Character acquisition, page processing and knowledge graph construction method and device, medium

Patent metadata

Field	Value
Publication number	US-12131809-B2
Application number	US-201916982185-A
Country	US
Kind code	B2
Filing date	Nov 25, 2019
Priority date	Nov 25, 2019
Publication date	Oct 29, 2024
Grant date	Oct 29, 2024

Abstract

A method and a device for acquiring character information in a picture, a non-transitory storage medium, a page processing method, and a knowledge graph construction method are disclosed. The method for acquiring character information in a picture includes: acquiring a picture and extracting at least one piece of character information in the picture; and checking-and-correcting the at least one piece of character information based on a knowledge graph.

First claim

What is claimed is:

1. A method for acquiring character information in a picture, comprising:
acquiring a picture and extracting at least one piece of character information in the picture; and
checking-and-correcting the at least one piece of character information based on a knowledge graph,
wherein the checking-and-correcting the at least one piece of character information based on a knowledge graph comprises:
identifying character information having an error in the at least one piece of character information based on the knowledge graph; and
correcting the character information having an error based on the knowledge graph,

wherein the at least one piece of character information comprises a plurality of pieces of character information; and
the identifying character information having an error in the at least one piece of character information based on the knowledge graph comprises:
obtaining a plurality of entities respectively based on the plurality of pieces of character information in the picture, and selecting an entity from the plurality of entities as a to-be-checked-and-corrected entity for a process of determining whether character information corresponding to the to-be-checked-and-corrected entity has an error; and
determining whether the character information corresponding to the to-be-checked-and-corrected entity has an error according to a hierarchical structure of the knowledge graph, and identifying the character information corresponding to the to-be-checked-and-corrected entity as the character information having an error when the character information corresponding to the to-be-checked-and-corrected entity has an error,

wherein the determining whether the character information corresponding to the to-be-checked-and-corrected entity has an error according to a hierarchical structure of the knowledge graph comprises:
grading the plurality of entities according to the hierarchical structure of the knowledge graph;
determining a level of the to-be-checked-and-corrected entity in the hierarchical structure of the knowledge graph;
calculating a similarity between the to-be-checked-and-corrected entity and each entity that is at a same level and has a same relationship as the to-be-checked-and-corrected entity in the knowledge graph, to obtain a plurality of entity similarities related to the to-be-checked-and-corrected entity; and
when a maximum value of the plurality of entity similarities is smaller than a predetermined entity similarity threshold, determining that the to-be-checked-and-corrected entity is a to-be-checked-and-corrected entity having an error and that the character information corresponding to the to-be-checked-and-corrected entity has an error.

2. The method for acquiring character information according to claim 1 , wherein the correcting the character information having an error based on the knowledge graph comprises:
determining a number of all entities that are at the same level and have the same relationship as the to-be-checked-and-corrected entity in the knowledge graph, to obtain an entity number;
when the entity number is equal to 1,
directly replacing character information corresponding to the to-be-checked-and-corrected entity having an error with character information corresponding to the entity that is at the same level and has the same relationship as the to-be-checked-and-corrected entity in the knowledge graph, or
calculating a probability that the entity that is at the same level and has the same relationship as the to-be-checked-and-corrected entity is the to-be-checked-and-corrected entity having an error, to obtain an entity probability, and when the entity probability is larger than a predetermined entity probability, replacing character information corresponding to the to-be-checked-and-corrected entity having an error with character information corresponding to the entity that is at the same level and has the same relationship as the to-be-checked-and-corrected entity; and

when the entity number is larger than 1, performing a following method comprising:
determining at least two candidate entities based on the plurality of entity similarities;
calculating a probability that each of the at least two candidate entities is the to-be-checked-and-corrected entity having an error, to obtain a candidate probability for the each of the at least two candidate entities; and
replacing character information corresponding to the to-be-checked-and-corrected entity having an error with character information corresponding to a candidate entity corresponding to a maximum candidate probability.

3. The method for acquiring character information according to claim 2 , wherein the determining at least two candidate entities based on the plurality of entity similarities comprises:
sorting all entities that are at the same level and have the same relationship as the to-be-checked-and-corrected entity in the knowledge graph based on the plurality of entity similarities in a descending order to obtain a sequence, selecting a predetermined number of entities at a beginning of the sequence as the at least two candidate entities; or
sorting all entities that are at the same level and have the same relationship as the to-be-checked-and-corrected entity in the knowledge graph based on the plurality of entity similarities in an ascending order to obtain a sequence, and selecting a predetermined number of entities at an end of the sequence as the at least two candidate entities.

4. The method for acquiring character information according to claim 2 , wherein the calculating a probability that each of the at least two candidate entities is the to-be-checked-and-corrected entity having an error comprises:
determining whether the to-be-checked-and-corrected entity comprises a next-lower-level to-be-checked-and-corrected entity, to obtain a first determination result, wherein the next-lower-level to-be-checked-and-corrected entity is all entities that are subordinate to the to-be-checked-and-corrected entity and at a next lower level of the to-be-checked-and-corrected entity;
determining whether the to-be-checked-and-corrected entity corresponds to a relevant to-be-checked-and-corrected entity, to obtain a second determination result, wherein the relevant to-be-checked-and-corrected entity is all entities, that are at the same level as the to-be-checked-and-corrected entity in the hierarchical structure of the knowledge graph, related to the to-be-checked-and-corrected entity, and have a different relationship with an entity at a next higher level to which the to-be-checked-and-corrected entity is subordinate; and
selecting a method for calculating the probability that the each of the at least two candidate entities is the to-be-checked-and-corrected entity having an error, based on the first determination result and the second determination result;
the first determination result is a first determination sub-result when determining that the to-be-checked-and-corrected entity comprises the next-lower-level to-be-checked-and-corrected entity, or a second determination sub-result when determining that the to-be-checked-and-corrected entity does not comprise the next-lower-level to-be-checked-and-corrected entity; and
the second determination result is a third determination sub-result when determining that the to-be-checked-and-corrected entity corresponds to the relevant to-be-checked-and-corrected entity or a fourth determination sub-result when determining that the to-be-checked-and-corrected entity does not correspond to the relevant to-be-checked-and-corrected entity.

5. The method for acquiring character information according to claim 4 , wherein the selecting a method for calculating the probability that the each of the at least two candidate entities is the to-be-ch

Assignees

• Boe Technology Group Co Ltd

Inventors

• Liu Nan
• Wang Chuan
• Miao Xinyu
• Lei Yiming
• Wang Hong

Classifications

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06V30/26

  Techniques for post-processing, e.g. correcting the recognition result · CPC title

• G16H10/40 Primary

  for data related to laboratory analysis, e.g. patient specimen analysis · CPC title

• G06V30/412 Primary

  Layout analysis of documents structured with printed lines or input boxes, e.g. business forms or tables · CPC title