JP2020016946A - Information processing device, information processing method, program, and document reading system - Google Patents

Abstract

To provide an information processing device and a method capable of identifying an item name of each item of a document and an item value with respect to character data subjected to character recognition on an image of an atypical document.SOLUTION: An information processing device 100 includes: a storage unit which stores item name registration information and a mapping model representing correlation values between a position of an item name of each item of a document and a position of an item value corresponding to the item name; an acquisition unit which acquires image data of an atypical document; a character recognition unit which recognizes each character on the image data; an item name identification unit which identifies the item name of the document on the basis of recognized character information and the item name registration information; a range identification unit which identifies a range where item values having correlation values exceeding a threshold exist from item values corresponding to the identified item names; an item value identification unit which extracts each of characters belonging to the identified range and identifies item values corresponding to the identified item names on the basis of each of the characters; and an output information generation unit which generates output information for outputting information on the identified item names and information on the identified item values in association with each other.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an information processing device, an information processing method, a program, and a form reading system.

2. Description of the Related Art Conventionally, there is known a technology for automating a task using a form, which has been performed manually, by a computer or the like. Examples of such techniques, using OCR (O ptical C haracter R ecognition ), has been conducted to and Image image analysis of a form read by a scanner or the like to recognize the form of the character (character recognition) .

  Further, with respect to the above-described technology, when specifying which item of the form corresponds to the recognized character, the item name and item of each item indicated in the layout information of the form stored in advance. There is known a method of specifying an item name or an item value of an item for character data by arranging values.

  Japanese Patent Application Laid-Open No. H11-163,086 discloses that, with respect to image data of a public ID card, the image data is divided using layout information defined for each type of public ID card, and the image data is divided using optical character recognition. An information input device for converting divided image data into text data is shown. This layout information includes the item names “first and last name”, “address”, and “birth date” of each item of the public identification card. The information input device specifies the item names and item values of the items of the image data of the public identification card based on such layout information.

JP-A-176-173710

  However, the method of specifying item names and item values using the form layout information as described above requires that the layout of the form to be specified needs to be determined in advance. There was a problem that could not be applied to fixed form images.

  Therefore, the present invention provides an information processing apparatus, an information processing method, a program, and the like, which can specify an item name and an item value of an item of a form with respect to a character recognized on a non-standard form image. The purpose is to provide.

  An information processing apparatus according to an aspect of the present invention includes a mapping model representing a correlation value between a position of an item name of each item of a form and a position of an item value corresponding to the item name, and an item name indicating the item name. A storage unit for storing registration information, an acquisition unit for acquiring image data of an irregular form, a character recognition unit for recognizing each character on the image data, and registration of character information and an item name indicating each recognized character. An item name specifying unit that specifies an item name of each item on a form based on the information, and an item whose correlation value is equal to or greater than a threshold value among item values corresponding to the position of the specified item name using a mapping model. A range specifying unit that specifies a range in which a value can be located, and an item value that extracts each character belonging to the specified range and specifies an item value corresponding to an item name specified based on the extracted character Identifying part and information indicating the identified item name And an output information generation unit that generates output information for outputting in association with information indicating a constant item value.

  In an information processing method according to an aspect of the present invention, a computer includes a mapping model representing a correlation value between a position of an item name of each item of a form and a position of an item value corresponding to the item name; Storing the item name registration information shown, acquiring image data of an irregular form, recognizing each character on the image data, character information indicating each recognized character, and item name registration information Identifying the item name of each item on the form based on the above, and using a mapping model, among the item values corresponding to the position of the identified item name, an item value whose correlation value is equal to or greater than a threshold is located. Specifying a range to obtain, extracting each character belonging to the specified range, specifying an item value corresponding to the item name specified based on each extracted character, and specifying the item name Shows And generating an output information for outputting in association with the information indicating the item value specified as information.

  A program according to an aspect of the present invention provides a computer with: a mapping model representing a correlation value between a position of an item name of each item of a form and an item value position corresponding to the item name; and an item indicating an item name. A storage function for storing name registration information, an acquisition function for acquiring image data of an irregular form, a character recognition function for recognizing each character on the image data, character information indicating each recognized character, and an item name An item name specifying function for specifying the item name of each item on the form based on the registration information, and a mapping model, wherein, among the item values corresponding to the position of the specified item name, the correlation value is equal to or greater than the threshold value. A range specifying function for specifying a range in which an item value can be located, and an item for extracting each character belonging to the specified range and specifying an item value corresponding to an item name specified based on each extracted character Value identification function and special It is an output information generating function item name information and in association with information indicating the specified field values indicating generate output information for outputting a, to realize.

  According to the above aspect, the information processing apparatus specifies the position of the item name of each item in the non-standard form, and regarding the item value corresponding to these item names, the position of the item name in each item and the item value It specifies using the correlation with a position. Specifically, the information processing apparatus specifies a range in which an item value having a correlation value equal to or larger than a threshold value can be located in each item name, and specifies item values belonging to the range as corresponding to these item names. I do. Therefore, the information processing apparatus can specify the item name and item value of each item for each character on the form without knowing the arrangement of the item name and item value of each item. Therefore, the information processing apparatus can specify the item name and the item value of each item for each character recognized on the form, even in an irregular form.

  According to the present invention, there is provided an information processing apparatus, an information processing method, a program, and the like that can specify an item name and an item value of each item of a form with respect to a character whose character is recognized on an image of an irregular form. Can be provided.

It is a figure for explaining the example of system composition of the form reading system concerning a 1st embodiment. FIG. 3 is a diagram illustrating an example of a functional configuration of the server device according to the first embodiment. FIG. 4 is a diagram for describing an outline of pairing of the server device according to the first embodiment. FIG. 4 is a diagram for describing an outline of pairing of the server device according to the first embodiment. FIG. 4 is a diagram for describing an outline of pairing of the server device according to the first embodiment. FIG. 4 is a diagram illustrating an operation example of the server device according to the first embodiment. FIG. 2 is a diagram illustrating an example of a hardware configuration of a server device and a terminal according to the first embodiment. It is a figure for explaining the outline of the form reading system concerning a 2nd embodiment. It is a figure showing an example of functional composition of a server device concerning a 2nd embodiment. It is a table | surface which shows an example of the pairing execution condition of the server apparatus which concerns on 2nd Embodiment. It is a table | surface which shows an example of the year / month insertion condition of the server apparatus which concerns on 2nd Embodiment.

  A preferred embodiment of the present invention will be described with reference to the accompanying drawings. In each of the drawings, the components denoted by the same reference numerals have the same or similar configurations.

[First Embodiment]
First, a first embodiment of the present invention will be described. In the present embodiment, a non-standard paper medium form is read as an image, and the item name and item value of each item of the form are specified from the image and converted into text data (hereinafter, also referred to as “text data”). An example in which the output is performed will be described. The term "atypical" as used herein refers to a state in which the format of the target form is not fixed, and the non-standard form has a plurality of issuing organizations and a plurality of patterns, so that the layout is fixed. Also includes forms that do not have to be.

<1. System Configuration>
An example of a system configuration of a form reading system 1 according to the embodiment will be described with reference to FIG.

  The form reading system 1 is a system for reading a form on a paper medium and converting it into text data. As shown in FIG. 1, the form reading system 1 includes a server device 100, a terminal 200 used by a user, and a scanner 300. The server device 100, the terminal 200, and the scanner 300 are connected to each other via a network N.

  The network N is configured by a wireless network or a wired network. Examples of the network include a mobile phone network, a PHS (Personal Handy-phone System) network, a wireless LAN (Local Area Network), 3G (3rd Generation), LTE (Long Term Evolution), and 4G (4th Generation (M)). (Registered trademark), infrared communication, Bluetooth (registered trademark), a wired LAN, a telephone line, a power line network, a network conforming to IEEE 1394, and the like.

  The server device 100 is an information processing device capable of performing data analysis / analysis processing such as communication with the terminal 200 and the scanner 300 and image analysis. By executing a predetermined program, the server device 100 performs image analysis on image data of a form read and output by the scanner 300 in cooperation with the scanner 300, and performs an image analysis on the form data. A function of analyzing an item value corresponding to the item name is realized.

  The terminal 200 is a terminal device such as a smartphone, a laptop terminal, or a desktop terminal capable of communicating with the server device 100 and outputting an item name and an item value of each item of the form to the user as text data. By executing a predetermined program, the terminal 200 realizes a function of transmitting and receiving information on a form and outputting a screen on a form in cooperation with the server apparatus 100.

  The scanner 300 is an image input device that converts a document such as a character, a photograph, or a picture into digital image data. Note that the scanner 300 may be any device as long as it can read a paper medium form and convert it into image data, and may be a terminal device such as a smartphone or a laptop terminal having a camera function.

<2. Functional Configuration>
The functional configuration of the server device 100 according to the present embodiment will be described with reference to FIG. As shown in FIG. 2, the server device 100 includes a control unit 110, a storage unit 120, and a communication unit 130. The server device 100 specifies an item name and an item value of each item of the form from the acquired image of the form. The server device 100 causes the terminal 200 to output the specified result. Specifying the item name of each item and the item value corresponding to the item name in this way by the server apparatus 100 is also referred to as “pairing the item name and the item value” or simply “pairing”.

  The control unit 110 includes an acquisition unit 111, a character recognition unit 113, an item name specification unit 114, a range specification unit 115, an item value specification unit 116, and an output information generation unit 117. Further, the control unit 110 may include, for example, a pattern classification unit 112 or a construction unit 118.

  The acquisition unit 111 acquires image data of a form. More specifically, the acquisition unit 111 may acquire a paper form read by the scanner 300 via the communication unit 130 and converted into image data, or a form stored in the storage unit 120 in advance. May be obtained.

The pattern classification unit 112 classifies the form pattern indicated in the image data based on the form image data acquired by the acquisition unit 111. Pattern classification unit 112, for example, a convolutional neural network (C onvolutional N eural N etwork: CNN) using extracts the feature quantity of the image data may be performed from the image classification. Specifically, the pattern classification unit 112 extracts a feature amount by subdividing the image data and repeatedly executing a convolution operation and sampling. In addition, the pattern classification unit 112 may classify each format pattern based on pattern information indicating a format pattern of a form stored in the storage unit 120 in advance. Further, the pattern classifying unit 112 classifies each of the format patterns so that the form indicated by the acquired image data is a type of form such as a receipt, a medical certificate, a resident's card, a health insurance card or a license. You may classify to what kind. The processing related to pairing of the item name and the item value in the item name specifying unit 114, the range specifying unit 115, and the item value specifying unit 116 may be performed for each pattern of the form classified by the pattern classification unit 112.

  The character recognition unit 113 recognizes each character on the image data of the form acquired by the acquisition unit 111. Further, the character recognition unit 113 may recognize each character on the image data of the form for each pattern classified by the pattern classification unit 112, for example. Specifically, the character recognizing unit 113 may extract a pixel value representing each character on the image data of the form, and recognize each character on the form based on the extracted pixel value. More specifically, the character recognizing unit 113 recognizes each character on the form by extracting a pixel value serving as a feature amount from the image data of the form and performing pattern recognition on the extracted pixel value. You may.

  The character recognizing unit 113 further performs (1) a layout analysis process of a form for identifying a character area, an image area, and a ruled line on the form as pre-processing for extracting the pixel value, and (2) a line of the identified character area. (3) The above-described character extraction processing for each extracted line may be performed.

  The item name specifying unit 114 specifies the item name of each item of the form based on the character information indicating each character recognized by the character recognizing unit 113 and the item name registration information stored in the storage unit 120. Here, the “item name registration information” is information indicating the item name of each item of the form registered in advance. The item name registration information includes, for example, when the form is a receipt issued by a medical institution (hereinafter also referred to as “medical cost receipt”), “issue date” of an item indicating a date on which the medical expense receipt was issued, A list of item names to identify each item, such as "Initial re-examination fee", which indicates the cost required when consulting a doctor, or "Surgery", which indicates the cost of surgery This is information that has been uploaded.

  The item name specifying unit 114 searches character information using, for example, a part or all of the item names registered in the item name registration information as a search keyword, and matches each character or a character obtained by combining each character that matches the search keyword. A column may be specified as an item name. In this search, the item name specifying unit 114 performs a search by absorbing various noises such as a swaying of a spelling, a difference in a synonym, an expression by a different name, a deviation from a main theme and a search word, or a typographical error. May be included.

  The range specifying unit 115 uses the mapping model stored in the storage unit 120 to set the position of the item name and the item value corresponding to the position of the item name specified by the item name specifying unit 114. A range in which an item value whose correlation value with the position (hereinafter, also referred to as “position correlation value”) is equal to or larger than a threshold is specified. Furthermore, the range specifying unit 115 may specify a range in which an item value corresponding to the item name can be located, excluding a range in which the specified item name is located. The range specifying unit 115 may specify the range in which the item name is located, for example, based on the coordinates of a character frame of a character string forming the item name. According to such a configuration, the server device 100 can narrow the range in which the item value can be located, and thus can omit redundant character extraction processing and item value identification processing.

  The “mapping model” is a model representing a correlation value (position correlation value) between the position of the item name of each item of the form and the position of the item value corresponding to the item name. The mapping model is, for example, a model in which the correlation value between the position of the item name of each item of the form and the position of the item value is mapped on the form for each form pattern classified by the pattern classification unit 112. The correlation value is calculated by the later-described constructing unit 118 by statistically analyzing a plurality of pieces of relative position information between the item name and the item value of each item, which is learning data. In addition, the mapping model may be obtained by smoothing the data of the relative position in the statistical analysis. Further, the mapping model may be a heat map which is a visualization graph expressing data of a relative position between an item name and an item value of each item, and a correlation level of a relative position of a plurality of item values corresponding to the item name. May be represented in two dimensions. Details of the mapping model will be described later in the construction unit 118.

  The range specifying unit 115 further uses the mapping model corresponding to the form pattern classified by the pattern classification unit 112 to determine the position of the item value equal to or more than the predetermined position correlation value corresponding to the position of the specified item name. The indicated range may be specified. According to such a configuration, the server device 100 can specify the item value based on the relative position between the learned standard item name and the item value for each form pattern.

  The item value specifying unit 116 extracts each character belonging to the range specified by the range specifying unit 115. By extracting each character in this way, the item value specifying unit 116 narrows down characters to be paired with an item name described later. Further, the item value specifying unit 116 may extract each of the above-mentioned characters by limiting the character type to the item value corresponding to the item name specified by the item name specifying unit 114. Specifically, when the specified item name is “surgery”, the item value specifying unit 116 further specifies a character type from 0 to 9 that is a character type that can take an item value from each character belonging to the specified range. Only characters of numbers, "dots" and "circles" are extracted.

  The item value specifying unit 116 specifies an item value corresponding to the item name specified by the item name specifying unit 114 based on the extracted characters. The item value specifying unit 116 may specify the item value by, for example, (1) constructing a character string of the item value from the extracted characters, (2) a character string and an item name of the constructed item value It is conceivable to use a method of performing pairing of the item name and the item value based on the position correlation value (pair-likeness) of the item value with respect to.

  The method of constructing the above (1) in the item value specifying unit 116 includes, for example, (a) combining the extracted characters in units of lines, and (a) combining the characters combined in units of lines with a predetermined rule. (C) recombining the characters in the divided group, (d) if the combined character is a number, combine the number with the unit of the number, etc. It may be constructed using a method.

As a more specific construction method, when combining the extracted characters in (a) above on a line-by-line basis, the item value specifying unit 116 may use, for example, the coordinate bbox (x _min , y _Min , x _max , y _max ) are grouped on a line-by-line basis by determining whether the center of the y-coordinate of the character frame coordinates of the other character is included in the y-direction range of _min .

  In dividing the characters combined in line units (a), the item value specifying unit 116 divides the characters based on a predetermined threshold. As an example of a method of dividing a combined character based on the predetermined threshold, (a) loosening the threshold between “flat”, “composed”, “year”, and “month” and a numerical value, (b) "Flat", "H" and "~" are determined to be the beginning of the character string. (C) "Day", "dot", "yen" and "~" are determined to be the end of the character string. (D) In the combined characters, if there is a ruled line between the coordinate bbox of the first character and the coordinate bbox of the last character, the group may be divided. In addition, the item value specifying unit 116 may perform the word division using natural language processing to divide the combined characters. As an example of a method of performing word division using this natural language processing, a word n-gram model, which is one of probabilistic language models, is obtained by dividing an object in units of words and modeling the set in units of n consecutive words. A method of dividing by finding the n-gram probability using is considered.

  In recombining the characters in the divided groups in (c) above, when there are characters such as “hand” and “jutsu” in the same group, for example, the item value specifying unit 116 Combine to construct a string such as "surgery".

  When combining the numerical value of (d) and the unit of the numerical value, the item value specifying unit 116 determines, for example, that the form is a medical expense receipt and the character string constructed in (c) is each medical fee score. In the case of the numerical value “48” representing “”, a combination of the numerical value and “point” representing the unit of the numerical value is performed to construct a character string of the item value of “48 points”.

  Here, referring to FIGS. 3 to 5, a method in which server apparatus 100 specifies a range in which an item value corresponding to an item name can be located using a mapping model, and specifies each character belonging to the specified range. An example of an extraction method and a method of specifying an item value based on each extracted character will be described. In this example, the form to be processed is a medical expense receipt, the specific target item is an item indicating the cost of the operation, and the item name is “surgery”.

  FIG. 3 is a schematic diagram illustrating an example of an image of a medical expense receipt in a state where the position correlation value according to the mapping model is not mapped. For simplicity of description, as shown in FIG. 3, the character strings of the item value candidates corresponding to the item name T1 “surgery” are represented by a character string C1 “55 (dot)” and a character string C2 “anesthesia”. And the character string C3 “32 (dots)”.

  FIG. 4 is a schematic diagram showing an example of a state in which a position correlation value is mapped on the image of the medical expense receipt shown in FIG. FIG. 4 is also an enlarged view of the area F shown in FIG. As shown in FIG. 4, the mapping area A1 indicated by the mapping model is an area in which the relative positions of a plurality of item values corresponding to the item name T1 are mapped based on the width and height of the item value. Further, the position correlation value at this relative position is set uniformly within the mapping area A1. For example, if the position correlation value is “0.1” when there is a correlation and the position correlation value is “0” when there is no correlation, the position correlation value mapped in the mapping area A1 is uniformly “0.1”.

  As a specific method, first, the range specifying unit 115 uses the mapping model based on the item name “surgery” to specify the mapping area A1 as a range in which an item value whose position correlation value is larger than the threshold “0” can be located. I do. Next, when the range specifying unit 115 specifies the mapping area A1, the item value specifying unit 116 specifies the characters C11 “5” and C12 “5” to which the center position belongs in the area of the mapping area A1, and the character C21 “ "Hemp" and the character C22 "Drunken" are extracted. Next, the item value specifying unit 116 constructs a character string C1 “55” from the characters C11 “5” and C12 “5” and a character string C2 “anaesthesia” from the characters C21 “hemp” and the character C22 “drunk”. I do.

  FIG. 5 is a schematic diagram showing an example of a state where the correlation value is mapped on the image of the medical expense receipt shown in FIG. FIG. 5 is also an enlarged view of the area F shown in FIG. Regarding the mapping areas A2 and A3 indicated by the mapping model of this example, the difference from the example of A1 in FIG. 4 is that (1) the relative position of the item value corresponding to the item name T1 “surgery” is the magnitude of each position correlation value. (2) The point where the width and height of the item value are not taken into account. For example, the mapping areas A2 and A3 are also heat maps in which light and shade of colors are classified according to the magnitude of the position correlation value. The mapping areas A2 and A3 indicate that the position correlation value decreases as the color approaches dark gray, and increases as the color approaches white.

  As a specific method, first, the range specifying unit 115 uses a mapping model based on the item name T1 “surgery” to set mapping areas A2 and A3 as ranges in which an item value having a correlation value larger than the threshold “0” can be located. To identify. Next, in the character C1 “55” and the character C2 “anesthesia” to which the center position belongs in the areas of the mapping areas A2 and A3, the item value specifying unit 116 determines the position of the center position of each character on the mapping areas A2 and A3. Get the position correlation value. In this example, the center position of the character string C1 “55 (point)” is a position correlation value “0.9”, and the center position of the character string C2 “anesthesia” is a position correlation value “0.5”. When the threshold value of the position correlation value for specifying as the item value is set to “0.85”, the item value specifying unit 116 acquires the character string C1 “55 (point)” of the position correlation value of 0.85 or more. , The character string C2 “anesthesia” of the position correlation value smaller than 0.85 is discarded. Therefore, the item value specifying unit 116 specifies the character string C1 “55 (dot)” as the item value corresponding to the item name T1 “surgery”.

  The item value specifying unit 116 further includes: (1) a correlation value between a character type or a word forming an item name and a character type or a word forming an item value; and (2) an inter-character value or a character string based on ruled line information. May be added to the position correlation value to calculate the total value. The item value specifying unit 116 may specify the character string having the highest value among the calculated total values as the item value.

  In the correlation value between the character type and the word in the above (1), the item value specifying unit 116 specifically specifies each treatment such as the item name “surgery” and the item name “anesthesia” when the form is a medical expense receipt. For words representing names, the correlation value with item values composed of character types indicating numerical values is relatively high, and the correlation value with item values composed of character types other than numerical values is relatively low. May be set.

  In the correlation value between each character or each character string based on the ruled line information of (2), the item value specifying unit 116 specifically compares the correlation value of the adjacent item value via the ruled line surrounding the item name. May be set to be relatively high and the correlation values of the other item values to be relatively low.

  The output information generating unit 117 includes information indicating the item name specified by the item name specifying unit 114 (hereinafter, also referred to as “item name specifying information”) and information indicating the item value specified by the item value specifying unit 116 (hereinafter, referred to as “item name specifying information”). , Which is also referred to as “item value specifying information”), and generates output information to be output by the terminal 200 or the like. This output information may be, for example, fixed-length or variable-length text data listing the item names and item values of each item on the form.

  The construction unit 118 calculates a mapping model by statistically analyzing a plurality of pieces of relative position information between the item name and the item value of each item for each pattern of the form classified by the pattern classification unit 112. This relative position information is so-called learning data in constructing a mapping model. In the mapping model, for example, a position correlation value corresponding to the frequency of the relative position between the item name and the item value to be the learning data in units of 1pxl may be set.

Specifically, the construction unit 118 may construct a mapping model by the following processes (1) to (5) for each form pattern. (1) First, the construction unit 118 calculates the relative position between the item name and the item value of each item to be a plurality of learning data, x of the center coordinates (x, y) of the character string frame of the item name and the item value. The difference between the component and the y component is obtained. Hereinafter, the respective differences will be referred to as x _edges and y _edges . (2) Next, the construction unit 118 performs a process of excluding outliers for each of the x _edges and the y _edges . As a method of the exclusion processing, for example, a method of repeating a test using an Smirnov-Grubbs test until an outlier is not detected can be considered.

(3) Next, the construction unit 118 obtains the minimum and maximum values of the remaining x _edges and y _edges excluding outliers. (4) Next, the construction unit 118 constructs a mapping model by adding a predetermined margin to the obtained minimum value and maximum value and forming a two-dimensional histogram. (4) Further, the construction unit 118 may apply a Gaussian filter to the constructed mapping model to perform a smoothing process, that is, a so-called blurring process. (5) Further, the construction unit 118 may normalize the smoothed mapping model to have a maximum value of 1.

  The storage unit 120 stores the item name registration information and the mapping model. Further, the storage unit 120 indicates the image data of the form, the form pattern information, the item name specifying information, the item value specifying information, the information on the association between the item name specifying information and the item value specifying information, the character type or the correlation value between words. Information or information indicating a correlation value between each character or each character string based on the ruled line information may be stored.

  The communication unit 130 receives image data of a form from the scanner 300 and transmits output information to the terminal 200 via the network N.

<3. Operation example>
An operation example of the server device 100 according to the present embodiment will be described with reference to FIG. FIG. 6 is a flowchart showing the flow of text data processing by reading image data of a form in server device 100. The process illustrated in FIG. 6 is started in a state where the mapping model and the item name registration information are stored in the storage unit 120 in advance. The order of the processing described below is an example, and may be changed as appropriate.

  The acquisition unit 111 acquires the image data of the form (S10). When the acquisition unit 111 acquires the image data of the form, the character recognition unit 113 recognizes each character on the image data (S11).

  When the character recognizing unit 113 recognizes each of the above-mentioned characters, the item name specifying unit 114 uses the character information indicating the recognized character and the item name registration information stored in the storage unit 120 to store each item of the form. The item name is specified (S12). When the item name specifying unit 114 specifies the item name, the range specifying unit 115 uses the mapping model stored in the storage unit 120 to output, from the item values corresponding to the position of the specified item name, the form A range in which an item value whose correlation value between the position of the item name of each item and the position of the item value corresponding to the item name is equal to or larger than a threshold is specified is specified (S13).

  The item value specifying unit 116 extracts each of the characters belonging to the range specified by the range specifying unit 115, and specifies an item value corresponding to the specified item name based on the extracted characters (S14). When the item value specifying unit 116 specifies the item value, the output information generating unit 117 outputs output information for outputting the information indicating the specified item name and the information indicating the specified item value in association with each other. It is generated (S15).

<4. Hardware Configuration>
With reference to FIG. 7, an example of a hardware configuration when the above-described server device 100 and terminal 200 are realized by a computer 800 will be described. Note that the function of each device can also be realized by being divided into a plurality of devices.

  As shown in FIG. 7, the computer 800 includes a processor 801, a memory 803, a storage device 805, an input I / F unit 807, a data I / F unit 809, a communication I / F unit 811 and a display device. 813.

The processor 801 controls various processes in the computer 800 by executing a program stored in the memory 803. For example, each function unit included in the control unit 110 of the server device 100 and each function unit included in the control unit of the terminal 200 can be realized as a program that is temporarily stored in the memory 803 and that mainly operates on the processor 801. It is. The processor 801, for construction of mapping model may be used GPU (G raphics P rocessing U nit ).

  The memory 803 is a storage medium such as a RAM (Random Access Memory). The memory 803 temporarily stores a program code of a program executed by the processor 801 and data necessary for executing the program.

  The storage device 805 is a non-volatile storage medium such as a hard disk drive (HDD) or a flash memory. The storage device 805 stores an operating system and various programs for implementing the above-described configurations. In addition, the storage device 805 can also store a table for registering item name registration information and a DB for managing the table. Such programs and data are loaded into the memory 803 as needed, and are referred to by the processor 801.

  The input I / F unit 807 is a device for receiving an input from a user. Specific examples of the input I / F unit 807 include a keyboard, a mouse, a touch panel, various sensors, and a wearable device. The input I / F unit 807 may be connected to the computer 800 via an interface such as a USB (Universal Serial Bus).

  The data I / F unit 809 is a device for inputting data from outside the computer 800. A specific example of the data I / F unit 809 includes a drive device for reading data stored in various storage media. The data I / F unit 809 may be provided outside the computer 800. In that case, the data I / F unit 809 is connected to the computer 800 via an interface such as a USB.

  The communication I / F unit 811 is a device for performing data communication with a device external to the computer 800 via a network N by wire or wirelessly. The communication I / F unit 811 may be provided outside the computer 800. In that case, the communication I / F unit 811 is connected to the computer 800 via an interface such as a USB, for example.

  The display device 813 is a device for displaying various information. Specific examples of the display device 813 include, for example, a liquid crystal display, an organic EL (Electro-Luminescence) display, and a display of a wearable device. The display device 813 may be provided outside the computer 800. In that case, the display device 813 is connected to the computer 800 via, for example, a display cable or the like. In the case where a touch panel is employed as the input I / F unit 807, the display device 813 can be configured to be integrated with the input I / F unit 807.

[Second embodiment]
Next, a second embodiment of the present invention will be described. In the present embodiment, an example in which the form reading system according to the present invention is used for payment of medical insurance benefits by an insurance company will be described. Hereinafter, the points different from the first embodiment will be mainly described. It should be noted that the form reading system according to the present invention can be used not only for payment of medical insurance benefits but also for various insurance procedures.

  In this embodiment, for the sake of simplicity, in order to claim medical insurance benefits in the payment operation, the insured submits a claim to the insurance company as (a) claiming medical insurance benefits. (B) Inpatient / treatment report to report the contents of hospitalization / treatment to be billed, and (c) Medical expenses receipt issued for hospitalization / treatment to be billed And In this example, the form to be read by the form reading system is the medical expense receipt described in (c) above. The form to be read by the form reading system may be any form as long as it is a form related to medical insurance, and may be, for example, a medical statement, a discharge certificate, or a document issued during a public period.

<1. Overview>
With reference to FIG. 8, the flow of the medical insurance benefit payment operation using the form reading system according to the present embodiment will be described with the following processes (1) to (7).

  As shown in FIG. 8, (1) First, in order to apply for a medical insurance benefit claim, the contents of the paper-based claim documents (a) and (b) submitted by the insured are data-entry. The user (hereinafter, also referred to as “input person”) uses the terminal 200a to display data (hereinafter, also referred to as “application data”) indicating the contents of the above (a) and (b) on the Web screen for this application registration. "). In the case where the application for this claim has been received from a Web screen instead of a paper claim document, the insured person may be the input person of the application data.

  (2) Next, the scanner 300 reads the medical expense receipt of the paper medium (c) submitted together with the above (a) and (b) and converts it into image data by the operation of the input person. I do.

  (3) Next, the server device 500 according to the present embodiment acquires the converted image data of the medical expense receipt, and calculates the number of operations, the medical treatment date / hospitalization period, etc. described in the medical expense receipt. Identify the item name and item value of each item.

  (4) Next, the server device 500 associates the specified item name with the item value and outputs it as text data.

  (5) Next, the server device 500 collates the report content of (b) input in (1) with the text data of the receipt output in (4). Specifically, when the item to be compared is the “medical treatment day” of the report content, the server device 500 specifies the input value “2012/08/15” of the medical treatment date and the item corresponding to the medical treatment date. Is compared with the item value of the item name “issue date” specified by the server device 500, and the degree of matching indicating the degree of matching is calculated. It should be noted that the medical treatment date and the issue date are the same in the treatment of the claim target in this example.

  (6a) When the predetermined degree of matching is set to 80%, the server apparatus 500 inputs the information in (1) above if the matching result indicates that the degree of matching is 95% and exceeds the predetermined degree of matching. It is determined that there is no possibility of an input error in the contents of the report on the medical treatment date, and the process proceeds to the next report item check. (6b) On the other hand, when the matching result indicates that the matching degree is 60% and equal to or less than the predetermined matching degree, the server apparatus 500 may indicate an input error in the medical treatment day report content input in (1). It is determined that there is. The server device 500 generates input error information indicating the medical treatment date in order to notify the user of the possibility of an input error of the medical treatment date as the determination result. The terminal 200b outputs the input error information generated by the server 500 on a screen or the like.

  (7) The user confirms the input error information output using the terminal 200b on a screen or the like, and re-enters the contents of the hospitalization / treatment report on the Web screen for registering the application for a benefit claim. I do.

  According to the above configuration, the server device 500 only needs to manually re-check or re-enter the report contents entered by the insured person or entered by the input person, for only the items having an input error (mismatch). For this reason, even if the accuracy of converting the receipt into text data is not 100%, the server device 500 can compare the report content and the receipt content with the load of manually inputting the contents of the receipt data and the human eyes. The load for collation can be reduced. In addition, the server device 500 can more accurately collate the contents of the report and the contents of the receipt with the eyes of a person, so that the quality of work can be improved. In addition, by using the form reading system according to the present embodiment, the insurance company can reduce the load of the data entry business and the collation business in the life insurance benefit payment business and improve the business quality.

<2. Functional Configuration>
An example of a functional configuration of the server device 500 according to the present embodiment will be described with reference to FIG. As shown in FIG. 9, the server device 500 includes a collating unit 119 in addition to the functional units of the server device 100 according to the first embodiment. The range specifying unit 115, the item value specifying unit 116, the output information generating unit 117, and the storage unit 120 of the server device 500 have the following functions in addition to the functions of these functional units according to the first embodiment.

When performing pairing for both the medical treatment day item and the hospitalization period item, the range specifying unit 115 specifies a range in which the item value of the hospitalization period item can be located based on the result of performing the pairing of the medical treatment day item, The range may be specified by excluding the range indicated by the coordinates bbox (x _min , y _min , x _max , y _max ) of the character string frame of the character string of the medical treatment date item. According to such a configuration, in specifying the item values of the medical treatment day / hospitalization period item, the server device 500 prevents the item values of the medical treatment day item and the hospitalization period item from overlapping, and furthermore, the The accuracy of specifying the item value can be improved.

  When specifying an item value in a predetermined item of the medical expense receipt, the item value specifying unit 116 performs a pairing execution condition (hereinafter, also referred to as a “pairing execution condition”) for each item according to the characteristics or the like of the item. ) May be added. The item value specifying unit 116 performs the pairing (specifies the item value) when the pairing execution condition is satisfied.

  With reference to FIG. 10, an example of the pairing execution condition for each of the predetermined items will be described. In this example, description will be given using items indicating a medical treatment date and a hospitalization period in a medical expense receipt as predetermined items. Items indicating the medical treatment day and hospitalization period vary depending on whether the target treatment is inpatient treatment or outpatient treatment in the medical receipt, or whether the item name is `` issued '' It has the characteristic of being shown in a completely different notation from "medical care" and "hospitalization" such as "day" and "birth date". Therefore, the item value specifying unit 116 sets a pairing execution condition according to the characteristics for the item indicating the medical treatment day and the hospitalization period. Hereinafter, the items indicating the medical treatment day and the hospitalization period are also referred to as “medical treatment day item” and “hospitalization period item”, respectively, and are also collectively referred to as “medical treatment day / hospitalization period item”.

  As shown in FIG. 10, a pairing execution condition table 700 listing examples of the pairing execution conditions in the medical treatment day / hospitalization period item includes a condition No. 701 for identifying each condition, and whether or not an item name is acquired. Whether to execute the pairing set according to the item name acquisition condition 702 indicating the condition, the other condition 703 indicating the condition other than the item name acquisition condition, and the AND condition of the item name acquisition condition 702 and the other condition 703 And a pairing execution status 704 indicating whether or not the pairing is performed.

  As shown in FIG. 10, as an example, the pairing execution condition when the condition No. 701 is “P003” is as follows: (1) In the item name acquisition presence / absence condition 702, there is no acquisition of the item name of the medical treatment day item and the acquisition of the item name of the hospitalization period And (2) Other condition 703 indicates a case where the entry / exit category is “outpatient”. Under this pairing execution condition, the item value specifying unit 116 does not perform the pairing of the medical treatment day item, but performs the pairing of the hospitalization period item. Here, the “outside / outside division” is an item indicating whether the target treatment is inpatient treatment or outpatient treatment.

  As another example, the pairing execution condition whose condition No. 701 is “P005” is as follows: (1) In the item name acquisition presence / absence condition 702, when the item name of the medical treatment day item is acquired and the item name of the hospitalization period is not acquired, And (2) Other conditions 703 indicate a case where the entrance / exit category is “hospitalization” and the name of the acquired medical treatment day item is other than “issue date” and “childbirth date”. Under this pairing execution condition, the item value specifying unit 116 performs pairing using the acquired medical treatment day item name for the hospitalization period item. On the other hand, if there are a plurality of medical day items for the medical day items in the pairing execution conditions, the item value specifying unit 116 performs the second item name (performs the pairing with the hospitalization period item). (Other than the item name).

  As another example, the pairing execution condition whose condition number is “P006” is as follows: (1) In the item name acquisition presence / absence condition 702, when the item name of the medical treatment day item is acquired and the item name of the hospitalization period is not acquired, And (2) Other condition 703 indicates a case where the out-of-service category is “hospitalization” and the acquired medical treatment day item name is “issue date” or “birth date”. Under this pairing execution condition, the item value specifying unit 116 executes the pairing of the medical treatment day item and does not execute the pairing of the hospitalization period item.

  According to the above configuration, the server device 500 determines whether or not the item name of the item to be paired changes depending on the content of another item, or the character from which the item name of the item to be paired is extracted. Pairing can be performed according to the characteristics of the item to be paired even when an item name indicated by a completely different notation that is difficult to predict corresponds.

  The item value specifying unit 116 may exchange the values of the start and end timings when the start and end timings of the period are reversed in the specified item value in the item indicating the period. For example, in the hospitalization period item, if the specified start date is “2018/11/15” and the end date is inverted to “2018/11/1”, the item value specifying unit 116 The start date may be "2018/11/1" and the end date may be "2018/11/15" by exchanging the date values. According to such a configuration, the item value specified in the item indicating the period can be corrected, so that the server device 500 can specify the item value with high accuracy.

  The item value specifying unit 116 deletes a part of the character string of the item value when a predetermined character or character string that does not overlap when specifying the item value is continuous in the item indicating the date or the period. May be. Specifically, the item value specifying unit 116 determines that the predetermined characters or character strings that do not overlap are “Heisei”, “Month”, “Flat”, “Year”, “Day”, “Naru” or At least one of "H" is set, and when these characters or character strings are continuous, only the first character or character string may be left and the rest may be deleted.

  The item value specifying unit 116 determines whether an item representing a date or a period (hereinafter, also referred to as a “date item”) does not include at least one of “year” and “month” when specifying the item value. The break position of “year” or “month” may be determined and inserted. Here, an example of inserting “year” or “month” will be described with reference to FIG. As shown in FIG. 11, a year / month insertion condition table 900 listing examples of insertion of “year” or “month” includes a condition No. 901 for identifying each condition, and a condition 902 indicating the condition of the insertion. And processing contents 903 indicating the contents of processing for inserting at least one of "year" and "month" set according to the condition 902.

  As illustrated in FIG. 11, as an example, when the condition No. 901 is “D002” and the insertion condition is “D002”, and the condition 902 includes only the month for the presence or absence of the date, the item value specifying unit 116 is extracted as the item value of the date item. The break position is determined from the number and distance of the numbers, and the character “year” is inserted at the determined break position.

  As another example, if the condition No. 901 is "D003" and the insertion condition is "D003", there is no "year" and "month" for the presence / absence of the condition 902, and if the number of extracted digits is eight, the item value is specified. The unit 116 estimates that the item value of the date item is represented in the format of YYYYMMDD (year = YYYY, month = MM, day = DD), and inserts the characters “year” and “month” according to the format. I do.

  The output information generation unit 117 provides the user with input data of a procedure relating to a form whose matching degree calculated by the matching unit 119 described below is equal to or less than a predetermined matching degree (for example, the application data in the example of FIG. 8 described above). Generate input error information indicating the item name of the item. According to such a configuration, the report content entered by the input person may be re-checked or re-input by a person only for an item having an input error (mismatch). Can be reduced.

  Here, “input error information” is information indicating that there is an error in a value input by a user for each item of input data of a procedure related to a form. The input error information may list, for example, only the item names of the items corresponding to the input errors, or may add the degree of matching in addition to each item name. Here, the “matching degree” indicates the degree of matching between two item values to be compared, and is a quantification of the degree of matching in a set element of these item values. Here, the “predetermined degree of matching” is a value that sets a threshold value for determining whether or not an input item value is regarded as an error. As a specific value of the predetermined matching degree, a value such as 80 or 90% may be appropriately set according to the contents of the business using the form reading system according to the present embodiment.

  The collation unit 119 includes an input item value of each item of the input data of the procedure related to the form and an item value specifying unit 116 of the item name specified by the item name specifying unit 114 corresponding to the item name of each item. Is compared with the specified item value, and the degree of matching indicating the degree of matching between the input item value and the specified item value is calculated.

  The storage unit 120 stores item input information indicating an item name and an item value input by an input user in each item of input data of a procedure related to a form such as insurance application data. Here, the “item input information” is information indicating an item name of each item of input data of a procedure related to a form associated with an item value input by an input person corresponding to the item name. For example, in the case of a DB table, the item input information is the item name “medical date” in the example of FIG. 8 and the item value (input value) “2012/08/15” input by the input person corresponding to the item name. Is included in the same record.

  The storage unit 120 may store, for example, information indicating the pairing execution condition table 700 shown in the example of FIG. 10 and the year / month insertion condition table 900 shown in the example of FIG.

  Note that the present embodiment is an exemplification for describing the present invention, and is not intended to limit the present invention to only the embodiment. Further, the present invention can be variously modified without departing from the gist thereof. Further, those skilled in the art can adopt an embodiment in which each element described below is replaced with an equivalent one, and such an embodiment is also included in the scope of the present invention.

[Modification]
Although the present invention has been described based on the above embodiment, the following cases are also included in the present invention.

  (1) The item value specifying unit 116 according to the above-described embodiment performs the re-specifying process by limiting the target character type and range when a plurality of item values paired with the specified item name exist. You may. The item value specifying unit 116 specifically (1) converts each character belonging to the range of the character string frame of a plurality of paired item values into an item corresponding to the item name specified by the item name specifying unit 114. Extract only characters that can take values. (2) A range is specified by connecting the coordinates of the character frames of the extracted characters. (3) Thereafter, similarly to the above embodiment, each character belonging to the specified range is extracted, and the item value is specified again based on the extracted characters. According to such a configuration, the item value for the item name of each item can be specified with higher accuracy.

  (2) At least a part of each configuration in the server devices 100 and 500 according to the above embodiment may be included in the terminal 200 (200a, 200b). For example, the function of the matching unit 119 of the server device 500 may be provided in the control unit of the terminal 200. That is, the terminal 200 compares the input item value of each item of the input data of the procedure related to the form with the specified item value of the specified item name corresponding to the item name of each item. May be. In the collation processing on the terminal 200 side, the server device 500 transmits the item name specifying information and the item value specifying information to the terminal 200. Further, for example, the function of the output information generation unit 117 of the server device 100 or 500 may be provided in the control unit of the terminal 200. That is, the terminal 200 may generate output information or input error information to be output on a screen or the like. For the generation processing on the terminal 200 side, the server devices 100 and 500 transmit, for example, information in which the item name specifying information and the item value specifying information are associated with each other to the terminal 200. In addition, when the function of the matching unit 119 is provided on the server device side, the server devices 100 and 500 generate input error information on the terminal 200 side. Information indicating the collation result between the input item value of the item and the item value of each of the specified items is transmitted to the terminal 200.

  DESCRIPTION OF SYMBOLS 1 ... Form reading system, 100, 500 ... Server apparatus (information processing apparatus), 110 ... Control part, 111 ... Acquisition part, 112 ... Pattern classification part, 113 ... Character recognition part, 114 ... Item name specification part, 115 ... Range Specification unit, 116: item value specification unit, 117: output information generation unit, 118: construction unit, 119: collation unit, 120: storage unit, 130: communication unit, 200, 200a, 200b: terminal, 300: scanner, 800 ... Computer, 801 ... Processor, 803 ... Memory, 805 ... Storage device, 807 ... Input I / F unit, 809 ... Data I / F unit, 811 ... Communication I / F unit, 813 ... Display device

Claims (7)

A mapping model representing a correlation value between the position of the item name of each item of the form and the position of the item value corresponding to the item name, and a storage unit for storing item name registration information indicating the item name,
An acquisition unit that acquires image data of an irregular form,
A character recognition unit that recognizes each character on the image data,
An item name specifying unit that specifies an item name of each item of the form based on the character information indicating the recognized characters and the item name registration information;
Using the mapping model, of the item values corresponding to the position of the specified item name, a range specifying unit that specifies a range in which the correlation value can be located at an item value equal to or greater than a threshold,
An item value specifying unit that extracts the characters belonging to the specified range and specifies an item value corresponding to the specified item name based on the extracted characters.
An output information generation unit that generates output information for outputting information indicating the specified item name and information indicating the specified item value in association with each other,
Information processing device. The storage unit stores form pattern information indicating a pattern of the form,
Based on the image data, a pattern classification unit that classifies the pattern of the form shown in the image data,
The mapping model is configured such that, for each pattern of the form, a correlation value between the position of the item name and the position of the item value calculated by statistically analyzing a plurality of pieces of relative position information of the item name and the item value of each item. Is mapped on the form,
The range specifying unit specifies, using the mapping model corresponding to the classified form pattern, a range indicating a position of an item value equal to or more than the predetermined correlation corresponding to the position of the specified item name. ,
The information processing device according to claim 1. The range specifying unit excludes a range in which the specified item name is located, and specifies a range in which an item value corresponding to the specified item name can be located,
The information processing device according to claim 1. The storage unit stores item input information indicating an item name and an item value input by an input user in each item of input data of a procedure related to the form,
The input item value of each item of the input data is compared with the specified item value of the specified item name corresponding to the item name of each item, and the input item value and the input item value are compared. A matching unit that calculates a degree of matching indicating the degree of matching with the specified item value;
The output information generation unit generates, for a user, input error information indicating an item name of an item of the input data in which the calculated matching degree is equal to or less than a predetermined matching degree,
The information processing device according to claim 1. Computer
Storing a mapping model representing a correlation value between the position of the item name of each item of the form and the position of the item value corresponding to the item name, and storing item name registration information indicating the item name;
Obtaining image data of an irregular form,
Recognizing each character on the image data;
Identifying the item name of each item of the form based on the character information indicating the recognized characters and the item name registration information;
Using the mapping model, among the item values corresponding to the position of the specified item name, specifying a range in which the correlation value can be located at an item value equal to or greater than a threshold,
Extracting the characters belonging to the specified range, and specifying an item value corresponding to the specified item name based on the extracted characters;
Generating output information for outputting information indicating the specified item name and information indicating the specified item value in association with each other,
Information processing method. On the computer,
A mapping model representing a correlation value between the position of the item name of each item of the form and the position of the item value corresponding to the item name, and a storage function of storing item name registration information indicating the item name,
An acquisition function for acquiring image data of an irregular form,
A character recognition function for recognizing each character on the image data,
An item name specifying function for specifying an item name of each item of the form based on the character information indicating the recognized characters and the item name registration information;
Using the mapping model, of the item values corresponding to the positions of the specified item names, a range specifying function for specifying a range in which the correlation value can be located at or above a threshold item value,
An item value specifying function of extracting the characters belonging to the specified range and specifying an item value corresponding to the specified item name based on the extracted characters;
An output information generation function of generating output information for outputting information indicating the specified item name and information indicating the specified item value in association with each other,
program. A mapping model representing a correlation value between the position of the item name of each item of the insurance-related form and the position of the item value corresponding to the item name; item name registration information indicating the item name; A storage unit that stores an item name and an item input information indicating an item value input by an input person in each item of the application data of the insurance procedure,
An acquisition unit that acquires image data of a form related to atypical insurance,
A character recognition unit that recognizes each character on the image data,
An item name specifying unit that specifies an item name of each item of the form based on the character information indicating the recognized characters and the item name registration information;
Using the mapping model, of the item values corresponding to the position of the specified item name, a range specifying unit that specifies a range in which the correlation value can be located at an item value equal to or greater than a threshold,
An item value specifying unit that extracts the characters belonging to the specified range and specifies an item value corresponding to the specified item name based on the extracted characters.
The item value of each item input from the input person of the application data is compared with the specified item value of the specified item name from the image data of the form corresponding to the item name of each item. A matching unit that calculates a degree of matching indicating the degree of matching between the input item value and the specified item value;
An output information generating unit that generates input error information indicating an item name of the application data in which the calculated degree of match is equal to or less than a predetermined degree of match for the user.
Form reading system.