JP2016122404A - Information processing apparatus, information processing method, program, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically extract information easily and quickly from a paper document converted into an image.SOLUTION: An information processing apparatus comprises: first extraction means that extracts a plurality of areas from document data converted into an image; second extraction means that extracts, from the plurality of areas, an area including a first character or word; and third extraction means that extracts information different from the first character or word from the area extracted by the second extraction means.SELECTED DRAWING: Figure 5

Description

  The disclosed technology relates to an information processing apparatus, an information processing method, a program, and a storage medium.

  IT is rapidly progressing in the medical field, centering on electronic medical records. On the other hand, medical information on various paper media still exists in the hospital. For example, medical information on paper media includes medical information documents, explanation consent documents, documents required for entrance and exit, medical documents such as diagnostic documents, and documents for office procedures such as order slips, reservation slips, and application forms. is there.

  In an environment where medical information (paper documents) on paper media and electronic information such as electronic medical records are mixed, it is desirable that not only electronic information but also paper documents can be quickly searched and utilized.

  Therefore, in order to ensure readability of paper documents, the paper documents are digitized with a scanner, and basic information such as type information indicating the type of paper document, clinical department information, and patient number is manually registered and stored in the electronic system. A linking workflow is generally implemented. However, there are cases where the number of paper documents used in hospitals can reach thousands, and each hospital has its own format. Therefore, it takes a lot of time to register the above basic information from paper documents. It takes time and effort.

  In order to save labor for registration of basic information included in a paper document, in Patent Document 1, a barcode is added to a paper document, and the barcode is read by a barcode reader so that the basic information included in the paper document is obtained. A method of extracting and registering is disclosed.

  Patent Document 2 discloses that a character string to be extracted from a form (name of the issuer of the form) is stored and the form is recognized by collating the character string with a form recognition result. Yes.

Patent No. 5356905 JP 2001-31694 A

  However, in the method of Patent Document 1, it is necessary to add a barcode to a paper document in advance when digitizing a large amount of medical records and questionnaires, which requires a complicated manual operation and a heavy load. Further, in the method of Patent Document 2, since the entire character string to be extracted and the recognition result of the form are collated, if the collation cannot be performed, a new whole character string and a recognition result different from the characters that could not be matched. It takes time to recognize a form because it is necessary to collate the same as in the past.

  The disclosed technology has been made in view of such a situation, and an object thereof is to automatically and quickly extract information from a paper document.

  In addition, the present invention is not limited to the above-described object, and is a function and effect derived from each configuration shown in the embodiment for carrying out the present invention, which is another object of the present invention. It can be positioned as one.

  An information processing apparatus according to the disclosed technique includes a first extraction unit that extracts a plurality of regions from imaged document data, and a second extraction unit that extracts a region including a first character or word from the plurality of regions. And third extraction means for extracting information different from the first character or word from the area extracted by the second extraction means.

  According to the disclosed technique, information can be automatically extracted from an imaged paper document easily and quickly.

It is a figure which shows an example of a structure of the information processing system which concerns on 1st Embodiment. It is a block diagram which shows an example of a function structure of the information processing apparatus which concerns on 1st Embodiment. It is a flowchart which shows an example of operation | movement of the information processing apparatus which concerns on 1st Embodiment. It is a flowchart which shows an example of the procedure of the setting process of the candidate area | region in FIG.3 S120 which concerns on 1st Embodiment. It is a flowchart which shows an example of the procedure of the identification process of the extraction target in FIG.3 S140 which concerns on 1st Embodiment. It is a schematic diagram which shows an example of the setting process of the candidate area | region in step S120 of FIG. 4, and the identification process of the extraction object in FIG.6 S140 based on 1st Embodiment. It is a schematic diagram which shows an example of the knowledge structure which concerns on 1st Embodiment. It is a flowchart which shows an example of the process sequence of the information processing method by the information processing system which concerns on 2nd Embodiment. It is a flowchart which shows an example of the procedure of the correction process of the candidate area | region in step S'230 of FIG. 8 which concerns on 2nd Embodiment. It is a schematic diagram which shows an example of the correction process of the candidate area | region in step S'230 of FIG. 8 which concerns on 2nd Embodiment. It is a flowchart which shows an example of the process sequence of the information processing method by the information processing system which concerns on 3rd Embodiment. 12 is a flowchart illustrating an example of a procedure of a candidate area narrowing process in step S240 of FIG. 11 according to the third embodiment. It is a schematic diagram which shows an example of the narrowing-down process of the candidate area | region in FIG.11 S240 which concerns on 3rd Embodiment. It is a flowchart which shows an example of the process sequence of the information processing method by the information processing system which concerns on 4th Embodiment. It is a flowchart which shows an example of the procedure of the extraction process of the medical department in FIG.3 S140 based on 5th Embodiment. It is a flowchart which shows an example of the process sequence of the information processing method by the information processing system which concerns on 6th Embodiment. It is a schematic diagram which shows an example of the structural characteristic presence / absence of the extraction object regarding FIG. 16 which concerns on 6th Embodiment, and knowledge management of an extraction object. It is a flowchart which shows an example of the process sequence of the information processing method by the information processing system which concerns on 7th Embodiment. It is a schematic diagram which shows an example of the information processing of FIG. 17 which concerns on 7th Embodiment. It is a flowchart which shows an example of the procedure of the confirmation work assistance which identifies the extraction object based on the knowledge in FIG.3 S140 which concerns on 8th Embodiment, and is a transaction regulation object. It is a schematic diagram which shows an example of the information processing of FIG. 19 which concerns on 8th Embodiment.

  Hereinafter, the information processing apparatus according to the present embodiment will be described in detail with reference to the drawings. However, the components described in this embodiment are merely examples, and the scope of the technology of the present invention is determined by the scope of the claims, and is limited by the following individual embodiments. Do not mean.

(First embodiment)
First, the first embodiment will be described.

  FIG. 1 shows an example of the configuration of an information processing system according to the first embodiment.

  As shown in FIG. 1, the information processing system includes a registration unit 1 (information processing device) and a storage unit 2. The registration unit 1 and the storage unit 2 are connected to each other via a wired or wireless network 6 so as to communicate with each other. Further, the registration unit 1 and the storage unit 2 are communicably connected to various systems in the hospital (electronic medical record system 3, ordering system 4, and other systems 5) via a network 6. A plurality of registration units 1 and storage units 2 may be provided.

  The registration unit 1 will be described in detail. The registration unit 1 is an information processing apparatus such as a PC. The registration unit 1 includes a UI device 101, a CPU 102, a RAM 103, a communication IF 104, a UI display unit 105, a program storage area 106, and a data storage area 107.

  The UI device 101 includes at least one of a mouse, a digitizer, a keyboard, and the like, and is used for confirmation, correction, and transmission of registration information by a user.

  The CPU 102 can perform various controls, calculations, and UI display in the apparatus by interpreting and executing the program read from the program storage area 106 into the RAM 103. For example, the CPU 102 functions as a document image analysis unit 110, a candidate region setting unit 120, a candidate region recognition unit 130, an extraction information identification unit 140, and a registration unit 150 as shown in FIG. In addition, the CPU 102 and the RAM 103 included in the registration unit 1 may be one or plural. That is, when at least one processing device (CPU) and at least one storage device (RAM) are connected and at least one processing device executes a program stored in at least one storage device. The registration unit 1 functions as each unit described above.

  The document image analysis unit 110 acquires and analyzes a document image obtained by digitizing a paper document obtained by a scanner (not shown). Here, computerization with a scanner can be referred to as imaging. That is, the document image corresponds to an example of imaged document data. An imaged medical document is referred to as medical document data. The document image analysis unit 110 may directly acquire the digitized document image obtained by the scanner from the scanner, or if the document image obtained by the scanner is stored in the storage unit 2, the document image analysis The unit 110 may acquire a document image from the storage unit 2.

  The document image analysis unit 110 analyzes a digitized document image layout of a paper document, and divides it into a plurality of character areas and photo areas (area division) to extract areas. In other words, the document image analysis unit 110 corresponds to an example of a first extraction unit that extracts a plurality of regions from imaged document data.

  Note that the document image analysis unit 110 acquires the coordinates of each divided area and attribute information indicating whether each area is a character area or a photographic area for each area. The attribute information indicating the character area or the photographic area can be acquired by various known methods. The means for digitizing the paper document is not limited to the scanner, and may be other means.

  The candidate area setting unit 120 sets a candidate area as a target for extracting information from the area divided by the document analysis unit 110. Specifically, the candidate area setting unit 120 sets a character area as a candidate area. In other words, the candidate area setting unit 120 does not set the photo area as a candidate area among the areas divided by the document analysis unit 110. Note that the processing of the candidate area setting unit 120 may be omitted, and the type of the document may be identified by using the dictionary described later by the extraction information identification unit 140 without setting the candidate area. Although the time until the document type is identified by the process of the candidate area setting unit 120 is shortened, the above-described effect can be obtained even if the process of the candidate area setting unit 120 is omitted.

  The candidate area recognition unit 130 acquires character recognition information by performing processing for recognizing the contents of the candidate area set by the candidate area setting unit 120. The character recognition information is a recognition result of the contents of the candidate area.

  The extraction information identification unit 140 identifies the extraction target region from the candidate region based on the recognition result of the candidate region recognition unit 130, and identifies basic information from the description of the identified region. Specifically, the extraction information identification unit 140 identifies the extraction target region from the candidate region using knowledge such as a dictionary created in advance. Then, the extracted information identification unit 140 identifies, for example, the type of the document using knowledge such as a dictionary created in advance from the identified area. Details of knowledge such as a dictionary will be described later. Note that knowledge such as a dictionary may be stored in the RAM 103 or in the data storage area 107. Knowledge such as a dictionary may be stored in a ROM (not shown) included in the registration unit 1.

  The registration unit 150 registers (records) the document image in a predetermined storage unit using the information identified by the extracted information identification unit 140. For example, the registration unit 150 registers the paper document type identified by the extraction information identification unit 140 in the data storage area 107 as the registration information 10 in association with the document image. The registration unit 150 may store the registration information 10 in the storage unit 2.

  In the above example, the CPU 102 functions as each unit shown in FIG. 2, but the present invention is not limited to this, and the FPGA may have at least a part of the above functions. Further, the above functions may be distributed to a plurality of CPUs. Furthermore, the program storage area 106 may be provided inside the registration unit 1 or may be provided outside the registration unit 1. Further, one program storage area 106 may be constituted by a storage device such as a memory, or may be constituted by a plurality of storage devices.

  The communication IF 104 is connected to the network 6 and is a communication interface between the registration unit 1, the storage unit 2, and various servers 3 to 5 in the hospital.

  The UI display unit 105 is an LED, a liquid crystal panel, or the like that displays the state of the apparatus, image information, and registered contents.

  Specifically, the program storage area 106 and the data storage area 107 are a hard disk or a flash memory. However, the present invention is not limited to a specific storage medium. In the registration unit 1, registration information 10 is stored in the data storage area 107. The registration information 10 may be stored on the storage unit 2. In addition, you may store the registration information 10 of the registration part 1 directly linked | related with the system (for example, electronic medical record system 3) in a hospital.

  Assuming that registration information is placed in the storage unit 2, the storage unit 2 will be described in detail. The storage unit 2 is at least one storage medium such as an HDD or an SSD, and the storage unit 2 stores a binder pool 20. The binder pool 20 includes binders 201 and 202. Each binder contains medical documents. That is, the storage unit 2 manages medical documents in units called binders. The binder pool 20 may be stored in association with a hospital system (for example, the electronic medical record system 3). In the binder pool 20, registered materials are stored for each binder according to a predetermined rule so that information can be easily used. As a method of grouping the binder, for example, various types of materials may be stored for each patient, or each material may be stored for each type. For example, the registration unit 150 can store registration information including document images in the binder for each type based on the type of paper document identified by the extraction information identification unit 140.

  With the configuration described above, it is possible to refer to registration information throughout the information processing system.

  The network 6 may be an intranet operated in a hospital or organization, or may be the Internet.

  Since the electronic medical record system / order system is a widely spread and well-known device, description of a hardware configuration example and an operation flow is omitted.

  Next, an example of the processing procedure of the information processing method by the information processing system according to the present embodiment will be described.

  FIG. 3 is a flowchart illustrating an example of a processing procedure of an information processing method performed by the information processing apparatus according to the first embodiment.

  First, in step S110, the document image analysis unit 110 acquires a document image obtained by digitizing a paper document obtained by a scanner (not shown). Then, the document image analysis unit 110 analyzes the digitized document image layout of the paper document and divides it into character areas and photograph areas (area division). As a document image region dividing method, for example, a known method disclosed in Japanese Patent Laid-Open No. 2002-314806 can be used.

  Subsequently, in step S120, the candidate area setting unit 120 sets an area that is a candidate for extraction from the analysis result of the document image. Details of this processing will be described later.

  Subsequently, in step S130, the candidate area recognition unit 130 recognizes the character string in the candidate area and records the recognition information. The recognition information includes the recognition result of the character string and the number of characters, and the number of lines in the case of a paragraph. The recognition process can use a known character recognition technique.

  Subsequently, in step S140, the extraction information identification unit 140 of the information processing apparatus identifies the extraction target region based on the recognition result of the candidate region and the knowledge information, and identifies basic information from the extraction target region. Then, the registration unit 150 of the information processing apparatus registers the document image based on the identification information. Details of this processing will be described later.

  Next, the candidate area setting process in step S120 will be described.

  FIG. 4 is a flowchart illustrating an example of a procedure of candidate area setting processing in step S120 of FIG. 3 according to the first embodiment.

  First, in step S1201, area information acquired by document image analysis by the document image analysis unit 110, that is, coordinate information indicating the position of each area, and attribute information indicating whether each area is a character area or a photographic area, are stored in the document image. The analysis unit 110 inputs the candidate area setting unit 120.

  In step S1202, the candidate area setting unit 120 determines whether the area acquired by the document image analysis unit 110 is a character area based on the attribute information. If it is a character region, the candidate region setting unit 120 sets the character region as a candidate region in step S1203.

  Subsequently, in step S1204, the candidate area setting unit 120 determines whether there is an unprocessed area. If there is still an unprocessed area, the process enters step S1202, and the processes from step S1202 to step S1204 are repeatedly executed. If there is no unprocessed area, the candidate area setting process is terminated.

  Next, the extraction target identification process in step S140 will be described.

  FIG. 5 is a flowchart illustrating an example of the procedure of the extraction target identification process in step S140 of FIG. 3 according to the first embodiment.

  First, in step S1401, the candidate area setting unit 120 and the candidate area recognizing unit 130 input candidate area information to the extraction information identifying unit 140. The candidate area information includes the coordinate information of the candidate area obtained by the candidate area setting unit 120 and the character recognition information obtained by the candidate area recognition unit 130.

  Subsequently, in steps S1402 to S1407, the extraction information identification unit 140 identifies the extraction target region based on the character recognition information and knowledge information of the candidate region, and identifies the contents of the extraction target region. This part will be described in detail.

  First, in step S1402, the extraction information identification unit 140 determines whether there is a ending in the ending dictionary (see reference numeral 604 in FIG. 6) in the candidate area to be processed.

  If the ending is in the candidate area, in step S1403, the extraction information identification unit 140 identifies the candidate area as the extraction area. That is, the extraction information identification unit 140 corresponds to an example of a second extraction unit that extracts a region including the first character from a plurality of regions. The endings included in the ending dictionary correspond to an example of the first character or word. More specifically, the first character is the ending of a word composed of a plurality of characters. Moreover, although the ending part contained in the ending part dictionary is made into 1 character, it is not limited to this, A several character may be sufficient.

  In step S1404, the extraction information identification unit 140 extracts terms in the term dictionary (see reference numeral 605 in FIG. 6) from the extraction region. Here, the term included in the term dictionary corresponds to an example of information different from the first character. That is, the extracted information identification unit 140 corresponds to an example of a third extracting unit that extracts information that is different from the region extracted by the second extracting unit or the first character.

  In step S1405, based on the relationship between the term dictionary and the classification dictionary (see reference numeral 606 in FIG. 6), the type of the document is identified by the extracted term, and the extraction target identification process is terminated. In other words, the extracted information identification unit 140 corresponds to an example of a classifying unit that classifies document data using information extracted by the third extracting unit.

  If the ending part included in the ending dictionary is not in the candidate area, in step S1406, the extraction information identifying unit 140 determines whether there is an unprocessed candidate area. If there is an unprocessed candidate area, the processes from step S1402 to step S1405 are repeated. If there is no unprocessed candidate area, the extraction information identification unit 140 determines that there is no area corresponding to the type in the candidate area and determines that there is no type.

  Next, an example of extraction target identification processing in the present embodiment will be described in more detail while showing the contents of the dictionary.

  FIG. 6 is a schematic diagram illustrating an example of candidate area setting processing in step S120 and extraction target identification processing in step S140 of FIG. 5 according to the first embodiment.

  Reference numeral 601 denotes an example of a result of analysis by the document image analysis unit 110 for a certain document image. The document image is divided into regions surrounded by a frame, and a character region, a photographic region, or other attributes are given to each region.

  Reference numeral 602 denotes an example of a candidate area setting result obtained by the candidate area setting unit 120 from the analysis result of the document image. Each candidate area is sequentially recorded with an area number and coordinate information.

  Reference numeral 603 denotes a result of identification processing of an extraction target from a candidate area.

  In this embodiment, the ending dictionary 604, the term dictionary 605, and the classification dictionary 606 used for the extraction target identification process are stored in a ROM (not shown). The ending dictionary 604 records a common ending included in the type. The term dictionary 605 records terms included in the type. For example, the term dictionary 605 includes the terms “question” and “question”. That is, the term dictionary 605 includes a first reference character and a second reference character that are different from each other, and a ROM (not shown) that holds the term dictionary 605 corresponds to an example of a holding unit. The classification dictionary 606 records the classification related to the type. The dictionary may be stored in storage means other than the ROM (program storage area 106, data storage area 107, storage unit 2, etc.). In this case, the storage unit corresponds to an example of a holding unit.

  Process in the order of candidate areas. Since the character recognition information of the candidate region 01 obtained by the candidate region recognition unit 130 includes the ending “book” in the six-ending dictionary 04, the extraction information identification unit 140 identifies the candidate region as an extraction target region. .

  Further, the extraction information identification unit 140 determines that the term “explanation” in the term dictionary 605 is included in the extraction target area. Specifically, the extraction information identification unit 140 compares the terms included in the term dictionary 605 with the characters included in the extraction target region, and if the comparison results match, the terms included in the term dictionary 605 are extracted. It is judged that it was extracted from. In the present embodiment, the extraction information identification unit 140 compares the term “question” with the characters included in the extraction target area, and if they do not match, compares the next term in the term dictionary 605 with the characters included in the extraction target area. I do. That is, the extraction information identification unit 140, which is an example of the third extraction unit, compares the character included in the region extracted by the second extraction unit with the first reference character, and the comparison result matches. Extracts the character that matches the first reference character as information, and if the comparison result does not match, compares the character included in the region extracted by the second extracting means with the second reference character To do.

  The extracted information identification unit 140 determines from the term dictionary 605 that the term “explanation” is associated with the “type number” “02”. Therefore, the extraction information identification unit 140 determines that the type of “explanation / consent form” corresponding to “02” in the classification dictionary 606 is the document type of the extraction target (paper document). Then, the registration unit 150 records the type of “explanation / agreement” in the data storage area 107 or the storage unit 2 in association with the document image.

  As described above, the present embodiment sets a candidate area to be extracted based on the attribute information of each area in the document image, identifies the extraction target area from the candidate area based on the character recognition information and knowledge information of the candidate area, Type is acquired. However, the present invention is not limited to the above-described embodiment. For example, when extracting clinical department information, patient information (patient identification information such as patient ID) from a medical document (paper document), an extraction target The knowledge information may be replaced depending on the situation. The patient ID is, for example, a number.

  For example, in the case of medical department information extraction, the ending dictionary for type extraction may be a medical department dictionary including “family”. Furthermore, the term dictionary may be changed to a dictionary including words such as “children” and “skin”. The classification dictionary is not essential, but when used, the classification may be changed to “pediatric”, “dermatology”, etc. so that the classification dictionary is also classified in the medical department. In the present embodiment, knowledge is described in terms of a dictionary, but it may be called something other than a dictionary. When extracting patient information (patient ID, etc.), the ending dictionary for type extraction may be a dictionary including “ID”, “number”, and the like. In this case, the characters such as “ID” are often present at the beginning rather than at the end in the area, but in the present embodiment, the wording of the ending dictionary is used for the sake of simplicity. When extracting patient information (patient ID or the like) or the like, a term dictionary or the like is unnecessary because classification is not necessary. Details of a method for extracting clinical department information and patient information (patient ID and the like) will be described in a fifth embodiment to be described later.

  In the present embodiment, the classification of medical documents is classified into the classifications shown in FIG. 6 for easy management of document images. However, the classification is not limited to this, and the classification may be performed more finely. However, it may be classified more roughly. The words and the number of words included in each dictionary are not limited to the contents shown in FIG. 6 and can be arbitrarily changed.

  Further, in the present embodiment, the ending dictionary, the term dictionary, and the classification dictionary for type extraction are taken as an example. However, the names of the dictionaries may be other than the names shown in FIG. Different configurations may be used. For example, as shown in FIG. 7, the term dictionary may have not only associations between terms and terms and types but also associations with endings. In this case, if a ending is found, it is only necessary to check whether a term that can be combined with it is included and extract the term. For example, in step S1402, if the ending of “101” number “book” is found, in step S1404, instead of extracting all the terms contained in the term dictionary from the area, the ending “book” of “101” number is extracted. Only terms that can be combined with are extracted. That is, it is only necessary to extract “question”, “explanation”, and the like (it is not necessary to extract “question”), and the processing speed can be increased. Moreover, you may have collectively 601, 602, and 603 shown in the example of FIG. 6 as a dictionary. That is, the form of the dictionary is not limited to the above example, but may be other forms.

  In the present embodiment, the dictionary is provided inside the registration unit 1 as an example. However, the dictionary may be provided outside the registration unit 1. It may be defined and referred to externally. In the present embodiment, information corresponding to the type is output as “no type” in the document image that cannot be found, but other types of output, for example, the type may be unknown.

  As described above, according to the first embodiment, information can be easily automatically extracted from a paper document. In the above embodiment, since it is not necessary to add additional information such as a barcode to a paper document, it is possible to extract information such as the document type without taking time and effort as compared with the conventional case. Further, since it is not necessary to add additional information such as a barcode to a paper document, it is possible to easily extract information such as a document type from a document in an unknown format. That is, information such as the document type can be automatically extracted even if information is not manually assigned to the medical document and the format of the medical document is not known in advance.

  In the above embodiment, since the extraction area is identified using the ending dictionary, it is not necessary to collate with the term dictionary for all areas, and information such as the document type can be extracted at high speed. . In addition, when a word indicating the type itself, such as an “inquiry sheet”, is extracted from the document image, it may take a long time to extract due to the large number of words indicating the type. However, according to the present embodiment, since the combination of the ending and the term is used, it is possible to shorten the time for extracting the term indicating the type such as “questionnaire”. Here, in the medical field, there are various names of departments and documents for each hospital, so that the present embodiment can be used in the medical field to obtain a remarkable effect.

  In the above example, the document image type is identified in step S1405. However, this step is not essential, and the process may be terminated in step S1404. In this case, the operator can perform classification by referring to the terms extracted in step S1404.

(Second Embodiment)
Next, a second embodiment of the present invention will be described.

  In the first embodiment described above, the character area is set as the extraction candidate area from the analysis result of the document image. In the second embodiment, when a correct chunk area cannot be extracted by the analysis processing of the document image, it is merged with the area.

  Here, the hardware configuration of the information processing system and the functional configuration of the information processing apparatus according to the second embodiment are the same as those in FIGS.

  Next, an example of a processing procedure of the information processing method according to the present embodiment will be described.

  FIG. 8 is a flowchart illustrating an example of a processing procedure of an information processing method by the information processing system according to the second embodiment.

  First, in step S′210, the document image analysis unit 110 acquires a document image obtained by digitizing a paper document obtained by a scanner (not shown). Then, the document image analysis unit 110 analyzes the digitized document image of the paper document and divides it into a character area and a photograph area. This step is the same as step S110.

  Subsequently, in step S′220, the candidate area setting unit 120 sets an area that is a candidate for extraction of a character area from the analysis result of the document image. Specific processing is the same as that in step S120.

  Subsequently, in step S′230, the candidate area setting unit 120 corrects the candidate area. Details of this processing will be described later.

  Subsequently, in step S′240, the candidate area recognition unit 130 recognizes the character string in the corrected candidate area and records the recognition information. This step is the same as step S130.

  Subsequently, in step S′250, the extraction information identification unit 140 identifies the extraction target region based on the recognition result of the corrected candidate region and the knowledge information, and identifies the extraction target contents. This step is the same as step S140.

  Next, the candidate area correction processing in step S′230 will be described.

  FIG. 9 is a flowchart showing an example of the procedure of the candidate area correction process in step S′230 of FIG. 8 according to the second embodiment.

  First, in step S′2301, the candidate area set in step S′220 is input.

  Subsequently, in steps S ′ 2302 to S ′ 2306, an area to be merged is selected from the candidate areas and merged.

  In step S′2302, the candidate area setting unit 120 determines whether the interval between the two candidate areas to be processed is equal to or less than a predetermined threshold T1. That is, the candidate area setting unit 120 compares the interval between two adjacent candidate areas with the threshold T1. Here, the threshold value T1 corresponds to an example of a first threshold value.

  If the interval between the candidate areas is equal to or smaller than the predetermined threshold T1, in step S′2303, whether the difference between the character sizes in the two candidate areas that are further processed by the candidate area setting unit 120 is equal to or smaller than the predetermined threshold T2. Judge whether. Here, the threshold value T2 corresponds to an example of a second threshold value.

  If the difference between the character sizes in the candidate area is equal to or smaller than the predetermined threshold T2, the process proceeds to step S'2304. In step S′2304, if the candidate area setting unit 120 further includes a plurality of characters in one candidate area to be processed, whether or not the difference between the character intervals is equal to or less than a predetermined threshold T3. Judging. That is, when a plurality of characters are not included in one candidate area, step S ′ 2304 may not be executed. Here, the threshold value T3 corresponds to an example of a third threshold value.

  If the difference between the characters in the candidate area is equal to or smaller than a predetermined threshold T3, in step S′2305, the two candidate areas are determined to be merged areas, and in step S′2306, the candidate area setting unit 120 is determined. Merges the two candidate areas and updates the information of the candidate areas. That is, the candidate area setting unit 120 corresponds to an example of an area merging unit that merges the areas extracted by the first extracting unit based on the information about the area extracted by the first extracting unit. Further, in this embodiment, the extraction information identification unit 140 corresponding to an example of the second extraction unit extracts a region including the first character or word from the merged region.

  In step S′2307, the candidate area setting unit 120 determines whether there is an uncompared area. If there is still an uncompared area, the process enters step S ′ 2302 and repeats the processes from step S ′ 2302 to step S ′ 2306, but if there is no uncompared area, the candidate area correction process is terminated. To do.

  Next, an example of a candidate area correction process in step S′230 will be described.

  FIG. 10 is a schematic diagram showing an example of candidate area correction processing in step S′230 in FIG. 8 according to the second embodiment of this invention.

  Reference numeral 10001 denotes an example of a candidate area setting result in a certain document image. Since “same”, “will”, and “book” are separated, they are extracted as independent regions.

  10002 is an example of the result of the candidate area correction processing from 10001 in the case of type extraction. The difference in character string spacing when the distance between the areas from the candidate area 10001 is within a certain range T1, and the difference in character size in each area is within a certain range T2, and there are a plurality of characters in each area. Merge candidate regions within a certain range T3 as divided regions.

  In this embodiment, the candidate area is corrected based on the characteristics of the extraction target to make a meaningful area. In this embodiment, the candidate region merging conditions are the interval between candidate regions, the difference in character size in the candidate region, and the difference in interval between character strings in the candidate region. However, other conditions may be set. Good. In addition, division processing when candidate areas are overintegrated may be taken as an example. In the above embodiment, the candidate region merging conditions are all the intervals between the candidate regions (that is, the positions of the candidate regions), the character size differences in the candidate regions, and the character string interval differences in the candidate regions. However, at least one may be used. That is, the information regarding the region used for merging the regions is information indicating the position of the region extracted by the first extraction unit and at least one of the characters included in the region extracted by the first extraction unit.

  According to the second embodiment, a meaningful area can be extracted, and the accuracy of information extraction processing can be improved.

(Third embodiment)
Next, a third embodiment of the present invention will be described.

  In the second embodiment described above, the area is corrected to a meaningful area based on the analysis result of the document image. In the third embodiment, the candidate area is narrowed down based on the characteristics of the extraction target.

  Here, the hardware configuration of the information processing system according to the third embodiment is the same as the hardware configuration of the information processing system according to the first embodiment shown in FIG. The functional configuration of the information processing system according to the third embodiment is the same as the functional configuration of the information processing system according to the first embodiment shown in FIG.

  Next, a processing procedure of the information processing method according to the present embodiment will be described.

  FIG. 11 is a flowchart showing an example of a processing procedure of an information processing method by the information processing system according to the third embodiment of the present invention.

  First, in step S210, the document image analysis unit 110 analyzes a digitized document image of a paper document and divides it into a character area and a photograph area. Specific processing is the same as that in step S110.

  Subsequently, in step S220, the candidate area setting unit 120 sets an area that is a candidate for extraction of a character area from the analysis result of the document image. Specific processing is the same as that in step S120.

  Subsequently, in step S230, the candidate area recognition unit 130 recognizes the character string in the candidate area and records the recognition information. The specific process is the same as that in step S130.

  Subsequently, in step S240, the extraction information identification unit 140 narrows down the candidate areas based on the characteristics of the extraction target. Details of this processing will be described later.

  Subsequently, in step S250, the extraction information identification unit 140 identifies the extraction target region based on the recognition result of the candidate region and the knowledge information, and identifies the extraction target contents. The specific process is the same as in step S140.

  Next, the candidate area narrowing-down process in step S240 will be described. The candidate area narrowing process is also referred to as candidate area filtering process hereinafter.

  FIG. 12 is a flowchart illustrating an example of a procedure for narrowing down candidate areas in step S240 in FIG. 11 according to the third embodiment of this invention.

  First, in step S2401, the candidate area setting unit 120 inputs the candidate area set in step S220 to the extraction information identification unit 140.

  Subsequently, in steps S2402 to S2404, the extraction information identification unit 140 narrows down the candidate areas. In the case of type extraction, the candidate area is narrowed down using the characteristics that the type area is likely to be within a certain range from the top of the document image and that the type area is unlikely to exist in a multi-paragraph document. Set as. Here, the plurality of paragraphs may be two or more paragraphs or three or more paragraphs. Further, within a certain range is, for example, within the upper third of the entire document image. The fixed range may be within the upper half of the entire document image or may be another range. In the case of medical department extraction or patient information extraction, the narrowing range may be different from the type extraction. That is, the narrowing-down conditions for the candidate area may be changed according to the extraction target. In order to narrow down candidate areas, the above two conditions may be used, or one of the two conditions may be used. Further, other conditions such as the position of the document image in the horizontal direction may be added to the above two conditions.

  In step S2402, the extraction information identification unit 140 determines whether the candidate area to be processed is within a predetermined range. If it is within the predetermined range, in step 2403, the extraction information identification unit 140 further determines whether or not the number of rows in the candidate area is equal to or less than a predetermined threshold T. If it is less than or equal to the predetermined threshold T, in step 2404, the candidate area is left as a candidate area. Here, the threshold value T corresponds to an example of a fourth threshold value.

  In step 2405, candidate areas outside the predetermined range or candidate areas in which the number of character lines in the candidate area is equal to or greater than a predetermined threshold T are excluded from the candidate areas. This is based on the fact that information indicating the type of document image is unlikely to exist in a multi-line document. As described above, the extraction information identification unit 140 corresponds to an example of a region selection unit that selects a region to be processed by the second extraction unit.

  In step S2406, the extraction information identification unit 140 determines whether there is an unprocessed area. If there is still an unprocessed area, the process enters step S2402, and the processes from step S2402 to step S2405 are repeatedly executed. If there is no unprocessed area, the candidate area filtering process is terminated.

  Next, an example of candidate area narrowing processing in step S240 will be described.

  FIG. 13 is a schematic diagram illustrating an example of candidate area narrowing processing in step S240 of FIG. 11 according to the third embodiment of this invention.

  Reference numeral 1001 denotes an example of a candidate area setting result in a certain document image. The area surrounded by the frame is set as a candidate area.

  1002 is an example of the result of filtering candidate areas from 1001 in the case of type extraction. Only a region whose position is within a certain range from the candidate region 1001 and surrounded by a frame that is not a plurality of lines remains. These candidate areas become target areas for identification processing.

  In the present embodiment, the candidate area is narrowed down based on the characteristics of the extraction target, and the extraction target is identified from the remaining candidate areas. In the present embodiment, the condition for filtering candidate areas is set based on the characteristics of the type information, taking the type extraction as an example, but other conditions may be set. When extracting other information, filtering conditions may be set according to the characteristics of the extracted information. In the present embodiment, the position of the candidate area (step S2402) and the number of character lines in the candidate area (step S2403) are used for narrowing down the candidate areas. However, at least one piece of information may be used. According to the third embodiment, in addition to the effects of the first embodiment, the efficiency of information extraction processing can be improved.

(Fourth embodiment)
Next, a fourth embodiment of the present invention will be described.

  In the third embodiment described above, a candidate area is set from the analysis result of the document image, the candidate area is filtered according to the characteristics of the extraction target, and the extraction target is identified from the target candidate area. In the fourth embodiment, in the candidate region to be the target, the order of the extraction targetness is assigned, and the extraction target is identified in the order of the extraction targetness.

  Here, the hardware configuration of the information processing system according to the fourth embodiment is the same as the hardware configuration of the information processing system according to the first embodiment shown in FIG. The functional configuration of the information processing system according to the fourth embodiment is the same as the functional configuration of the information processing system according to the first embodiment shown in FIG.

  Next, a processing procedure of the information processing method according to the present embodiment will be described.

  FIG. 14 is a flowchart illustrating an example of a processing procedure of an information processing method by the information processing system according to the fourth embodiment of the present invention.

  First, in step S310, the document image analysis unit 110 analyzes an electronic document image of a paper document and divides it into a character area and a photograph area. Specific processing is the same as that in step S110.

  Subsequently, in step S320, the candidate area setting unit 120 sets an area that is a candidate for extraction of a character area from the analysis result of the document image. Specific processing is the same as that in step S120.

  Subsequently, in step S330, the candidate area recognition unit 130 recognizes the character string in the candidate area and records the recognition information. The specific process is the same as that in step S130.

  Subsequently, in step S340, the extraction information identification unit 140 narrows down the candidate area based on the characteristics of the extraction target. Specific processing is the same as that in step S240.

  Subsequently, in step S350, the extraction information identification unit 140 calculates the likelihood of extraction in the candidate region to be processed, and assigns the order of the likelihood of extraction. That is, processing priority is assigned to the candidate area. In other words, the extraction information identification unit 140 corresponds to an example of an assigning unit that gives priority to the region extracted by the first extracting unit. Details of the process of assigning the order of the extraction target will be described later.

  Subsequently, in step S360, the extraction information identification unit 140 identifies the extraction target region in the order of the extraction target determined in step S350 based on the recognition result and knowledge information of the candidate region, and identifies the extraction target contents. . The specific process is the same as in step S140.

  Next, a method for calculating the likelihood of extraction of candidate areas in step S350 will be described.

The document type area basically corresponds to an area that seems to be a title of a document image. The title basically has the characteristics that it is located on the document, has a large character size, and is close to the center line. However, since the formats of medical documents are diverse, the type area does not necessarily have the above-described characteristics. Here, using these characteristics, the likelihood of the type of the candidate area is obtained comprehensively by the following formula.
Type-likeness = w1 * {character size} + w2 * {reciprocal of the proximity to the center line} + w3 * {reciprocal of the number of regions at the top}
Here, W1, W2, and W3 are weights of the respective elements. Give high value to important elements. Here, the “upper part” indicates, for example, the range of the upper third of the entire document image, but is not limited thereto. It should be noted that at least one item out of the three terms may be used as the value indicating the type likeness shown in Equation 1. Further, the position of the candidate area is used to obtain the number of areas in the upper part. That is, the extraction information identification unit 140, which is an example of an assigning unit, assigns a priority based on at least one of the position of the region extracted by the first extracting unit and the size of the character included in the region.

  In addition, although the value which shows the kindness shown in Formula 1 is calculated | required by three terms, it is good also as calculating a kindness using four or more items. Further, for example, the extraction information identification unit 140 is set as a processing target from a region having a large value indicating the type-likeness.

  In the present embodiment, the extraction target likelihood (priority) of the candidate area is calculated, and the extraction target is identified in the order of the extraction target likelihood. In the present embodiment, the character size, the position of the region, and the number of regions are used as the elements related to the extraction target by using the type extraction as an example, but other characteristics may be used.

  According to the fourth embodiment, in addition to the effects of the first and third embodiments, it is possible to perform processing from a candidate region having a high possibility of extraction target region, and further improve the efficiency of extraction processing. Will be able to.

(Fifth embodiment)
Next, a fifth embodiment will be described.

  In the first, second, third, and fourth embodiments described above, the example in which the type information is extracted from the medical document has been mainly described. In the fifth embodiment, clinical department information or patient information is extracted from a medical document.

  Here, the hardware configuration of the information processing system according to the fifth embodiment is the same as the hardware configuration of the information processing system according to the first embodiment shown in FIG. The functional configuration of the information processing system according to the fifth embodiment is the same as the functional configuration of the information processing system according to the first embodiment shown in FIG. Further, since the processing procedure of the information processing method according to the fifth embodiment is the same except for step S140 of the information processing method according to the first embodiment shown in FIG. 3, the description of steps S110 to 130 is omitted. To do.

  Since the type extraction processing basically requires classification of the type region after the type region is identified, the type region is identified by the tail information, the type term in the type region is extracted, and the type identification is performed in three steps. Is done. Since the clinical department extraction is basically for extracting the name of the clinical department, it is processed in two steps: identification of the clinical department area and extraction of the clinical department name in the clinical department area. Extraction of patient information is the same as extraction of clinical departments.

  Here, the identification process of the clinical department extraction in step S140 of the present embodiment will be described.

  FIG. 15 is a flowchart showing an example of the procedure of the department extraction process in step S140 of FIG. 3 according to the fifth embodiment of the present invention.

  First, in step S4401, the candidate area setting unit 120 inputs candidate area information to the extraction information identification unit 140.

  Subsequently, in step S4402, the extraction information identification unit 140 determines whether there is a ending in the medical department ending dictionary in the candidate area to be processed.

  If there is a ending, in step S4403, the extraction information identification unit 140 identifies the candidate area as a clinical department area. In step S4404, the extraction information identification unit 140 extracts a term in the clinical department term dictionary in the area as a clinical department name.

  If there is no ending, it is determined in step S4405 whether there is an unprocessed candidate area. If there is an unprocessed candidate area, the processes from step S4402 to step S4404 are repeated. If there is no unprocessed candidate area, it is determined that there is no area corresponding to the medical department among the candidate areas, and there is no medical department information.

  In this embodiment, in addition to the type extraction, clinical department information or patient information is extracted from the document image. The knowledge information may be replaced according to the extraction target.

  According to the fifth embodiment, in addition to the effects of the first, second, and fourth embodiments, information other than the type information can be extracted.

(Sixth embodiment)
Next, a sixth embodiment will be described.

  In the first, second, third, fourth, and fifth embodiments described above, the example in which only one type of information is extracted from the type, the department, and the patient information has been mainly described. In the sixth embodiment, a case where a plurality of information is extracted from a document image will be described.

  Here, the hardware configuration of the information processing system according to the sixth embodiment is the same as the hardware configuration of the information processing system according to the first embodiment shown in FIG. The functional configuration of the information processing system according to the sixth embodiment is the same as the functional configuration of the information processing system according to the first embodiment shown in FIG.

  Next, a processing procedure of an information processing method by the information processing system according to the present embodiment will be described.

  FIG. 16 is a flowchart showing an example of a processing procedure of an information processing method by the information processing system according to the sixth embodiment of the present invention.

  First, in step S510, the document image analysis unit 110 divides an electronic document image of a paper document. Specific processing is the same as that in step S110.

  Subsequently, in step S520, the candidate area setting unit 120 sets a candidate area to be extracted from the result of the area division. Specific processing is the same as that in step S120.

  Subsequently, in step S530, the candidate area recognition unit 130 recognizes the character string in the candidate area and records the recognition information. The specific process is the same as that in step S130.

  Subsequently, in step S540, the extraction information identification unit 140 determines whether the extraction target has structural characteristics by referring to the information illustrated in FIG.

  When it is determined that there is a characteristic, in step S550, the extraction information identification unit 140 narrows down candidate areas based on the characteristic to be extracted. For example, in the case of extracting type information having structural characteristics, the extraction information identification unit 140 narrows the candidate area to the candidate area existing above the document image. The specific process is the same as that in step S340. Here, the “upper part” indicates, for example, the range of the upper third of the entire document image, but is not limited thereto.

  Subsequently, in step S560, the extraction information identification unit 140 switches knowledge information according to the extraction target based on the information illustrated in FIG.

  Subsequently, in step S570, the extraction information identification unit 140 identifies the extraction target based on the recognition result of the candidate area and the knowledge information. The specific process is the same as in step S140. The operator may input information indicating the extraction target to the registration unit 1 so that the registration unit 1 can grasp the extraction target, or the registration unit 1 automatically switches the extraction target in a predetermined order. Thus, the registration unit 1 may grasp the extraction target.

  Next, the presence / absence of structural characteristics of the extraction target and an example of knowledge management of the extraction target will be described.

  FIG. 17 shows a sixth embodiment of the present invention, and is a schematic diagram showing an example of the structural characteristics of the extraction target and the knowledge management of the extraction target related to FIG.

  Reference numeral 1401 denotes a management table of presence / absence of structural characteristics of an extraction target, and records whether the extraction target has structural characteristics. Since the type information is basically at the top of the document image, it is assumed that it has structural characteristics. Since clinical department information and patient information may be described anywhere in the document image, it is assumed that there is no structural characteristic.

  Reference numeral 1402 denotes an extraction target knowledge management table for managing knowledge necessary for extraction of an extraction target. For the type extraction, the ending dictionary 1 for type extraction, the term dictionary 1, and the classification dictionary 1 necessary for classification are used. The ending dictionary 2 and the term dictionary 2 for extracting the clinical department are used for extracting the clinical department. The patient information extraction ending dictionary 3 is used for patient information extraction.

  In the present embodiment, when a plurality of pieces of information are extracted, the knowledge information used for the candidate region setting process and the extraction target identification process based on the structure information is switched according to the extraction target information. In this embodiment, if the extraction target has a characteristic on the structure, the candidate area is narrowed down based on the structural characteristic of the extraction target. Calculation and ranking processing may be performed. In this embodiment, the knowledge of a plurality of pieces of extracted information is managed separately, but the knowledge may be managed collectively.

  According to the sixth embodiment, in addition to the effects of the first, second, third, and fifth embodiments, when extracting a plurality of pieces of information, it is possible to improve the efficiency of information extraction considering information characteristics. become.

  In the first, second, third, fourth, fifth, and sixth embodiments described above, the character area is set as the extraction candidate area from the analysis result of the document image. However, not only the character area but also other attribute areas within a predetermined range may be set widely as candidate areas to be extracted. In the first, second, third, fourth, and sixth embodiments described above, the extraction target region is identified based on the character recognition and knowledge of the candidate region, and the extraction information is identified. The result of character recognition in the candidate area may be corrected, and the extraction target may be identified based on the correction information and knowledge.

(Seventh embodiment)
Next, a seventh embodiment will be described.

  In the first, second, third, fourth, fifth and sixth embodiments described above, information to be extracted is extracted by analyzing a document image. In the seventh embodiment, both medical information and document images stored in an in-hospital system (for example, an electronic medical record system) are analyzed and information is extracted.

  Here, the hardware configuration of the information processing system according to the seventh embodiment is the same as the hardware configuration of the information processing system according to the first embodiment shown in FIG. The functional configuration of the information processing system according to the seventh embodiment is the same as the functional configuration of the information processing system according to the first embodiment shown in FIG.

  Next, a processing procedure of an information processing method by the information processing system according to the present embodiment will be described.

  FIG. 18 is a flowchart illustrating an example of a processing procedure of an information processing method by the information processing system according to the seventh embodiment of the present invention.

  First, in step S610, the extraction information identification unit 140 extracts a patient number from the document image. The patient number extraction process can use the fifth embodiment.

  Subsequently, in step S620, the extraction information identification unit 140 extracts relevant information on the patient from the electronic medical record system. The related information is related to the classification. Details of the related information will be described later.

  Subsequently, in step S630, the extraction information identification unit 140 confirms whether there is related information on the type classification. If there is related information, in step S640, the type classification is narrowed down using the related information. If there is no relevant information, the process enters step S650.

  Subsequently, in step S650, the type is identified from the type classification. Any of the first, second, and fourth embodiments can be used for the type extraction process.

  Next, an example of information processing by the information processing system according to the present embodiment will be described.

  FIG. 19 is a schematic diagram illustrating an example of the information processing of FIG. 18 according to the seventh embodiment of this invention.

  Reference numeral 1601 denotes a description example of the structure information of medical information in the electronic medical record system. Basic information includes patient information, visit date, first visit or revisit. The medical information includes S (main complaint) O (findings) A (examination) P (plan).

  Reference numeral 1602 is an example of related information related to the classification included in the medical record medical information. In the basic information, for example, terms such as first visit or revisit are mentioned. In the medical information, for example, terms such as an operation schedule or hospitalization treatment can be mentioned.

  A classification dictionary 1603 is originally used for the type extraction process.

  In the processing example in which terms related to the classification are extracted from the basic information and the classification classification candidates are narrowed down, first, related information “first visit” is extracted from 1601. In the case of “first visit”, since there is no possibility that the document image has a type such as consent form or record / report, it is excluded from the type candidates. Then, the type is determined and classified from the type numbers “01” and “10” that can be associated with the “first visit”.

  Further, when extracting terms related to the classification from the medical information, the type of the document image is identified from the classification of the range corresponding to the extracted related terms, as described above.

  In the present embodiment, the contents related to the extracted information are extracted from the electronic medical record system, and the extracted information candidates are narrowed down. In the present embodiment, the use of the electronic medical record system is taken as an example, but it may be linked with other related systems. In the present embodiment, information related to type extraction is given as an example, but other related information may be set. Further, in the present embodiment, the type extraction has been described as an example, but clinical department extraction or other information extraction may be performed. Furthermore, in the present embodiment, the type classification candidates are narrowed down by the related information, and the type is identified from the possible type classification. However, as in the first, second, third, fourth, and fifth embodiments, the classification classification is first identified, and the processing method is to check the extraction result with the classification classification narrowed down from the related information. Good.

  According to the seventh embodiment, in addition to the effects of the first, second, third, fourth, and sixth embodiments, it is possible to realize an information extraction mechanism in cooperation with the related system.

(Eighth embodiment)
Next, an eighth embodiment will be described.

  In the first, second, third, fourth, fifth, sixth, and seventh embodiments described above, the type information and the like are automatically extracted for medical atypical documents. . The eighth embodiment relates to information extraction in an atypical document in a general field.

  For example, in the case of banks, document and data capture related to operations such as opening an account, financing efforts, and mortgages are basically done manually. For example, in the case of foreign remittances denominated in US dollars, it is very laborious to check whether the locations of the parties involved in the transaction include prohibited countries or problematic corporations or individuals due to US OFAC regulations. Therefore, it is necessary to support business efficiency.

  Here, in order to improve business efficiency, an eighth embodiment in which necessary information is automatically extracted from documents having various formats and the documents are classified will be described. The hardware configuration of the information processing system according to the eighth embodiment is the same as the hardware configuration of the information processing system according to the first embodiment shown in FIG. The functional configuration of the information processing system according to the eighth embodiment is the same as the functional configuration of the information processing system according to the first embodiment shown in FIG. The hardware configuration of the information processing system according to the third embodiment is the same as the hardware configuration of the information processing system according to the first embodiment shown in FIG. The processing procedure of the information processing method according to the eighth embodiment is the same except for step S140 of the information processing method according to the first embodiment shown in FIG. 3, and thus description of steps S110 to 130 is omitted. To do.

  Next, extraction target identification processing based on knowledge in step S140 will be described.

  FIG. 20 is a flowchart showing an eighth embodiment of the present invention and identifying an extraction target based on knowledge in step S140 of FIG. 3 and showing an example of a procedure for confirming whether or not it is a transaction regulation target. .

  First, in step S7401, the candidate area setting unit 120 inputs candidate area information to the extraction information identification unit 140.

  Subsequently, in step S7402 to step S7406, the extraction information identification unit 140 checks whether or not the basic extraction item content is satisfied, and determines a transaction regulation target. Details will be described below.

  In step S7402, the extraction information identification unit 140 extracts a basic extraction item n. In step S7403, the contents m corresponding to the basic extraction item n is extracted.

  In step S7404, it is checked whether there is any candidate information corresponding to the content m of the basic extraction item n. If there is an item corresponding to the content m of the basic extraction item n, it is determined that the document needs to be further examined, and the process enters step S7407. If there is no item corresponding to the contents m of the basic extraction item n, the process enters step S7405 to check whether all the contents of the basic item n have been checked. If there are still unchecked contents, the process enters step S7403 and the processes from step S7403 to step S7404 are repeatedly executed. When all the contents of the basic extraction items n are checked, it is checked in step S7406 whether all the basic extraction items are checked. If there is still an unchecked basic extraction item, the process enters step S7402, and the processes from step S7402 to step S7406 are repeatedly executed. If none of the basic extraction items correspond to the contents of all the basic items, the process enters step S7412 and determines that the document image is not subject to restriction.

  Steps S7407 to S7413 are a scrutinization process in the case where there are items corresponding to the contents of the basic extraction items in step S7404. Details will be described below.

  In step S7407, an extraction item is extracted. In step S7408, the contents m 'corresponding to the extracted item n' is extracted.

  In step S7409, it is checked whether there is any candidate information corresponding to the content m 'of the extracted item n'. If there is an item corresponding to the content m 'of the extracted item n', the process enters step S7413, and the document is determined to be a restriction target. If there is nothing corresponding to the contents m ′ of the extracted item n ′, the process proceeds to step S74010 to check whether or not all the contents of the extracted item n ′ have been checked. If there are still unchecked contents, the process enters step S7408, and the processes from step S7408 to step S7409 are repeatedly executed. When all the contents of the extraction item n are checked, it is checked in step S7411 whether all the extraction items are checked. If there is still an unchecked extraction item, the process enters step S7407, and the processes from step S7407 to step S7411 are repeated. If none of the extracted items correspond to the contents of all the extracted items, the process proceeds to step S7412, and it is determined that the document image is not regulated.

  Next, an example of information processing by the information processing system according to the present embodiment will be described.

  FIG. 21 is a schematic diagram illustrating an example of the information processing of FIG. 20 according to the eighth embodiment of this invention.

  Reference numeral 1801 denotes an example of a form for overseas remittance business. Items to be checked for transaction restrictions include remittance currency, country name, and trader enclosed in an ellipse.

  Reference numeral 1802 denotes an example of knowledge used for checking a transaction regulation target. The knowledge information includes a content list of items included in 18030 basic extraction items, 18040 extraction items, 18031 basic extraction items, and items included in 18041 and 18042 extraction items. For example, the content number of 18030 basic extraction item 01 “remittance currency” is set to “0101” and the content is set to “USD”. Further, for example, there are a plurality of content lists corresponding to the 18040 extraction item 11 “country name”, which are listed in order. In addition to Japanese, descriptions in other languages are also recorded in association with each other.

  In the information processing, if there is anything corresponding to “remittance currency” or “USD” set as a basic extraction item, it is subject to scrutiny. Further, it is checked whether or not it corresponds to a list of trade prohibited country names set as extraction items or a list of problematic corporations and individuals set as extraction items.

  In the present embodiment, information is automatically extracted from a financial form using knowledge of financial operations. In this embodiment, the overseas remittance business is taken as an example in the financial business, but the present invention may be applied to other business related to document images. In the example of automating overseas remittance work, an example has been given in which items to be checked are managed by dividing them into basic extracted items and extracted items. However, they may be managed collectively or other structures.

  According to the eighth embodiment, in the information extraction in the field set as the target, if necessary architecture is replaced and the proposed architecture is applied, it can be applied to operations other than medical treatment.

  In the first, second, third, fourth, fifth, sixth, seventh and eighth embodiments described above, information is extracted from a scanned document image. You may make it a target. At that time, an image correction process for the camera input image may be added.

(Other embodiments)
Needless to say, the object of the present invention can also be achieved by supplying a system or apparatus with a computer-readable storage medium storing software program codes for realizing the functions of the above-described embodiments. Needless to say, this can also be achieved by the computer (or CPU or MPU) of the system or apparatus reading and executing the program code stored in the storage medium.

  In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention.

  As a storage medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a nonvolatile memory card, a ROM, or the like can be used.

  Further, the functions of the above-described embodiment are realized by executing the program code read by the computer. In addition, an OS (operating system) running on a computer performs part or all of actual processing based on an instruction of a program code, and the above-described embodiment is realized by the processing. Needless to say.

  In addition, it is good also as combining said some embodiment.

110 Document Image Analysis Unit 120 Candidate Area Setting Unit 130 Candidate Area Recognition Unit 140 Extraction Information Identification Unit 150 Registration Unit

Claims (20)

First extraction means for extracting a plurality of regions from the imaged document data;
Second extraction means for extracting a region including a first character or word from the plurality of regions;
Third extraction means for extracting information different from the first character or word from the area extracted by the second extraction means;
An information processing apparatus comprising:   The information processing apparatus according to claim 1, further comprising a classifying unit that classifies the document data using information extracted by the third extracting unit.   The information processing apparatus according to claim 1, wherein the first character is a character at the end of a word composed of a plurality of characters. A holding means for holding a first reference character and a second reference character different from each other;
The third extracting unit compares the character included in the region extracted by the second extracting unit with the first reference character, and if the comparison result matches, the third reference unit Characters that match characters are extracted as the information, and if the comparison results do not match, the characters included in the region extracted by the second extracting means are compared with the second reference characters. The information processing apparatus according to any one of claims 1 to 3. A region merging unit for merging the regions extracted by the first extracting unit based on the information about the region extracted by the first extracting unit;
5. The information processing apparatus according to claim 1, wherein the second extraction unit extracts an area including the first character or word from the merged area. 6.   6. The information on the area is information indicating a position of the area extracted by the first extracting unit and at least one of characters included in the area extracted by the first extracting unit. The information processing apparatus described.   The region merging unit has a plurality of regions extracted by the first extracting unit, wherein a region interval is equal to or smaller than a first threshold value, and a difference in character size included in each region is equal to or smaller than a second threshold value. The information processing apparatus according to claim 5 or 6, wherein the areas are merged.   The region merging unit merges regions of the plurality of regions extracted by the first extracting unit when the difference in character spacing included in each region is equal to or smaller than a third threshold value. 7. The information processing apparatus according to 7. A region selection unit that selects a region to be processed by the second extraction unit based on information about the region extracted by the first extraction unit;
The information processing apparatus according to any one of claims 1 to 8, wherein the second extraction unit extracts an area including the first character or word from the selected area. The information related to the area is information indicating at least one of characters included in the area extracted by the first extracting means and the position of the area extracted by the first extracting means,
The area selecting means selects an area extracted by the first extracting means that is in a predetermined range in the document data and that includes a line number of characters that is equal to or less than a fourth threshold. Item 10. The information processing device according to Item 9. A granting unit that gives priority to the area extracted by the first extracting unit;
11. The information processing apparatus according to claim 1, wherein the second extraction unit extracts an area including the first character or word in an order based on the priority.   12. The information according to claim 11, wherein the assigning unit assigns a priority based on at least one of a position of the region extracted by the first extracting unit and a size of a character included in the region. Processing equipment.   The information processing apparatus according to claim 1, wherein the document data is medical document data.   The information processing apparatus according to claim 13, wherein the third extraction unit extracts at least one of type information, medical department information, and patient identification information of the medical document data as the information.   When extracting the type information, the second extraction unit extracts the region including the first character or word from the region selected by the region selection unit, and extracts the medical department information or patient identification information. In this case, the second extracting unit extracts the region including the first character or word from the plurality of regions extracted by the first extracting unit without using the region selecting unit. The information processing apparatus according to claim 13 or 14 dependent on claim 9.   The claim 13 or the claim dependent on claim 2, wherein the classification means classifies the medical document data using information obtained from an electronic medical record and information extracted by the third extraction means. 14. The information processing apparatus according to 14.   The information processing apparatus according to claim 16, wherein the information obtained from the electronic medical record is information indicating whether or not a first visit is made. A first extraction step of extracting a plurality of regions from the imaged document data;
A second extraction step of extracting a region including a first character or word from the plurality of regions;
A third extraction step of extracting information different from the first character or word from the region extracted in the second extraction step;
An information processing method comprising:   A program causing a computer to execute each step according to claim 18.   A storage medium storing the program according to claim 19.

Images