JP2016200899A - Information processing apparatus, information processing method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate an appropriate attribute value of a document image.SOLUTION: An information processing apparatus includes: recognition means of extracting attribute values of document images by applying recognition processing to each of a plurality of document images belonging to a document image group; result determination means for determining whether the recognition processing has been successful or not, in accordance with a recognition condition; document type specifying means for specifying a document type of the document image; reference image specifying means for specifying a document image determined to have succeeded in recognition processing and having the identified document type, as a reference image, from among the document images belonging to the document image group; and estimating means for estimating an attribute value of the document image determined to have failed in recognition processing, on the basis of an attribute value of the document image specified by the reference image specifying means.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an information processing apparatus, an information processing method, and a program.

  In recent years, computerization has progressed in medical institutions such as hospitals, and an increasing number of institutions have introduced electronic medical records that manage patient diagnosis information. On the electronic medical record, patient diagnosis history, medication information, surgery information, affected area images, simple X-ray images, and the like are centrally managed, and data can be easily shared and reused. In addition, with the recent increase in storage capacity, documents used in the hospital and digital images such as CT, MRI, and simple X-rays can be stored in large amounts in the storage.

  However, even in hospitals where the core system such as electronic medical records is digitized, paper documents still exist in the hospital. In such hospitals, generated paper documents that need to be stored are digitized, and document types and patient information are assigned as attribute values and stored in a document management system or electronic medical record. . In hospitals of medium or larger scale, thousands of paper document scans may occur every day, and it is not uncommon to have personnel to digitize paper documents.

  As an attempt to support such a scan reading operation of a paper document, Patent Document 1 discloses that a document title or the like is extracted from a document image obtained by scanning by character recognition and presented to a user for visual confirmation and correction. A technique for registering is disclosed. Further, Patent Document 2 prints a document type and patient information on a paper document using a one-dimensional barcode or a two-dimensional code, reads the barcode from a document image obtained by scanning, and displays the document type and patient information. A technique for restoring is disclosed.

JP 2000-148788 A JP 2013-131044 A

  However, a method based on character recognition has a problem that sufficient recognition accuracy cannot be obtained depending on a document. In addition, a method of printing a QR code (registered trademark) on a paper document is more advantageous in recognition accuracy than character recognition, but it is necessary to have a device for printing information to be scanned and read on a paper document. For this reason, it is not possible to deal with paper documents issued from hospitals or clinics that do not have a printing device. In addition, there is a possibility that reading the QR code may fail.

  An object is to estimate a correct attribute value of a document image while using a document image recognition process.

  Therefore, the present invention is an information processing apparatus, and by recognition processing for each of a plurality of document images belonging to a document image group, the recognition means for extracting the attribute value of the document image and the recognition processing succeeded according to the recognition conditions. A result determination unit that determines whether or not, a document type identification unit that identifies a document type of the document image, a determination that the recognition process has been successful from the document images belonging to the document image group, and Based on the attribute value of the document image specified by the reference image specifying means for specifying the document image whose reference type is the reference type as the reference image, the recognition processing has failed. And estimating means for estimating the attribute value of the document image determined as.

  According to the present invention, a correct attribute value of a document image can be estimated.

It is a figure which shows a document management system. It is a figure which shows the hardware constitutions of a server apparatus. It is a flowchart which shows a document image registration process. It is a flowchart which shows an attribute value extraction process. It is explanatory drawing of an attribute pattern. It is a flowchart which shows an attribute pattern determination process. It is explanatory drawing of a continuous probability calculation process. It is a flowchart which shows an attribute value estimation process. It is a figure which shows an example of a reference type table. It is a flowchart which shows a 1st reference image detection process. It is a figure which shows an example of a document type table. It is explanatory drawing of an attribute value estimation process. It is a figure which shows an example of a confirmation screen. It is a figure which shows an example of a data structure of a correction history table. It is a flowchart which shows a reference table production | generation process. It is a figure which shows an example of the confirmation screen which concerns on the example of a change. It is a figure which shows an example of the confirmation screen which concerns on the example of a change. It is a figure which shows an example of the confirmation screen which concerns on the example of a change.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a diagram illustrating a document management system 100 according to the first embodiment. The document management system 100 generates a document image from a paper medium, extracts an attribute value indicated in the document image, registers and manages the document image and the attribute value in association with each other. The document management system 100 includes a reading device 110, a server device 120 as an example of an information processing device, and a client device 130, and these devices are communicatively connected via a wired or wireless network 140. ing. In the present embodiment, a case where the document management system 100 manages a document image indicating a medical document will be described as an example.

  The reading device 110 scans a paper document and generates a document image. The reading device 110 is, for example, a scanner device provided with an automatic paper feeding mechanism called an auto document feeder (hereinafter referred to as ADF). As another example, the reading device 110 may be a digital camera that captures a paper document. The reading device 110 includes a document image generation unit 111 and a transmission / reception unit 112. The document image generation unit 111 electronically reads (scans) an input paper document and generates one document image from one paper document. The transmission / reception unit 112 transmits / receives data to / from an external device via the network 140. When a plurality of paper originals are set on the reading device 110 by the user, the paper originals are sent page by page to a scanner unit (not shown). Then, the document image generation unit 111 generates a plurality of document images for each of the plurality of paper originals based on the result of the image reading process of the scanner unit. In the present embodiment, it is assumed that a paper document group composed of a plurality of paper documents is scanned together and a document image group including a plurality of document images is generated correspondingly. The document image group is transmitted to the server apparatus 120 by the transmission / reception unit 112.

  The server device 120 registers and manages the document image generated by the reading device 110. The server apparatus 120 includes a transmission / reception unit 121, an attribute value extraction unit 122, an attribute pattern specification unit 123, an attribute value estimation unit 124, an image management unit 125, a modification management unit 126, and a reference type table generation unit 127. ,have. The transmission / reception unit 121 transmits / receives data to / from an external device via the network 140. The transmission / reception unit 121 receives a document image group from the reading device 110, for example. The attribute value extraction unit 122 extracts an attribute value represented by a character or a barcode from each document image included in the document image group using a recognition technique. The attribute pattern specifying unit 123 specifies the attribute pattern of the document image group. The attribute pattern is information indicating characteristics relating to attribute values indicated by a plurality of document images included in the document image group. The attribute value estimation unit 124 estimates the attribute value of the document image group based on the attribute pattern. The image management unit 125 registers and manages document images. The correction management unit 126 corrects the attribute value obtained by the attribute value extraction unit 122 and records history information related to the correction process. The reference type table generation unit 127 generates a reference type table based on the history information related to the correction process. The reference type table and the process for generating the reference type table will be described later.

  The client device 130 provides a means for the user to check the document image generated by the reading device 110 and the attribute value of the document image. The client device 130 includes a display processing unit 131, an instruction receiving unit 132, and a transmission / reception unit 133. The display processing unit 131 displays a document image or the like on the display screen. The instruction receiving unit 132 receives an instruction input from the user. The transmission / reception unit 133 transmits / receives data to / from an external device via the network 140. The configuration of the document management system 100 is not limited to this embodiment. As another example, the server apparatus 120 and the client apparatus 130 may be provided integrally.

  FIG. 2 is a diagram illustrating a hardware configuration of the server device 120. The server device 120 includes a CPU 201, a ROM 202, a RAM 203, an HDD 204, a display unit 205, an input unit 206, and a network I / F unit 207. The CPU 201 reads a control program stored in the ROM 202 and executes various processes. The RAM 203 is used as a temporary storage area such as a main memory and a work area for the CPU 201. The HDD 204 stores various information such as image data and various programs. Note that the functions and processing of the server device 120 described later are realized by the CPU 201 reading a program stored in the ROM 202 or the HDD 204 and executing this program.

  The display unit 205 displays various information. The input unit 206 has a keyboard and a mouse and accepts various operations by the user. A network I / F unit 207 performs communication processing with an external apparatus such as an image forming apparatus via a network. Note that the hardware configurations of the reading device 110 and the client device 130 are substantially the same as the hardware configuration of the server device 120 described with reference to FIG. However, the reading device 110 further includes a scanner unit that electronically reads a document image.

  FIG. 3 is a flowchart showing document image registration processing by the server apparatus 120. The document image registration process is a process for registering a document image included in a document image group received from the reading device 110. In the present embodiment, the reading device 110 generates a document image corresponding to a medical document using a medical document as an input original, and the server device 120 uses a plurality of document images obtained by a series of scanning operations as a document image group. Shall be sent to In the document image registration processing, the server device 120 uses the character recognition technology to determine the document type from the document image, the patient name as patient identification information (patient ID), and the attribute value of the clinical department name as clinical department identification information (clinical department ID). To extract. Then, the server apparatus 120 registers the document image and the attribute value in association with each other after the user visually confirms the extracted attribute value. The document type, patient name (patient ID), and department name (clinical department ID) are referred to as attribute types, and the value of each attribute type is referred to as an attribute value. Each attribute value may be a character string uniquely numbered by combining alphanumeric characters and the like.

  Hereinafter, the outline of the document image registration process will be described with reference to FIG. 3, and details of each process will be described later. In step S <b> 301, the transmission / reception unit 121 receives a document image group from the reading device 110 via the network 140. The following processing is performed on the document image group received in S301 as a processing target. That is, following S301, in S302, the attribute value extraction unit 122 performs character recognition processing on each of the plurality of document images included in the document image group received in S301, thereby extracting the attribute value of the document image. To do. In the present embodiment, the attribute value extracting unit 122 extracts attribute values of document type, patient ID, and department name.

  In step S303, the attribute pattern specifying unit 123 specifies the attribute pattern of the document image group received in step S301. Next, in S304, the attribute value estimation unit 124 includes a part of the document image group including the document image determined to have failed in the recognition process based on the attribute value obtained in the attribute value extraction process (S302). More accurate attribute values are estimated for the document image than the attribute value extraction result.

  When the process of S304 is completed, the transmission / reception unit 121 transmits the processing target document image, the extracted attribute value, and the estimated attribute value to the client device 130. This process is an example of an output process for outputting a recognition result and an estimation result. In the client device 130, the display processing unit 131 displays the document image and the attribute value on the display unit 205. The user confirms the display contents, inputs a correction instruction for the estimated attribute value as appropriate, and inputs a document image registration instruction. When the instruction receiving unit 132 receives the correction instruction, the transmission / reception unit 133 transmits the correction instruction to the server device 120. When the instruction receiving unit 132 receives a registration instruction, the transmission / reception unit 133 transmits a registration instruction to the server device 120.

  In step S <b> 305, the correction management unit 126 determines whether the transmission / reception unit 121 has received a correction instruction from the client device 130. When the correction management unit 126 receives a correction instruction (Yes in S305), the correction management unit 126 advances the process to S306. If the correction management unit 126 has not received the correction instruction (No in S305), the process proceeds to S307. In S306, the correction management unit 126 corrects the attribute value according to the correction instruction. In step S <b> 307, the image management unit 125 determines whether the transmission / reception unit 121 has received a registration instruction. If the registration instruction is received (Yes in S307), the image management unit 125 advances the process to S308. If the registration instruction is not received (No in S307), the image management unit 125 advances the process to S305.

  In S308, the correction management unit 126 records the correction history. If no correction has been made, information indicating no correction is recorded as a correction history. In step S309, the image management unit 125 registers the document image in the storage unit such as the HDD 204 in association with the attribute value. This completes the document registration process.

  As described above, the document management system 100 according to the present embodiment can present not only the attribute value extracted from the scanned document by the recognition process but also the attribute value estimated from the extracted attribute value to the user. Then, the document management system 100 can register the document image with the correct attribute value after the user visually confirms and corrects the attribute value.

  FIG. 4 is a flowchart showing detailed processing in the attribute value extraction processing (S302) described with reference to FIG. In the attribute value extraction process, the attribute value extraction unit 122 extracts the document type, patient ID, and medical department ID by character recognition, and associates the extraction source document image with the attribute value. The process shown in FIG. 4 is an attribute value extraction process for one attribute value, and the attribute value extraction unit 122 executes the attribute value extraction process for the number of attribute values to be extracted, thereby performing the above three attribute values. And the three attribute values are associated with one document image of the extraction source.

  In step S401, the attribute value extraction unit 122 selects one document image as a processing target from the processing target document image group. In step S <b> 402, the attribute value extraction unit 122 extracts an area that is likely to include the attribute value to be extracted as an attribute area. A known document layout analysis technique or a character recognition technique can be used for the process of extracting the attribute area. For example, for the document type, it is possible to use the knowledge that the document title (document type) is generally at the top of the document and has a larger font size than the text. That is, the attribute value extraction unit 122 extracts a block of character areas in the document image as a character rectangle by document layout analysis, and calculates an attribute area by calculating the likelihood of the document title from the coordinate value and size of the extracted character rectangle. To do. In addition, it is possible to use the knowledge that the patient name and the department name are likely to be described immediately after the labels “patient name:”, “clinical department name:” indicating that they are described. it can. That is, the attribute value extraction unit 122 may extract the label from the document by character recognition, and extract a character rectangular area that appears immediately after as an attribute area.

  Next, in S403, the attribute value extraction unit 122 determines whether or not an attribute region has been extracted. If the attribute area can be extracted (Yes in S403), the attribute value extraction unit 122 advances the process to S406. If the attribute value extraction unit 122 cannot extract the attribute area (No in S403), the process proceeds to S404. In S404, the attribute value extraction unit 122 determines that the recognition process has failed. In step S <b> 405, the attribute value extraction unit 122 sets an attribute value “Unknown” for the attribute type of the document image to be processed. For example, when the patient ID is the processing target, the attribute value extraction unit 122 sets the attribute value “Unknown” for the attribute type of the patient ID of the document image.

  On the other hand, in S <b> 406, the attribute value extraction unit 122 generates character string information by character recognition processing on the character string rectangular image in the attribute area. Here, the character string information is an attribute value, and the process of S406 is an example of a recognition process for extracting an attribute value of a document image by a recognition process for each of a plurality of document images belonging to the document image group. In step S <b> 407, the attribute value extraction unit 122 determines whether the extracted character string information satisfies a feature condition defined for the attribute type to be extracted. Here, the feature condition is that the extracted character string is registered in the dictionary. Here, it is assumed that a character string that appears characteristically is registered in the dictionary and stored in advance in the HDD 204 or the like of the server device 120. The attribute value extraction unit 122 performs partial character string matching between the character string information extracted in S406 and the registered character string registered in the dictionary, and determines whether or not the feature condition is satisfied according to the matching result. To do.

  For example, in the document type example, it is assumed that words frequently appearing in the titles of medical documents such as “inquiry form”, “consent form”, “instructions”, “medical information provision form” are registered in the dictionary in advance. . The attribute value extraction unit 122 determines that the feature condition is satisfied when a word registered in the dictionary exists in the character string information, and determines that the feature condition is not satisfied when the word does not exist. In the example of the department ID, the names of hospital departments such as “neurosurgery”, “orthopedics”, “respiratory medicine”, “cardiovascular medicine”, “ophthalmology”, “dermatology” are listed in the dictionary. Assume that it is registered as a frequently occurring word.

  As another example, the attribute value extraction unit 122 may determine whether or not the character string information satisfies a feature condition such as the number of characters and the type of characters indicated by the attribute type to be extracted. In addition, when a patient ID using a numerical value or alphabet is used as the attribute type, the character type (numerical value, alphabet) of the patient ID of the application destination hospital or the pattern of the number of characters may be registered in advance as a feature condition. .

  Next, in S408, the attribute value extraction unit 122 advances the process to S409 when the character string information satisfies the characteristic condition (Yes in S408). If the character string information does not satisfy the feature condition (No in S408), the attribute value extraction unit 122 advances the process to S404. In step S409, the attribute value extraction unit 122 determines that the recognition process has been successful. In step S410, character string information is set as an attribute value for the attribute type of the document image to be processed. In step S <b> 411, the attribute value extraction unit 122 determines whether the processing in steps S <b> 401 to S <b> 410 has been performed on all document images included in the document image group to be processed. The attribute value extraction unit 122 ends the attribute value extraction process when the processing for all the document images has been completed (Yes in S411). If there is an unprocessed document image (No in S411), the attribute value extraction unit 122 advances the process to S401. As described above, the attribute value extraction unit 122 succeeds in the recognition process according to the recognition condition with the attribute region existing (S403) and the character string information obtained by the recognition satisfying the characteristic condition (S408). Determine whether or not. That is, the processes of S403, S408, S404, and S409 are an example of a result determination process that determines whether the recognition process is successful according to the recognition conditions.

  FIG. 5 is an explanatory diagram of attribute patterns determined in the attribute pattern determination process (S303) described with reference to FIG. As shown in FIG. 5A, documents are often managed for each patient in a ward or the like. For example, in the patient folder 500, medical documents corresponding to patients admitted to a ward are filed and managed for each patient. In such a case, a group of paper documents placed in the patient folder 500 is often digitized collectively by continuous scanning using ADF. As a result, a document image group 501 corresponding to the paper document group is obtained. Reference numeral 502 denotes an attribute value of each document image in the document image group 501. For example, in the document image 510 included in the document image group 501, the attribute value of the document type is “introduction letter” 511, the attribute value of the patient ID is “Taro Yamada” 512, and the attribute value of the department ID is “ "Otolaryngology" 513. As shown in FIG. 5A, in the document image group 501 obtained from the paper document group of the patient folder 500, the same attribute value tends to appear continuously for the attribute type of the patient ID.

  Further, as shown in FIG. 5B, in the hospital reception or the like, there are many cases where documents are managed for each document type. For example, in the questionnaire sheet tray 520, all the questionnaire sheets filled by a plurality of patients at the reception are managed. In such a case, the same attribute value of the document type tends to appear continuously in the document image group 521 corresponding to the paper document group placed in the questionnaire slip tray 520. In addition, as shown in FIG. 5C, in each department in a hospital, a medical document having a specific department name is often managed. For example, a paper document used in the otolaryngology is managed in the medical department tray 530 of the otolaryngology. In such a case, the same attribute value of the medical department ID tends to appear continuously in the document image group 531 corresponding to the paper document group placed in the medical department tray 530.

  As described above, the attribute values of a plurality of document images belonging to the document image group are not random, and show a certain tendency in the arrangement order corresponding to the scan order. The attribute pattern specifying unit 123 according to the present embodiment determines whether the document image to be processed is one of three attribute patterns: a patient ID continuous pattern, a document type continuous pattern, and a medical department ID continuous pattern. Here, the patient ID continuous pattern is a pattern in which the same attribute value of the patient ID continues as described with reference to FIG. As described with reference to FIG. 5B, the document type continuous pattern is a pattern in which the same attribute value of the document type is continuous. As described with reference to FIG. 5C, the clinical department ID continuous pattern is a pattern in which the same attribute value of the medical department ID is continuous.

  FIG. 6 is a flowchart showing detailed processing in the attribute pattern determination processing (S303) described with reference to FIG. Here, a case where the document images dI (t1) to dI (t8) of the document image group 700 of the scan job J1 shown in FIG. Note that a plurality of document images included in the document image group 700 to be processed are arranged in an arrangement order corresponding to the generation order by scanning. For convenience of explanation, FIG. 7A shows only the attribute value of the attribute type “patient ID” among the attribute values of the document image.

  In step S601, the attribute pattern specifying unit 123 sets an attribute type, and specifies, as a processing target, an attribute value corresponding to the set attribute type of the document image in the processing target document image group. Hereinafter, a case where the attribute type of the patient ID is set will be described as an example. In this case, in the example of FIG. 7A, the attribute type “patient ID” (att_P) and its attribute value are specified as the processing target.

属性ａｔｔ＿Ｘに対する属性値が連続する割合Ｐ（ａｔｔ＿Ｘ）を計算する式は、処理対象とする属性値の数をｎとすると、（式１）のように表される。（式１）は、ある属性の系列に対し、隣接する属性値が同じ値をカウントし、最大カウント数ｎ−１（全て同じ値であった場合）で正規化したものである。 Next, in step S602, the attribute pattern specifying unit 123 calculates a probability that the same attribute value of the attribute type set in step S601 continues, that is, a continuous probability, in the arrangement order of the plurality of document images to be processed. Specifically, the attribute pattern specifying unit 123 calculates the continuity probability using (Equation 1). F (X, Y) is calculated by the following equation.

An expression for calculating the ratio P (att_X) where the attribute values for the attribute att_X are continuous is expressed as (Expression 1), where n is the number of attribute values to be processed. (Expression 1) is obtained by counting the same value of adjacent attribute values for a certain attribute series, and normalizing by a maximum count number n−1 (when all are the same value).

  The continuous probability calculation process (S602) will be specifically described with reference to FIG. FIG. 7A is a diagram showing a document image group 700 obtained from the continuously input scan job J1. Each document image in the document image group 700 is managed in the order read in the same scan job. This scan order is the document image arrangement order. The order of arrangement of the document image group 700 is represented by dI (t1), dI (t2),..., DI (tn) in order from the first. Reference numeral 701 denotes an attribute att_P indicating the patient ID of each document image. An attribute value corresponding to the document image dI (tn) is represented by att_P (tn). Reference numeral 702 denotes attribute values of each document image.

For example, in the document image group 700 shown in FIG. 7A, according to (Equation 2), “1” is used if the attribute values of adjacent document images in the arrangement order are the same, and “0” if they are different. Is calculated. In the example of FIG. 7A, the attribute values att_P (t1) and att_P (t2) of the attribute type “patient ID” are both “Taro Yamada”. For this reason, “1” is calculated. On the other hand, since the attribute values att_P (t2) and att_P (t3) are different, “0” is calculated. In the document image group 700, the number n of attribute values is 8, and the number of consecutive attribute values is 5. Therefore, the continuity probability “0.71” is calculated from (Equation 1).
P (att_P) = 5 / (8-1) = 0.71

On the other hand, in the document image group 710 of the scan job J2 shown in FIG. 7B, the number n of attribute values is 8, and the number of consecutive attribute values is 0. Therefore, the continuity probability “0” is calculated from (Equation 1).
P (att_P) = 0 / (8-1) = 0

  Thus, (Expression 1) indicates a higher value as the number of consecutive attribute values increases. If the attribute values of a certain attribute type are all different values in one document image group, the continuous probability is the minimum value “0”, and if all are the same value, the continuous probability is the maximum value “1”. "

  In step S <b> 603, the attribute pattern specifying unit 123 determines whether the continuity probability is greater than or equal to a threshold value. The attribute pattern specific | specification part 123 advances a process to S604, when a continuous probability is more than a threshold value (it is Yes at S603). In step S604, the attribute pattern specifying unit 123 determines that the attribute pattern of the document group to be processed is a continuous pattern corresponding to the attribute type. For example, when the attribute type of the patient ID is a processing target, it is determined as a patient ID continuous pattern. On the other hand, in S603, when the continuous probability is less than the threshold (No in S603), the attribute pattern specifying unit 123 determines that the pattern is not a continuous pattern corresponding to the attribute type to be processed, and ends the attribute pattern determination process. To do. Here, the processing of S601 to S604 is an example of attribute pattern specifying processing for specifying an attribute pattern based on the continuity of attribute values.

  The attribute value extraction unit 122 determines the attribute pattern of the document image group by executing the process shown in FIG. 6 for each of the three attribute types of the document image group to be processed. Thereby, one or more attribute patterns are specified for one document image group.

  FIG. 8 is a flowchart showing detailed processing in the attribute value estimation processing (S304) described with reference to FIG. In step S <b> 801, the attribute value estimation unit 124 sets, as an estimation target, consecutive attribute types in the attribute pattern specified for the document image group to be processed in the attribute pattern specification process (S <b> 303). Hereinafter, a case where the estimation target is the attribute type of the patient ID will be described as an example.

  In step S <b> 802, the attribute value estimation unit 124 identifies a reference type based on the attribute pattern of the document image group to be processed. Here, the reference type is a document type having a high recognition rate of the attribute value of the attribute type to be estimated. FIG. 9 is a diagram illustrating an example of the reference type table. It is assumed that the reference type table 900 is stored in the HDD 104 or the like, for example. The reference type table 900 stores attribute types consecutive in the attribute pattern in association with document types set as reference types. That is, the process of S802 is an example of a reference type specifying process that specifies the reference type of the document image group based on the attribute pattern. In the patient ID continuous pattern, the document types “A”, “B”, and “F” corresponding to the attribute type of the patient ID are specified as reference types from the reference type table 900 of FIG.

  In step S <b> 803, the attribute value estimation unit 124 searches for a first reference image from a plurality of document images in the document image group. Here, the first reference image is a document image that refers to the attribute value of the same type when estimating the attribute value of the attribute type to be estimated. Here, the first reference image search process (S803) will be described in detail with reference to FIG.

  FIG. 10 is a flowchart showing detailed processing in the first reference image detection processing (S803). In step S <b> 1001, the attribute value estimation unit 124 sets a document image with the first arrangement order as a processing target among a plurality of document images in the processing target document image group. Hereinafter, the document image to be processed is referred to as a target image. In step S <b> 1002, the attribute value estimation unit 124 determines whether the target image recognition process has succeeded in the attribute value extraction process (S <b> 302). In the case of success, the attribute value is associated with the document type to be estimated of the target image. In addition, when it has failed, “Unknown” is associated with the document type of the estimation target of the target image. If the recognition process is successful (Yes in S1002), the attribute value estimation unit 124 advances the process to S1003. If the recognition process has failed (No in S1002), the attribute pattern specifying unit 123 advances the process to S1007.

  In S1003, the attribute value estimation unit 124 calculates the feature amount of the target image. For example, the attribute value estimation unit 124 analyzes the target image to extract a text area, a chart area, and a background area, and obtains a layout feature indicating the ratio of each extracted area. The attribute value estimation unit 124 further obtains a text representation indicating the ratio of hiragana, katakana, kanji, alphabet, and numbers in the text area, and a vector representation of the ratio of the RGB component of the color of the chart area. Then, the attribute value estimation unit 124 calculates a feature amount based on these. For the feature amount calculation processing, for example, Japanese Patent Laid-Open No. 2006-318219 can be referred to. It should be noted that the feature amount only needs to indicate the feature of the document image, and the specific type of feature amount and the processing for calculating the feature amount are not limited to the embodiment.

  In step S1004, the attribute value estimation unit 124 refers to the document type table and identifies the document type based on the feature amount. FIG. 11 is a diagram illustrating an example of a document type table. The document type table 1100 is assumed to be stored in the HDD 104, for example. The document type table 1100 stores document types and feature amounts in association with each other. In the document type, a document type used in the hospital is registered in advance. In the document type, for example, all document formats used in the hospital are registered as different document types. As another example, similar formats or types of formats may be grouped into a plurality of document types, and the document types of each group may be registered in the document types. As another example, similar documents may be classified using a known document classification technique or document clustering technique, and each classification item may be registered as a document type in the document type. The feature quantity includes a feature quantity vector of a document image of a corresponding document type as a feature quantity.

  The attribute value estimation unit 124 calculates the similarity between the feature amount of the document type table 1100 and the feature amount calculated in S1003, and specifies the document type with the similarity equal to or greater than the threshold as the document type of the target image. Note that the attribute value estimation unit 124 may specify the document type having the maximum similarity as the document type of the target image. Here, the process of S1004 is an example of a document type specifying process for specifying the document type of the document image.

  In step S1005, the attribute value estimation unit 124 checks whether the document type of the target image is a reference type. If the document type of the target image is the reference type (Yes in S1005), the attribute value estimation unit 124 advances the process to S1006. If the document type of the target image is not the reference type (No in S1005), the attribute pattern specifying unit 123 advances the process to S1007. In step S1006, the attribute value estimation unit 124 sets the target image as the first reference image, and ends the first reference image search process. In step S <b> 1007, the attribute value estimation unit 124 checks whether there is a document image next to the target image in the document image group arrangement order. If the next document image exists (Yes in S1007), the attribute value estimation unit 124 advances the process to S1008. If the next document image does not exist (No in S1007), the attribute value estimation unit 124 advances the process to S1009.

  In step S1008, the attribute value estimation unit 124 sets the next document image as a new processing target, and advances the processing to step S1002. In step S1009, the attribute value estimation unit 124 determines that the first reference image does not exist, and ends the first reference image search process. As described above, in the first reference image search process (S803), among the document images in the document image group, the document image that has been successfully recognized and is the reference type document image and that has the earliest arrangement order is the first. Specified as one document image. In other words, the process of S803 determines that the recognition process is successful from the document images belonging to the document image group, and the specified document type is referred to from the reference-type document images based on the arrangement order. It is an example of the reference image specific process specified as an image.

  Returning to FIG. 8, after the processing of S803, in S804, the attribute value estimation unit 124 determines whether or not the first reference image is detected in S803. If the first reference image is detected (Yes in S804), the attribute value estimation unit 124 advances the process to S805. If the first reference image is not detected (No in S804), the attribute value estimation unit 124 ends the attribute estimation process (S304). In step S <b> 805, the attribute value estimation unit 124 searches for the second reference image by using the document images in the order of arrangement after the document image set as the first reference image as a search target. Here, like the first reference image, the second search image is a document image that refers to the attribute value of the same type when estimating the attribute value of the attribute type to be estimated. The process in S805 is almost the same as the process in S803. However, in S805, the attribute value estimation unit 124 sets a document image in the arrangement order next to the first reference image as a processing target in S1001 illustrated in FIG. Then, the attribute value estimation unit 124 identifies a document image having a successful recognition process and a reference document type as the second reference image.

  Next, in S806, 124 determines whether a second reference image has been detected in S805. If the second reference image is detected (Yes in S806), the attribute value estimation unit 124 advances the process to S807. If the second reference image is not detected (No in S806), the attribute value estimation unit 124 ends the attribute estimation process (S304). In S807, the attribute value estimation unit 124 compares the attribute value of the attribute type of the patient ID of the first reference image and the second reference image. Next, in S808, when the two attribute values match (Yes in S808), the attribute value estimation unit 124 advances the process to S809.

  In step S809, the attribute value estimation unit 124 sets a document image positioned between the first reference image and the second reference image in the document image group arrangement order as a document image to be estimated. Then, the attribute value estimation unit 124 associates the matched attribute value with the estimation target document image as the estimation result. Then, the attribute value estimation unit 124 advances the process to S811. In S810, the attribute value estimation unit 124 sets a document image located between the first reference image and the second reference image in the document image group arrangement order as a document image to be estimated. Then, the attribute value estimation unit 124 associates the estimation target attribute values of the first reference image and the second reference image as estimation results with the estimation target document image. Then, the attribute value estimation unit 124 advances the process to S811. In S811, the attribute value estimation unit 124 sets the second reference image as the first reference image, and the process proceeds to S805. When a plurality of attribute patterns are specified for the document image group to be processed, the server apparatus 120 sets consecutive attribute types in each attribute pattern as estimation targets, and the attributes shown in FIG. It is assumed that value estimation processing is executed. In this embodiment, as described above, the attribute value estimation unit 124 includes a document image whose arrangement order is earlier than the estimation target document image, and a document image whose arrangement order is later than the estimation target document image. Are specified as reference images.

  FIG. 12 is an explanatory diagram of the attribute value estimation process (S304). Reference numeral 1200 denotes a document image group of the scan job J1. Reference numeral 1201 denotes an attribute of each document image, and reference numeral 1202 denotes an attribute value. In the example of FIG. 12, for convenience of explanation, the patient number is shown as the attribute value instead of the patient name as the patient ID. Reference numeral 1210 denotes an attribute value extracted in the attribute value extraction process (S301) for each document image. Of the document image group 1200, the document images dI (t1), dI (t3), dI (t6), and dI (t8) are determined to have been successfully recognized, and the document type is a reference type document image. And On the other hand, the document images dI (t2), dI (t4), dI (t5), and dI (t7) are determined to have been successfully recognized, and the document type is a document other than the reference type document image. To do.

  Based on the above assumptions, detailed processing in the attribute value estimation processing (S304) will be specifically described. In this case, in S803, the attribute value estimation unit 124 succeeds in the recognition processing, and among the document images whose document type is the reference type, the document image dI (t1) that is first positioned in the arrangement order is the first reference image. Set to. Since the first reference image exists (Yes in S804), the attribute value estimation unit 124 advances the process to S805. In step S805, the attribute value estimation unit 124 selects the document image dI (t3) in the order of arrangement after the document image dI (t1) among the document images having the recognition process and the document type being the reference type. 2 reference image. Since the second reference image exists (Yes in S806), the attribute value estimation unit 124 advances the process to S807.

  In step S807, the attribute value estimation unit 124 compares the patient IDs att_P (t1) and att_P (t3) of the document image dI (t1) and the document image dI (t3). Since both the patient IDs are equal to “123456” (Yes in S808), the attribute value estimation unit 124 advances the process to S809. In step S809, the attribute value estimation unit 124 determines the attribute value “123456” of the document image dI (t1) and the document image dI (t3), and the document image dI between the document image dI (t1) and the document image dI (t3). Estimated as the attribute value of (t2). It is assumed that an extraction result has been obtained for the document image dI (t3). As described above, the attribute value estimation unit 124 estimates the correct attribute value even when the extraction result is obtained, and associates the estimation result with the document image.

  Thereafter, in S811, the attribute value estimation unit 124 sets the document image dI (t3) as the first reference image, advances the processing to S805, and sets the document image dI (t6) as the second reference image. Thereafter, the attribute value estimation unit 124 advances the process to S807 and compares both attribute values. Since both attribute values do not match, the process proceeds to S810. In S810, the attribute value estimation unit 124 sets the document image dI (t4) and the document image dI (t5) between the document image dI (t1) and the document image dI (t3) as the estimation target document image. Then, the attribute value estimation unit 124 uses the attribute value “123456” of the document image dI (t3) and the attribute value “654321” of the document image dI (t6) as the estimation result of the estimation target document image. Associate with document image. Thereby, it is determined that the recognition has failed, and a correct attribute value can be estimated and associated with the document image in which “Unknown” is set.

  Further, after the attribute value estimation process (S304), the transmission / reception unit 121 associates the document image with the extraction result 1210 of the attribute value extraction process (S302) and the estimation result 1220 of the attribute value estimation process (S304). To the client device 130. Then, as illustrated in FIG. 12, the client device 130 displays the document image, the extraction result 1210, and the estimation result 1220 in association with each other on the display unit 205. Thus, the user can confirm not only the extraction result but also the estimation result, and can register or modify the attribute value of the document image.

  FIG. 13 is a diagram illustrating an example of a confirmation screen displayed by the display processing unit 131 of the client device 130. The confirmation screen 1300 is displayed after the client device 130 receives the document image, the extraction result, and the estimation result. Reference numeral 1301 denotes an area for displaying document images of a document image group to be processed. The document image is displayed in a size sufficient for the user to confirm the contents. Reference numerals 1302, 1303, and 1304 denote areas indicating attribute values extracted in the attribute value extraction process (S302) from the document image displayed in 1301. An area 1302 displays the extracted document type in an editable state. An area 1303 displays the extracted patient ID in an editable state. An area 1304 displays the extracted department ID in an editable state.

  Reference numerals 1305 and 1306 denote areas for displaying thumbnails of a plurality of document images included in the same document image group. Reference numeral 1305 denotes a thumbnail image indicating a document image in which the extracted attribute value is confirmed by the user and the document image and its attribute value are already registered. Reference numeral 1306 denotes a thumbnail image indicating a document image whose extracted attribute value has not been confirmed, that is, an unregistered document image. In the thumbnail image 1305 of the registered document image, a check mark icon for distinguishing it from the thumbnail image 1306 of the unregistered document image is displayed.

  Reference numeral 1307 denotes an area indicating the estimation result of the attribute value estimation process (S304). When 1307 is selected, the instruction receiving unit 132 receives a correction instruction for the attribute value being displayed in 1307. In FIG. 13, only one estimation result is shown for one attribute value, but when a plurality of estimation results are obtained, a plurality of estimation results may be displayed. Reference numeral 1308 denotes a button for inputting a registration instruction for the document image displayed in 1301. That is, when 1308 is selected, the instruction receiving unit 132 receives a registration instruction for the document image being displayed and the attribute value.

  The confirmation screen 1300 is displayed when the user launches an application for registering scan information on the client device 130. In the initial state, the first document image in the order of arrangement in the group of document images to be processed is displayed in 1301, and the attribute values of the first document image are displayed in 1302, 1303, and 1304. In the example of FIG. 13, “Unknown” indicating that the extraction of the attribute value has failed is displayed in the document type extraction result, the correct patient ID “Taro Yamada” is correct for the patient ID, and correct for the clinical department The medical treatment ID “Otolaryngology” is displayed. Further, in the example of FIG. 13, since there are estimation results for the extraction results of 1302, the estimation results are displayed in a list box 1307 entitled “Correction candidates”.

  When the attribute values displayed in 1302, 1303, and 1304 are different from the attribute values displayed in the document image displayed in 1301, the user edits the attribute values displayed in the respective areas. And rewrite. Further, when the user selects a correction candidate “interview form” in the list box 1307, the content “Unknown” of 1302 can be rewritten to the content “interview form” 1307.

  Assume that the user confirms that the attribute values displayed in 1302, 1303, and 1304 are correct and presses the registration button 1308. In this case, the display state of the document thumbnail corresponding to the document image 1301 indicating that it is not registered displayed in 1306 shifts to the display state 1305 indicating that it is already registered. Further, a document image in the next arrangement order is displayed in 1301, and the user repeats the same confirmation operation until all unregistered document images are confirmed and registered.

  In this way, not only the attribute value extracted from the document image by the recognition technique but also the attribute value as the estimation result is presented to the user, and the user confirms and registers the attribute value, thereby giving the document image an incorrect attribute value. The risk of being lost. In addition, it is possible to reduce the load of the user's editing operation by recommending other attribute values for the attribute values that have failed to be extracted by the recognition technique and causing the user to select them.

  At this time, the transmission / reception unit 133 transmits a registration instruction for the document image displayed in 1301 to the server apparatus 120. Then, when receiving the registration instruction, the server apparatus 120 advances the process to S308 in the document image registration process shown in FIG. Further, in S308, the server device 120 records the presence / absence of correction by the user on the confirmation screen 1300 and the corrected attribute value as a correction history in association with the document image.

  FIG. 14 is a diagram illustrating an example of a data structure of the correction history table. The revision history table 1400 is a relational database table. The correction history table 1400 is stored in a storage unit such as the HDD 204, for example. A column 1401 stores document image names indicating the respective document images. Reference numeral 1402 denotes a column indicating the attribute pattern of the document image group to which the document image belongs. Reference numeral 1403 denotes a column indicating the document type. The document type is determined and stored from the layout characteristics of the document image. A column 1404 stores the patient ID as the attribute value extracted in the attribute value extraction process (S302). Here, a patient number is shown as a patient ID corresponding to FIG. A column 1405 stores whether or not the attribute value extracted in the attribute value extraction process (S302) is corrected, and the corrected attribute value. That is, if there is no correction and the attribute value extracted in the attribute value extraction process (S302) is registered in association with the document image, “no correction” is stored. When correction is performed, the corrected patient ID is stored. 14 shows only the correction history of the patient ID, it is assumed that the server device 120 stores a similar table for the document type and the medical department ID.

  FIG. 15 is a flowchart showing reference type table generation processing by the reference type table generation unit 127. The reference type table generation unit 127 generates the reference type table 900 from the modification history table 1400 in the reference type table generation process. In step S1501, the reference type table generation unit 127 sets one unprocessed attribute pattern. In step S <b> 1502, the reference type table generation unit 127 sets a document image having a set attribute pattern among the registered document images as a processing target. Then, the reference type table generation unit 127 calculates, for each document type, the ratio of the document image in which the correction information “No correction” is recorded with respect to all the document images of the target document type. In the example of FIG. 14, when the document type A is the target, the number of all document images of the target document type (document type A) is 2, and the number of “uncorrected” document images is 2. The ratio of none is 1. On the other hand, when the document type B is the target, the number of all document images of the target document type (document type B) is 2, the number of “uncorrected” document images is 0, and the ratio of no correction is 0. It becomes. In the document type C, the ratio without correction is 0.5.

  In step S <b> 1503, the reference type table generation unit 127 determines a document type for which the ratio of no correction is equal to or greater than a threshold as the reference type. In step S1504, the reference type table generation unit 127 determines whether the processing in steps S1501 to S1503 has been performed on all attribute patterns. If there is an unprocessed attribute pattern (Yes in S1504), the reference type table generation unit 127 advances the process to S1501. When there is no unprocessed attribute pattern (No in S1504), the reference type table generation unit 127 advances the process to S1505. In S1505, the reference type table generation unit 127 generates a reference type table 900 in which attribute types that are consecutive in each attribute pattern are associated with the reference types determined in S1503 for each attribute pattern. This completes the reference table generation process.

  As described above, according to the document management system 100 according to the present embodiment, it is possible to estimate a correct attribute value of a document image while using the document image recognition process. That is, the document management system 100 can estimate a correct value for an attribute value that has failed to be extracted by the recognition process, and can present this value to the user. Also, the document management system 100 can estimate the possibility of an error for the attribute value extracted by the recognition process and present it to the user. Therefore, at the time of document registration, the user only needs to perform an operation of selecting a correct attribute value from the extraction result and the estimation result displayed on the display screen, and the burden of editing work by the user can be reduced.

  As a first modification of the document management system 100 according to the first embodiment, the attribute value extraction unit 122 of the server device 120 recognizes a one-dimensional or two-dimensional barcode instead of character recognition. Thus, the attribute value may be extracted.

  Next, a second modification example will be described. For example, there may be a distributed scanning environment in which a document is scanned on demand when a document such as a reception or a department is generated. In such an environment, the attribute pattern specifying unit 123 of the server device 120 determines the attribute pattern of the image document group based on the device used for the scan, the location where the scan was performed, or the department where the scan was performed. You may specify.

  Specifically, the transmission / reception unit 121 acquires related information of the document image group together with the document image group from the reading device 110. Here, it is assumed that the related information indicates the device used for the scan, the location where the scan was performed, or the department where the scan was performed. In addition, the server device 120 stores an attribute pattern table that associates related information with attribute patterns in the storage unit. And the attribute pattern specific | specification part 123 specifies this attribute pattern from related information with reference to this attribute pattern table. Further, the server device 120 records the related information of the document image group that has been processed in the past and the attribute pattern identification result for the document image group in association with each other as a history, and based on this history information, the attribute information is recorded. A pattern table may be generated.

  As a third modification, the recognition condition for the attribute value extraction unit 122 of the server device 120 to determine the success or failure of the recognition process is not limited to the embodiment. For example, the attribute value extraction unit 122 may determine the success or failure of the recognition result according to the recognition condition that the recognition accuracy is determined to be successful when the recognition accuracy is equal to or greater than a threshold value.

  Further, as a fourth modification example, the client device 130 may display these information after receiving the document image, the extraction result, and the estimation result from the server device 120. It is not limited to the embodiment. FIG. 16 is a diagram illustrating a confirmation screen according to the fourth modification example. In the confirmation screen 1600, reference numeral 1601 denotes an area for displaying document images belonging to the document image group to be processed. In FIG. 16A, two document images 1602 and 1603 are displayed in the area 1601, but the number of document images displayed in the area 1601 is not particularly limited. In this example, the document image displayed in the area 1601 can be switched according to a user operation. In an area 1601, scan dates 1604 and 1605 of the document images 1602 and 1603 are displayed in association with the document images 1602 and 1603, respectively.

  Further, in the document images 1602 and 1603, an area used for character recognition, that is, an attribute area is displayed as a marker. In the character image 1602, an area 1606 where the patient ID is recognized, an area 1607 where the department ID is recognized, and 1608 where the document type is recognized are displayed as markers. Similarly, in the document image 1603, an area 1609 where the patient ID is recognized, an area 1610 where the department ID is recognized, and 16011 where the document type is recognized are displayed as markers. In this example, the attribute area is displayed as a marker. However, the display processing unit 131 only needs to display the attribute area so that it can be distinguished from other areas, and specific processing for that purpose is limited to the embodiment. Is not to be done. As another example, the display processing unit 131 may display an outer frame of the attribute area. The areas 1606 and 1609 in which the patient ID is recognized, the areas 1607 and 1610 in which the medical department ID is recognized, and 1608 and 1611 in which the document type is recognized may be highlighted in different display forms. For example, the areas 1606 and 1609 in which the patient ID is recognized are emphasized as if a red marker is drawn. On the other hand, the areas 1607 and 1610 in which the department ID is recognized are emphasized as if a blue marker is drawn, and the areas 1608 and 1611 in which the document type is recognized are emphasized as if a yellow marker is drawn. In this way, the operator can easily distinguish patient ID, department ID, and document type at a glance.

  When one of the document images displayed in the area 1601 is selected by the user, a check mark indicating that the document image is being selected is displayed on the selected document image. In the example of FIG. 16, a document image 1602 is selected, and a check mark 1612 is displayed on the document image 1602. Reference numeral 1620 denotes a recognition result area for displaying a recognition result for the selected document image. In the recognition result area 1620, patient ID, medical department ID, and document type recognition results 1621, 1622, and 1623 are displayed.

  Reference numeral 1630 denotes a registration information area for displaying registration information for the selected document image. The display contents of the registration information area 1630 are shown in FIG. In the registration information area 1630, a patient ID column 1631, a department ID column 1634, and a document type column 1635 are displayed. In each of the columns 1631 to 1635, the recognition result of each item is displayed as an initial value. The user can check the contents of each column and input correct information if there is an error.

  Further, a search button 1632 corresponding to the patient ID column 1631 is displayed, and the user can search for another patient ID by selecting the search button 1632. Further, patient information 1633 is displayed in association with the patient ID column 1631. Patient information 1633 is patient information associated with the patient ID displayed in the patient ID column 1631. The client device 130 transmits the displayed patient ID to the server device 120 when the patient ID is displayed in the patient ID column 1631 as an initial value or according to a user operation. The server device 120 has a patient DB in which a patient ID and patient information are associated with each other, and sends patient information corresponding to the received patient ID to the client device 130. Then, the client device 130 displays the received patient information as patient information 1633. The patient DB may be provided in a device (not shown) other than the server device 120. In this case, the client device may acquire patient information from a device including the patient DB.

  In the date / time column 1636, a check column 1637 for “specify date / time” and a check column 1638 for “use scan date / time” are displayed. The user can select either. Further, when the user selects the “Specify date and time” check field 1637, the user inputs the date and time in the input field 1638. In the comment field 1640, the user can freely input information. When the registration button 1641 is selected, the display contents of the registration information area 1630 are registered.

  Further, as shown in FIG. 17, when a part of the document image is designated by the enlargement frame 1700, the document image in the enlargement frame 1700 is enlarged and displayed in the upper right area 1701. Thereby, the user can confirm the content of the document image in more detail. In addition, since the document image is enlarged in a state in which the marker indicating the recognition result is displayed, it is possible for the operator to easily grasp the suitability of the recognition result of a fine part. For example, when the marker is not displayed at the last character of the patient ID, the operator can determine that it is a recognition error.

  In the example illustrated in FIG. 17, the area 1701 is provided at the same position as the recognition result area 1620 in FIG. 16, but may be provided at a position different from the recognition result area 1620. For example, the area 1701 may be displayed side by side in the horizontal direction or the vertical direction of the recognition result area 1620, or the area 1701 may be displayed in a pop-up window. By comparing the display contents of the recognition result area 1620 with the enlarged document image displayed in the area 1701, the operator can easily grasp the suitability of the recognition result even when the characters in the document image are small. .

  Further, as a fifth modification example, the document management system 100 may process a document extending over a plurality of pages. FIG. 18A is a diagram showing a confirmation screen for a document group including a document in which one document covers a plurality of pages. Here, it is assumed that the document images 1800 and 1801 have obtained recognition results for the patient ID “123456”, the clinical department “otolaryngology”, and the document type “inquiry sheet”. In addition, it is assumed that the document images 1802 and 1802 have obtained the recognition results of the patient ID “654321”, the clinical department “otolaryngology”, and the document type “consent form”. In this case, the user can input a batch registration instruction for display contents in the registration information area 1630 after selecting document images 1800 and 1801 having the same attributes. In this case, the server apparatus 120 receives a batch registration instruction for the document images 1800 and 1801 via the client apparatus 130. Then, according to the batch registration instruction, the server apparatus 120 registers the display contents of the registration information area 1630 in association with each of the document images 1800 and 1801.

  It is also conceivable that the user erroneously selects document images 1800 and 1802 having different attribute values and inputs a batch registration instruction. In this case, the server device 120 estimates that there is an error in the registered content because the recognition results of the document image 1800 and the document image 1802 are different. Then, the server device 120 transmits to the client device 130 an attribute type that may be erroneous. On the other hand, the client device 130 displays information indicating that there is a possibility of an error. FIG. 18B is a diagram illustrating a display example in the recognition result area 1620 when there is a possibility of an error in the patient ID and the document type. In the recognition results of the document images 1800 and 1802, the patient ID and the document type are different. Therefore, the display processing unit 131 displays information notifying the user of the possibility of an error such as “the patient ID may be incorrect” together with the patient ID related to the batch registration instruction. Further, notification information such as “the document type may be incorrect” is displayed together with the document type related to the batch registration instruction. Thereby, the user can confirm the possibility of an error.

  In the example shown in FIG. 18B, the display of the recognition result of the medical department is omitted, but the recognition result of the medical department may be displayed as in the recognition result area 1620 of FIG. Further, when the recognition result of the medical department is different among a plurality of document images to be registered, a warning such as “the medical department may be wrong” may be displayed. Even when a warning is displayed, the document image may be registered using the registration button 1641, or registration may be prohibited. If document images with different patient IDs are selected, the recognition result area 1620 shown in FIG. 16 may display the recognition result of the document image with a new scan date or the old scan date. The recognition result of the document image may be displayed. When a document image with a different patient ID is selected, a recognition result of a document image with a new scan date may be displayed in the registration information area 1630 in the patient ID column 1631 or the like. The recognition result of the old document image may be displayed.

<Other embodiments>
The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media. Then, the computer (or CPU, MPU, etc.) of the system or apparatus reads and executes the program.

  As mentioned above, according to each embodiment mentioned above, the correct attribute value of a document image can be estimated.

  The preferred embodiments of the present invention have been described in detail above, but the present invention is not limited to such specific embodiments, and various modifications can be made within the scope of the gist of the present invention described in the claims.・ Change is possible.

110 Reading device 120 Server device 130 Client device 122 Attribute value extraction unit 124 Attribute value estimation unit

Claims (10)

Recognition means for extracting attribute values of the document image by recognition processing for each of the plurality of document images belonging to the document image group;
A result judging means for judging whether or not the recognition processing is successful according to the recognition condition;
Document type specifying means for specifying the document type of the document image;
A reference image specifying means for determining, from the document images belonging to the document image group, that the recognition process has been successful, and specifying the document image of which the specified document type is a reference type as a reference image;
And estimating means for estimating the attribute value of the document image determined to have failed in the recognition process based on the attribute value of the document image specified by the reference image specifying means. Information processing device. Attribute pattern specifying means for specifying an attribute pattern based on continuity of attribute values of the document image group;
A reference type specifying unit for specifying the reference type of the document image group based on the attribute pattern;
The information processing apparatus according to claim 1, wherein the reference image specifying unit specifies the document image of the reference type specified by the reference type specifying unit as the reference image.   The reference image specifying means specifies the reference image based on an arrangement order from the document images determined to have succeeded in the recognition process and specified with the reference type. Item 3. The information processing apparatus according to item 1 or 2.   The reference image specifying means includes the document image whose arrangement order is earlier than the document image for which the recognition process has failed, and the document image whose arrangement order is after the document image for which the recognition process has failed. The information processing apparatus according to claim 3, wherein the information processing apparatus is specified as the reference image.   5. The information processing apparatus according to claim 1, wherein the document type specifying unit specifies the document type based on a feature amount of the document image. 6. The document image is an image showing a medical document,
6. The device according to claim 1, wherein the recognizing unit extracts at least one of a document type, patient identification information, and clinical department identification information as the attribute value of the document image. Information processing device.   The attribute pattern specifying unit is configured to determine the attribute pattern based on at least one of identification information of a device used for scanning related to generation of the document image group, a location where the scan is performed, and a department where the scan is performed. The information processing apparatus according to claim 2, wherein the information processing apparatus is specified.   The information processing apparatus according to claim 1, further comprising an output unit that outputs a recognition result by the recognition unit and an estimation result by the estimation unit. An information processing method executed by an information processing apparatus,
A recognition step of extracting attribute values of the document image by a recognition process for each of the plurality of document images belonging to the document image group;
A result determination step for determining whether or not the recognition process is successful according to the recognition condition;
A document type specifying step for specifying a document type of the document image;
A reference image specifying step of determining, from among the document images belonging to the document image group, that the recognition processing has been successful, and specifying the document image whose specified document type is a reference type as a reference image;
An estimation step of estimating the attribute value of the document image determined to have failed in the recognition process based on the attribute value of the document image specified in the reference image specifying step. Information processing method. Computer
Recognition means for extracting attribute values of the document image by recognition processing for each of the plurality of document images belonging to the document image group;
A result judging means for judging whether or not the recognition processing is successful according to the recognition condition;
Document type specifying means for specifying the document type of the document image;
A reference image specifying means for determining, from the document images belonging to the document image group, that the recognition process has been successful, and specifying the document image of which the specified document type is a reference type as a reference image;
A program for functioning as an estimation unit that estimates the attribute value of the document image determined to have failed in the recognition processing based on the attribute value of the document image specified by the reference image specifying unit.

Images