Granted Patent B2

US 12,314,661 · App. 18/082,919 · Granted May 27, 2025

Inventor: Michael Zatsepin (Novokuznetsk, RU)

Assignee: ABBYY Development Inc.

G06F40/284G06F40/263G06V30/153

Abstract

Claims (72)

1. A method, comprising:

identifying, by a processing device, a document comprising a plurality of words in one or more natural languages;

for each word of at least a subset of words of the document:

generating a plurality of sets of tokens representing the word, wherein each set of tokens of the plurality of sets of tokens represents the word using a corresponding plurality of tokens defined for a corresponding natural language of a set of natural languages, and

identifying, based on the plurality of sets of tokens, a primary natural language associated with the word;

associating each natural language of the set of natural languages with a corresponding word count indicating a number of words of the subset of words for which the natural language has been identified as the primary natural language;

identifying, among the set of natural languages, a natural language associated with a maximum word count; and

associating the identified natural language with the document.

2. The method of claim 1 , further comprising:

identifying, based on the plurality of sets of tokens, an alternative natural language associated with the word.

3. The method of claim 1 , further comprising:

iteratively performing the operations of:

updating the subset of words of the document by removing one or more words associated with the identified natural language,

associating each natural language of the set of natural languages with a corresponding word count indicating a number of words of the updated subset of words for which the natural language has been identified as the primary natural language,

identifying, among the set of natural languages, a natural language associated with a maximum word count, and

associating the identified natural language with the document.

4. The method of claim 3 , wherein the operations are iteratively performed until a number of words of the document that associated with at least one language reaches a certain high threshold.

5. The method of claim 3 , wherein the operations are iteratively performed until a number of words for which a language that has been identified as a primary or alternative language by a current iteration falls below a certain low threshold.

6. The method of claim 1 , further comprising:

performing, based on the identified natural language, a natural language processing task with respect to the document.

7. The method of claim 1 , wherein identifying the document further comprises:

performing optical character recognition (OCR) of an image.

8. The method of claim 1 , wherein the subset of words is identified by applying one or more filtering criteria to a plurality of words comprised by the document.

9. The method of claim 1 , wherein identifying the document further comprises:

splitting an input image into a plurality of portions;

removing one or more portions satisfying one or more geometric criteria;

performing optical character recognition (OCR) of at least a subset of remaining portions.

10. A system comprising:

a memory; and

a processing device operatively coupled to the memory, the processing device configured to:

identify a document comprising a plurality of words in one or more natural languages;

for each word of at least a subset of words of the document:

generate a plurality of sets of tokens representing the word, wherein each set of tokens of the plurality of sets of tokens represents the word using a corresponding plurality of tokens defined for a corresponding natural language of a set of natural languages, and

identify, based on the plurality of sets of tokens, a primary natural language associated with the word;

associate each natural language of the set of natural languages with a corresponding word count indicating a number of words of the subset of words for which the natural language has been identified as the primary natural language;

identify, among the set of natural languages, a natural language associated with a maximum word count; and

associate the identified natural language with the document.

11. The system of claim 10 , wherein the processing device is further configured to:

identify, based on the plurality of sets of tokens, an alternative natural language associated with the word.

12. The system of claim 10 , wherein the processing device is further configured to:

iteratively perform the operations of:

updating the subset of words of the document by removing one or more words associated with the identified natural language,

associating each natural language of the set of natural languages with a corresponding word count indicating a number of words of the updated subset of words for which the natural language has been identified as the primary natural language,

identifying, among the set of natural languages, a natural language associated with a maximum word count, and

associating the identified natural language with the document.

13. The system of claim 10 , wherein the processing device is further configured to:

perform, based on the identified natural language, a natural language processing task with respect to the document.

14. The system of claim 10 , wherein identifying the document further comprises:

performing optical character recognition (OCR) of an image.

15. The system of claim 10 , wherein the subset of words is identified by applying one or more filtering criteria to a plurality of words comprised by the document.

16. The system of claim 10 , wherein identifying the document further comprises:

splitting an input image into a plurality of portions;

removing one or more portions satisfying one or more geometric criteria;

performing optical character recognition (OCR) of at least a subset of remaining portions.

17. A non-transitory computer-readable storage medium including executable instructions that, when executed by a processing device, cause the processing device to:

identify a document comprising a plurality of words in one or more natural languages;

for each word of at least a subset of words of the document:

generate a plurality of sets of tokens representing the word, wherein each set of tokens of the plurality of sets of tokens represents the word using a corresponding plurality of tokens defined for a corresponding natural language of a set of natural languages, and

identify, based on the plurality of sets of tokens, a primary natural language associated with the word;

associate each natural language of the set of natural languages with a corresponding word count indicating a number of words of the subset of words for which the natural language has been identified as the primary natural language;

identify, among the set of natural languages, a natural language associated with a maximum word count; and

associate the identified natural language with the document.

18. The non-transitory computer-readable storage medium of claim 17 , further comprising executable instructions that, when executed by the processing device, cause the processing device to:

identify, based on the plurality of sets of tokens, an alternative natural language associated with the word.

19. The non-transitory computer-readable storage medium of claim 17 , further comprising executable instructions that, when executed by the processing device, cause the processing device to:

iteratively perform the operations of:

updating the subset of words of the document by removing one or more words associated with the identified natural language,

associating each natural language of the set of natural languages with a corresponding word count indicating a number of words of the updated subset of words for which the natural language has been identified as the primary natural language,

identifying, among the set of natural languages, a natural language associated with a maximum word count, and

associating the identified natural language with the document.

20. The non-transitory computer-readable storage medium of claim 17 , further comprising executable instructions that, when executed by the processing device, cause the processing device to:

perform, based on the identified natural language, a natural language processing task with respect to the document.

Assignments (2)

Continuity (1)

References Cited (10)