Information processing apparatus, information processing method, and storage medium

1. An information processing apparatus that extracts one or more candidate character strings to be a candidate of an item value corresponding to a predetermined item, from among a plurality of character strings included in a document image to be processed, the information processing apparatus comprising:

one or more hardware processors; and

one or more memories storing one or more programs configured to be executed by the one or more hardware processors, the one or more programs including instructions for:

obtaining information on an attribute group of a preset extraction target among a predetermined plurality of attribute groups;

identifying the attribute group to which each of the plurality of character strings belongs from among the plurality of attribute groups based on a feature vector corresponding to each of the plurality of character strings, and extracting the character string whose identified attribute group is the same as the attribute group of the extraction target from among the plurality of character strings as the candidate character string; and

outputting the extracted candidate character string,

wherein information indicating a nearby group is obtained, the nearby group being the attribute group located near the attribute group of the extraction target,

wherein the character string whose identified attribute group is the same as the nearby group is also extracted from among the plurality of character strings as the candidate character string in addition to the character string whose identified attribute group is same as the attribute group of the extraction target, and

wherein a barycentric position of each of the plurality of attribute groups is calculated and the attribute group that has a barycentric position within a range of a predetermined threshold from a barycentric position of the attribute group of the extraction target is set as the nearby group, from among the plurality of attribute groups.

2. The information processing apparatus according to claim 1 , wherein

the one or more programs further include an instruction for converting data of each of the plurality of character strings to the feature vector, wherein

the attribute group to which each of the plurality of character strings belongs is identified based on the converted feature vector corresponding to each of the plurality of character strings.

3. The information processing apparatus according to claim 1 , wherein the attribute group to which each of the plurality of character strings belongs is identified by using a classifier that classifies an inputted character string to one of the plurality of attribute groups.

4. The information processing apparatus according to claim 3 , wherein the classifier is generated based on a learned model generated by machine learning in which character strings included in a plurality of document images different from the document image to be processed are used as data for learning.

5. The information processing apparatus according to claim 3 , wherein the classifier is generated based on a learned model generated by machine learning in which character strings included in a plurality of document images are converted to the feature vectors, respectively, and the plurality of converted feature vectors are used as data for learning.

6. The information processing apparatus according to claim 1 , wherein

the one or more programs further include an instruction for setting the attribute group of the extraction target, wherein

based on the feature vector corresponding to a character string designated from among one or more second character strings included in a document image different from the document image to be processed, the attribute group to which the designated character string belongs is identified from among the plurality of attribute groups, and the identified attribute group to which the designated character string belongs is set as the attribute group of the extraction target, and

information indicating the set attribute group of the extraction target is obtained.

7. The information processing apparatus according to claim 6 , wherein

the one or more programs further include an instruction for converting data of each of the character strings into the feature vector, wherein

the designated second character string among the one or more second character strings included in the document image used to set the attribute group of the extraction target is converted to the feature vector, and

the attribute group to which the designated character string belongs is identified based on the feature vector corresponding to the designated character string.

8. The information processing apparatus according to claim 6 , wherein the attribute group to which the designated character string belongs is identified by using a classifier that classifies an inputted character string to one of the plurality of attribute groups.

9. The information processing apparatus according to claim 8 , wherein the classifier is generated based on a learned model generated by machine learning in which character strings included in a plurality of document images different from the document image to be processed are used as data for learning.

10. The information processing apparatus according to claim 8 , wherein the classifier is generated based on a learned model generated by machine learning in which character strings included in a plurality of document images are converted to the feature vectors, respectively, and the plurality of converted feature vectors are used as data for learning.

11. The information processing apparatus according to claim 6 , wherein

the one or more programs further include an instruction for comparing an image feature of each of one or more of the extracted candidate character strings and an image feature of the designated character string and calculating a similarity between the character strings, wherein

a priority of the candidate character string whose similarity is high is raised, and

the candidate character string with a high priority among the one or more extracted candidate character strings is preferentially outputted.

12. The information processing apparatus according to claim 11 , wherein the priority of the candidate character string whose similarity is low is reduced.

13. The information processing apparatus according to claim 11 , wherein

information indicating a first relationship between each of the one or more extracted candidate character strings and the character string located in a vicinity of the candidate character string in the document image to be processed among the plurality of character strings is obtained,

information indicating a second relationship between the designated character string and the character string located in a vicinity of the designated character string in the document image different from the document image to be processed among the character strings included in the document image different from the document image to be processed is obtained,

for each of the one or more candidate character strings, the first relationship and the second relationship are compared and a similarity between the character strings is calculated as a first similarity, and

the priority of each of the one or more candidate character strings is changed based on the first similarity.

14. The information processing apparatus according to claim 11 , wherein

information indicating a position of each of the one or more extracted candidate character strings in the document image to be processed is obtained,

information indicating a position of the designated character string in the document image different from the document image to be processed is obtained,

for each of the one or more candidate character strings, the position of the candidate character string and the position of the designated second character string are compared and a similarity between the character strings is calculated as a second similarity, and

the priority of each of the one or more candidate character strings is changed based on the second similarity.

15. The information processing apparatus according to claim 11 , wherein

information on a size of each of the one or more extracted candidate character strings is obtained,

information indicating a size of the designated character string is obtained,

for each of the one or more candidate character strings, the size of the candidate character string and the size of the designated second character string are compared and a similarity between the character strings is calculated as a third similarity, and

the priority of each of the one or more candidate character strings is changed based on the third similarity.

16. A non-transitory computer readable storage medium storing a program for causing a computer to perform a control method of an information processing apparatus that extracts one or more candidate character strings to be a candidate of an item value corresponding to a predetermined item, from among a plurality of character strings included in a document image to be processed, the control method comprising the steps of:

obtaining information on an attribute group of a preset extraction target among a predetermined plurality of attribute groups;

identifying the attribute group to which each of the plurality of character strings belongs from among the plurality of attribute groups based on a feature vector corresponding to each of the plurality of character strings, and extracting the character string whose identified attribute group is the same as the attribute group of the extraction target from among the plurality of character strings as the candidate character string; and

outputting the extracted candidate character string,

wherein information indicating a nearby group is obtained, the nearby group being the attribute group located near the attribute group of the extraction target,

wherein the character string whose identified attribute group is the same as the nearby group is also extracted from among the plurality of character strings as the candidate character string in addition to the character string whose identified attribute group is same as the attribute group of the extraction target, and

wherein a barycentric position of each of the plurality of attribute groups is calculated and the attribute group that has a barycentric position within a range of a predetermined threshold from a barycentric position of the attribute group of the extraction target is set as the nearby group, from among the plurality of attribute groups.

17. An information processing method of extracting one or more candidate character strings to be a candidate of an item value corresponding to a predetermined item, from among a plurality of character strings included in a document image to be processed, the information processing method comprising the steps of:

obtaining information on an attribute group of a preset extraction target among a predetermined plurality of attribute groups;

identifying the attribute group to which each of the plurality of character strings belongs from among the plurality of attribute groups based on a feature vector corresponding to each of the plurality of character strings, and extracting the character string whose identified attribute group is the same as the attribute group of the extraction target from among the plurality of character strings as the candidate character string; and

outputting the extracted candidate character string,

wherein information indicating a nearby group is obtained, the nearby group being the attribute group located near the attribute group of the extraction target,

wherein the character string whose identified attribute group is the same as the nearby group is also extracted from among the plurality of character strings as the candidate character string in addition to the character string whose identified attribute group is same as the attribute group of the extraction target, and

wherein a barycentric position of each of the plurality of attribute groups is calculated and the attribute group that has a barycentric position within a range of a predetermined threshold from a barycentric position of the attribute group of the extraction target is set as the nearby group, from among the plurality of attribute groups.