation of an original segment of text from a document into a first segment of text using a first OCR engine;
converting the graphical representation of the original segment of text into a second segment of text using a second OCR engine;
estimating an error probability for the first segment of text by matching the first segment of text with the second segment of text;
responsive to the error probability exceeding a predetermined threshold value, generating a suspect interpretation, for the first segment of text, the suspect interpretation associated with a suspected OCR error;
grouping the suspect interpretation with other similar suspect interpretations to form a cluster of suspect interpretations that are each associated with one or more occurrences of the suspected OCR error in the document;
generating a representative question for the cluster;
determining a correct answer for the representative question;
and
generating a corrected segment of text based on the correct answer, the corrected segment of text matching the original segment of text;
and
a processor configured to execute the computer program instructions stored on the computer-readable storage medium.

9 more elements

Dependencies are inferred from the claim text.

Efficient identification and correction of optical character recognition errors through learning in a multi-engine environment

Life of this patent

Efficient identification and correction of optical character recognition errors through learning in a multi-engine environment

Life of this patent

Abstract

Claims20