Text entity recognition
View Patent ↗Various embodiments enable the identification of semi-structured text entities in an imager. The identification of the text entities is a relatively simple problem when the text is stored in a computer and free of errors, but much more challenging if the source is the output of an optical character recognition (OCR) engine from a natural scene image. Accordingly, output from an OCR engine is analyzed to isolate a character string indicative of a text entity. Each character of the string is then assigned to a character class to produce a character class string and the text entity of the string is identified based in part on a pattern of the character class string.
1. A computer-implemented method, comprising:
receiving an output from an optical character recognition (OCR) performed on an image;
determining, from the output, first values associated with a first pixel region of the image and second values associated with a second pixel region of the image;
determining a character string indicative of a text entity from the output when one or more differences between the first values and the second values exceeds a threshold;
assigning each character of the character string to a character class to produce a character class string;
identifying the character string as being the text entity;
identifying a text entity type associated with the text entity; and
autocorrecting at least one character in the character string to provide a corrected character, the at least one character belonging to a character class not associated with the text entity type.
2. The computer-implemented method of claim 1 , wherein autocorrecting at least one character in the character string includes replacing the at least one character belonging to the character class not associated with the text entity with at least one second character belonging to a second character class associated with the text entity.
3. The computer-implemented method of claim 1 , wherein the analyzing of the output isolates the character string by performing one or more heuristic tests.
4. The computer-implemented method of claim 3 , further comprising:
converting the output into text lines; and
omitting characters that do not fit a pattern indicative of the text entity.
5. The computer-implemented method of claim 1 , wherein identifying the character string as being the text entity includes:
determining a matching score for the character string, the matching score being based at least in part on a number of edits made to the character class string.
6. The computer-implemented method of claim 5 , wherein determining the matching score includes:
assigning costs to edits made to the character class string, wherein the costs are based on a similarity between characters.
7. The computer-implemented method of claim 6 , wherein the costs assigned to the edits made to the character class string are represented in an N by M matrix where N represents a number possible ASCII characters and M represents a number of possible character classes.
8. A computer-implemented method, comprising:
receiving an output from an optical character recognition (OCR) performed on an image;
determining, from the output, first values associated with a first pixel region of the image and second values associated with a second pixel region of the image;
determining a character string indicative of a text entity from the output when one or more differences between the first values and the second values exceeds a threshold;
assigning each character of the character string to a character class to produce a character class string;
identifying the character string as being the text entity; and
validating the character string by comparing the character string to text entity-specific patterns, wherein the validating provides at least one corrected character for the character string.
9. The computer-implemented method of claim 8 , wherein validating the character string by comparing the character string to text entity-specific patterns includes:
determining that the text entity is associated with a pattern indicative of a URL; and
comparing the character string to a determined number of URLs.
10. The computer-implemented method of claim 8 , wherein the analyzing of the output isolates the character string by performing one or more heuristic tests.
11. The computer-implemented method of claim 10 , further comprising:
converting the output into text lines; and
omitting characters that do not fit a pattern indicative of the text entity.
12. The computer-implemented method of claim 8 , wherein identifying the character string as being the text entity includes:
determining a matching score for the character string, the matching score being based at least in part on a number of edits made to the character class string.
13. The computer-implemented method of claim 12 , wherein determining the matching score includes:
assigning costs to edits made to the character class string, wherein the costs are based on a similarity between characters.
14. The computer-implemented method of claim 13 , wherein the costs assigned to the edits made to the character class string are represented in an N by M matrix where N represents a number possible ASCII characters and M represents a number of possible character classes.
15. A computing device, comprising:
a processor;
a display screen; and
memory including instructions that, when executed by the processor, cause the computing device to:
receive an output from an optical character recognition (OCR) performed on an image;
determine, from the output, first values associated with a first pixel region of the image and second values associated with a second pixel region of the image;
determine a character string indicative of a text entity from the output when one or more differences between the first values and the second values exceeds a threshold;
assign each character of the character string to a character class to produce a character class string;
identify the character string as being the text entity; and
validate the character string by comparing the character string to text entity-specific patterns, wherein the validation provides at least one corrected character for the character string.
16. The computing device of claim 15 , wherein the instructions, when executed to validate the character string by comparing the character string to text entity-specific patterns, further enables the computing device to:
determine that the text entity is associated with a pattern indicative of a URL; and
compare the character string to a determined number of URLs.
17. The computing device of claim 15 , wherein the instructions, when executed by the processor, further cause the computing device to:
identify a text entity type associated with the text entity; and
autocorrect at least one character in the character string, the at least one character belonging to a character class not associated with the text entity type.
18. The computing device of claim 15 , wherein the instructions, when executed to identify the character string as being the text entity, further enables the computing device to:
determine a matching score for the character string, the matching score being based at least in part on a number of edits made to the character class string.
19. The computing device of claim 18 , wherein the instructions, when executed to determine the matching score, further enables the computing device to:
assign costs to edits made to the character class string, wherein the costs are based on a similarity between characters.
20. The computing device of claim 19 , wherein the costs assigned to the edits made to the character class string are represented in an N by M matrix where N represents a number possible ASCII characters and M represents a number of possible character classes.