Automated categorization and assembly of low-quality images into electronic documents

1. An apparatus comprising:

a memory configured to store:

an optical character recognition (OCR) algorithm; and

a natural language processing (NLP) algorithm;

a hardware processor communicatively coupled to the memory, the hardware processor configured to:

receive an image of a page of a physical document;

convert, by executing the OCR algorithm, the image into a set of text;

identify one or more errors in the set of text, the one or more errors associated with noise in the image, wherein each error of the one or more errors is assigned to an error type of a plurality of error types;

generate a feature vector from the set of text, the feature vector comprising:

a first plurality of features obtained by executing the NLP algorithm on the set of text; and

a second plurality of features, wherein each feature of the second plurality of features is associated with an error type of the plurality of error types and provides a measure of a quantity of errors of the one or more errors that are assigned to the associated error type;

assign, based on the feature vector, the image to a first document category of a set of document categories, wherein:

documents assigned to the first document category share one or more characteristics; and

the feature vector is associated with a probability that the physical document associated with the image comprises the one or more characteristics, wherein the probability is greater than a threshold; and

in response to assigning the image to the first document category, store the image in a database as a page of an electronic document belonging to the first document category.

2. The apparatus of claim 1 , wherein at least one error type of the plurality of error types is associated with at least one of:

non-ascii characters;

non-English words;

stray letters other than “a” and “I”;

numbers comprising one or more letters; and

misplaced punctuation marks.

3. The apparatus of claim 1 , wherein:

the image comprises at least one of:

a scanned image generated by scanning the page of the physical document; and

a faxed image generated by faxing the page of the physical document; and

the noise in the image is associated with at least one of:

ruled lines on the page of the physical document;

hole punches on the page of the physical document;

uneven contrast in the image;

one or more folds in the page of the physical document;

handwritten text in a margin of the page of the physical document;

a background color of the image; and

interfering strokes in the image, wherein the physical document comprises double-sided pages.

4. The apparatus of claim 1 , wherein:

the feature vector further comprises a third plurality of features; and

for each error of the one or more identified errors, the third plurality of features comprises an identification of a line number and a character offset of the error in the set of text.

5. The apparatus of claim 1 , wherein:

the physical document comprises two or more pages; and

the hardware processor is further configured to:

receive one or more images of additional pages of the physical document; and

identify a page number in the set of text, wherein storing the image in the database as the page of the electronic document comprises storing the image at a position within the electronic document corresponding to the identified page number.

6. The apparatus of claim 1 , wherein the hardware processor is further configured to:

receive a second image of a page of a second physical document;

convert, by executing the OCR algorithm, the second image into a second set of text;

identify one or more errors in the second set of text, the one or more errors in the second set of text associated with noise in the second image, wherein each error of the one or more errors in the second set of text is assigned to an error type of the plurality of error types;

generate a second feature vector from the second set of text, the second feature vector comprising:

a first plurality of features obtained by executing the NLP algorithm on the second set of text; and

a second plurality of features, wherein each feature of the second plurality of features is associated with an error type of the plurality of error types and provides a measure of a quantity of errors of the one or more identified errors in the second set of text that are assigned to the associated error type;

determine, based on the second feature vector, a set of probabilities, wherein each probability of the set of probabilities is associated with a document category of the set of document categories and corresponds to a probability that the second document belongs to the associated document category;

determine that each probability of the set of probabilities is less than a second threshold; and

in response to determining that each probability of the set of probabilities is less than the second threshold, transmit the second image to a device of a system administrator for manual classification.

7. The apparatus of claim 1 , wherein the hardware processor is further configured to:

receive an identification number assigned to the image;

use the identification number to obtain, from a second database, a name of an individual associated with the physical document, wherein the name of the individual is linked to the identification number in the second database;

identify a name in the set of text; and

verify that the name identified in the set of text corresponds to the name of the individual.

8. A method comprising:

receiving an image of a page of a physical document;

converting, by executing an optical character recognition algorithm, the image into a set of text;

identifying one or more errors in the set of text, the one or more errors associated with noise in the image, wherein each error of the one or more errors is assigned to an error type of a plurality of error types;

generating a feature vector from the set of text, the feature vector comprising:

a first plurality of features obtained by executing a natural language processing algorithm on the set of text; and

a second plurality of features, wherein each feature of the second plurality of features is associated with an error type of the plurality of error types and provides a measure of a quantity of errors of the one or more errors that are assigned to the associated error type;

assigning, based on the feature vector, the image to a first document category of a set of document categories, wherein:

documents assigned to the first document category share one or more characteristics; and

the feature vector is associated with a probability that the physical document associated with the image comprises the one or more characteristics, wherein the probability is greater than a threshold; and

in response to assigning the image to the first document category, storing the image in a database as a page of an electronic document belonging to the first document category.

9. The method of claim 8 , wherein at least one error type of the plurality of error types is associated with at least one of:

non-ascii characters;

non-English words;

stray letters other than “a” and “I”;

numbers comprising one or more letters; and

misplaced punctuation marks.

10. The method of claim 8 , wherein:

the image comprises at least one of:

a scanned image generated by scanning the page of the physical document; and

a faxed image generated by faxing the page of the physical document; and

the noise in the image is associated with at least one of:

ruled lines on the page of the physical document;

hole punches on the page of the physical document;

uneven contrast in the image;

one or more folds in the page of the physical document;

handwritten text in a margin of the page of the physical document;

a background color of the image; and

interfering strokes in the image, wherein the physical document comprises double-sided pages.

11. The method of claim 8 , wherein:

the feature vector further comprises a third plurality of features; and

for each error of the one or more identified errors, the third plurality of features comprises an identification of a line number and a character offset of the error in the set of text.

12. The method of claim 8 , wherein:

the physical document comprises two or more pages; and

the method further comprises:

receiving one or more images of additional pages of the physical document; and

identifying a page number in the set of text, wherein storing the image in the database as the page of the electronic document comprises storing the image at a position within the electronic document corresponding to the identified page number.

13. The method of claim 8 , further comprising:

receiving a second image of a page of a second physical document;

converting, by executing the OCR algorithm, the second image into a second set of text;

identifying one or more errors in the second set of text, the one or more errors in the second set of text associated with noise in the second image, wherein each error of the one or more errors in the second set of text is assigned to an error type of the plurality of error types;

generating a second feature vector from the second set of text, the second feature vector comprising:

a first plurality of features obtained by executing the NLP algorithm on the second set of text; and

a second plurality of features, wherein each feature of the second plurality of features is associated with an error type of the plurality of error types and provides a measure of a quantity of errors of the one or more identified errors in the second set of text that are assigned to the associated error type;

determining, based on the second feature vector, a set of probabilities, wherein each probability of the set of probabilities is associated with a document category of the set of document categories and corresponds to a probability that the second document belongs to the associated document category;

determining that each probability of the set of probabilities is less than a second threshold; and

in response to determining that each probability of the set of probabilities is less than the second threshold, transmitting the second image to a device of a system administrator for manual classification.

14. The method of claim 8 , further comprising:

receiving an identification number assigned to the image;

using the identification number to obtain, from a second database, a name of an individual associated with the physical document, wherein the name of the individual is linked to the identification number in the second database;

identifying a name in the set of text; and

verifying that the name identified in the set of text corresponds to the name of the individual.

15. A system comprising:

a database;

a memory configured to store a set of document categories; and

a hardware processor communicatively coupled to the memory, the hardware processor configured to:

receive an image of a page of a physical document;

convert the image into a set of text;

identify one or more errors in the set of text, the one or more errors associated with noise in the image, wherein each error of the one or more errors is assigned to an error type of a plurality of error types;

generate a feature vector from the set of text, the feature vector comprising a plurality of features, one or more features of the plurality of features associated with an error type of the plurality of error types and providing a measure of a quantity of errors of the one or more errors that are assigned to the associated error type;

assign, based on the feature vector, the image to a first document category of a set of document categories, wherein:

documents assigned to the first document category share one or more characteristics; and

the feature vector is associated with a probability that the physical document associated with the image comprises the one or more characteristics, wherein the probability is greater than a threshold; and

in response to assigning the image to the first document category, store the image in the database as a page of an electronic document belonging to the first document category.

16. The system of claim 15 , wherein at least one error type of the plurality of error types is associated with at least one of:

non-ascii characters;

non-English words;

stray letters other than “a” and “I”;

numbers comprising one or more letters; and

misplaced punctuation marks.

17. The system of claim 15 , wherein:

the image comprises at least one of:

a scanned image generated by scanning the page of the physical document; and

a faxed image generated by faxing the page of the physical document; and

the noise in the image is associated with at least one of:

ruled lines on the page of the physical document;

hole punches on the page of the physical document;

uneven contrast in the image;

one or more folds in the page of the physical document;

handwritten text in a margin of the page of the physical document;

a background color of the image; and

interfering strokes in the image, wherein the physical document comprises double-sided pages.

18. The system of claim 15 , wherein:

the physical document comprises two or more pages; and

the hardware processor is further configured to:

receive one or more images of additional pages of the physical document; and

identify a page number in the set of text, wherein storing the image in the database as the page of the electronic document comprises storing the image at a position within the electronic document corresponding to the identified page number.

19. The system of claim 15 , wherein the hardware processor is further configured to:

receive a second image of a page of a second physical document;

convert the second image into a second set of text;

identify one or more errors in the second set of text, the one or more errors in the second set of text associated with noise in the second image, wherein each error of the one or more errors in the second set of text is assigned to an error type of the plurality of error types;

generate a second feature vector from the second set of text, the second feature vector comprising a second plurality of features, one or more features of the second plurality of features associated with an error type of the plurality of error types and providing a measure of a quantity of errors of the one or more identified errors in the second set of text that are assigned to the associated error type;

determine, based on the second feature vector, a set of probabilities, wherein each probability of the set of probabilities is associated with a document category of the set of document categories and corresponds to a probability that the second document belongs to the associated document category;

determine that each probability of the set of probabilities is less than a second threshold; and

in response to determining that each probability of the set of probabilities is less than the second threshold, transmit the second image to a device of a system administrator for manual classification.

20. The system of claim 15 , further comprising a second database configured to store:

a set of names; and

a set of identification numbers, wherein:

each name of the set of names is linked to an identification number of the set of identification numbers; and

the hardware processor is further configured to:

receive a first identification number assigned to the image;

locate the first identification number in the second database;

in response to locating the first identification number in the second database, identify a first name linked to the first identification number in the second database;

identify a second name from the set of text; and

verify that the second name identified in the set of text corresponds to the first name.