JP6462930B1 - Character recognition apparatus, method and program - Google Patents

Abstract

A character recognition apparatus, method, and program capable of calculating likelihood for each item and changing a display mode of a read item based on the calculated likelihood.
A scan item setting unit 231 for setting a scan item for scanning image data converted by scanning a document and an OCR function, and a character displayed in the scan item of the image data A character recognition unit 232 that reads information by OCR to generate text data, a text data display unit 233 that displays image data and text data of a reading item in parallel, and a likelihood calculation unit that calculates the likelihood of text data A character recognition device 200 is provided that includes a control unit 230 that includes a display change unit 235 that changes the display mode of the displayed text data.
[Selection] Figure 2

Description

  The present disclosure relates to a character recognition apparatus, method, and program for reading character information from image data.

  2. Description of the Related Art A technique for generating digital data in which input information is converted into a predetermined character code by reading a document with handwritten characters by an image scanner or the like and performing OCR (Optical Character Recognition) processing has become widespread.

  A character identification system that performs character recognition by machine learning from image data obtained by scanning a handwritten document or the like as an image is disclosed (for example, see Patent Document 1). In the character recognition system disclosed in Patent Document 1, a character image input receiving unit that receives an input of a sample character image, a character component extraction that extracts a character component based on the sample character image, and a pseudo character based on the character component There has been proposed a system including a pseudo character model generation unit that generates a model, and an identification dictionary generation that generates a character identification pattern based on the pseudo character model and generates an identification dictionary.

Japanese Patent Application Laid-Open No. 2015-069256

  By the way, since the correct answer rate of the character recognition by OCR is not necessarily perfect, after performing a character recognition process, the human visual confirmation is generally performed. At that time, the accuracy of character recognition (likelihood of characters that have undergone character recognition processing) varies from character to character, but it is inefficient because all items and characters are checked uniformly regardless of the likelihood. there were.

  Therefore, in the present disclosure, when character recognition is performed by OCR processing from image data obtained by scanning a handwritten document or the like as an image, the likelihood is calculated for each item, and the display mode of the read item is set based on the calculated likelihood. A character recognition apparatus, method, and program capable of generating text data with high accuracy without performing unnecessary visual confirmation processing by grasping the reliability of the reading processing for each reading item by changing the data will be described.

A character recognition device according to an aspect of the present disclosure is a character recognition device that reads character information from image data obtained by scanning a document as an image, and includes character information displayed in a reading item set in the image data of the document. Character recognition unit that performs character recognition and generates text data, and displays the image data and text data of the read item for each read item in a comparable manner, and edits the text data by inputting characters into the text data A text data display unit for enabling a state, a likelihood calculation unit for calculating the likelihood of character recognition for each reading item, and a display change for changing the mode of the place where the text data is displayed based on the calculated likelihood e Bei and parts, and the display changing unit for reading items of text data likelihood for each read item is equal to or less than a predetermined first threshold value, the text data The display column is changed to a setting not to display a character string, and the image data and text data of the read item whose likelihood for each read item is equal to or greater than a predetermined second threshold value is not displayed, and the likelihood for each read item is the first. If greater than the first threshold value is less than the second threshold value, Ru changing the manner of a portion displaying the text data of the read item in accordance with the likelihood of each read entry.

A character recognition method according to an aspect of the present disclosure is a character recognition method for reading character information from image data obtained by scanning a document as an image, and is displayed on a reading item set in the image data of the document, which is performed by a character recognition unit. The character recognition step for generating the text data by performing the character recognition of the character information that has been read, and the text data display unit, for each reading item, the image data of the reading item and the text data are displayed so as to be comparable, and the text data A text data display step for making the text data editable by inputting characters to the likelihood, a likelihood calculating step for calculating the likelihood of character recognition for each reading item, and a calculation performed by the likelihood calculating unit based on the likelihoods, performed by the display change section, e Bei a display change step of changing the mode of a portion displaying text data, the display In the further step, the text data reading item whose likelihood for each reading item is equal to or less than a predetermined first threshold value is changed to a setting for not displaying a character string from the text data display column, and the likelihood for each reading item is changed. When the image data and text data of the reading item equal to or greater than the predetermined second threshold are not displayed, and the likelihood for each reading item is greater than the first threshold and less than the second threshold, the likelihood for each reading item depending on the Ru changing the manner of a portion displaying the text data of the read item.

A character recognition program according to an aspect of the present disclosure is a character recognition program that reads character information from image data obtained by scanning a document as an image, and the characters displayed in the reading items set in the image data of the document performs character recognition of the information, text data by performing a character recognition step of generating text data, read comparably displaying the read image data and text data entry for each item, the character input to the text data The text data display step for making the text editable, the likelihood calculation step for calculating the likelihood of character recognition for each reading item, and the mode of the place where the text data is displayed is changed based on the calculated likelihood is executed and a display changing step, the electronic computer, the display change step, the likelihood of each read entry is given 1 For text data reading items that are less than or equal to the threshold value, the text data display column is changed to a setting that does not display a character string, and the image data and text data of the reading item that has a likelihood for each reading item equal to or greater than a predetermined second threshold value. Is not displayed and the likelihood of each reading item is larger than the first threshold value and less than the second threshold value, the mode of the portion where the text data of the reading item is displayed is changed according to the likelihood of each reading item. to.

  According to the present disclosure, when character recognition is performed by OCR processing from image data obtained by scanning a handwritten document or the like as an image, the likelihood is calculated for each item, and the display mode of the read item is set based on the calculated likelihood. Since it can be changed, it is possible to grasp the reliability of the reading process for each reading item. Thereby, it is possible to generate highly accurate text data without performing useless visual confirmation processing.

It is a functional block diagram showing a character recognition system concerning one embodiment of this indication. It is a functional block diagram which shows an example of a structure of the character recognition apparatus 200 shown in FIG. It is a schematic diagram which shows the account transfer request document which is an example of the image data memorize | stored in image data DB221 of FIG. It is a schematic diagram which shows the example of the state by which the image data of the account transfer request document of FIG. 3 is displayed on the display part. FIG. 5 is a schematic diagram illustrating an example of a state in which image data and text data of a reading item extracted from the image data of FIG. 4 are displayed in parallel. It is a schematic diagram which shows the mode of the feature extraction and vector conversion by the character recognition part 232 of FIG. It is a schematic diagram which shows the mode of determination of the character type by the character recognition part 232 of FIG. FIG. 6 is a schematic diagram illustrating an example of a state in which color is applied to the right end of text data from a state in which image data and text data of a reading item in FIG. It is a flowchart which shows the character recognition method by the character recognition system 1 shown in FIG. It is a figure which shows the example of the financial institution table memorize | stored in the memory | storage part 240 shown in FIG. It is a mimetic diagram showing an example of a state where image data of a reading item extracted from image data and text data are displayed in parallel in a character recognition system concerning one embodiment of this indication. It is a mimetic diagram showing an example of a state where image data of a reading item extracted from image data and text data are displayed in parallel in a character recognition system concerning one embodiment of this indication. It is a mimetic diagram showing an example of a state where image data of a reading item extracted from image data and text data are displayed in parallel in a character recognition system concerning one embodiment of this indication.

  An embodiment of the present disclosure will be described with reference to the drawings. Note that the embodiments described below do not unduly limit the content of the present disclosure described in the claims. In addition, all the components shown in the embodiments are not necessarily essential components of the present disclosure.

(Embodiment 1)
<Configuration>
FIG. 1 is a block diagram of a character recognition system 1 according to the first embodiment of the present disclosure. The character recognition system 1 is a system that reads character information from image data obtained by scanning, for example, a handwritten application form or an account transfer request form as an image, and receives an application form or an account transfer request form from a customer. It is used by a user company to read handwritten character information described in an application form or fund transfer request form.

  The character recognition system 1 includes a user system 100, a character recognition device 200, and a network NW. The user system 100 and the character recognition device 200 are connected via a network NW. The network NW is configured by the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), or the like. The network NW may be wired communication or wireless communication, and includes a communication method called 4G such as LTE (Long Term Evolution) and a communication method based on 5G.

  The user system 100 scans a document such as an application form or an account transfer request form composed of a plurality of items, and converts it into image data. The user system 100 includes a scanner device 110 and a user terminal 120, and is connected to be communicable with each other via, for example, USB (registered trademark) or a LAN.

  The scanner device 110 is a device that scans a document such as an application form or an account transfer request form and converts it into image data. In the first embodiment, the scanner device is used. However, any device that can convert a paper document into electronic data may be used. For example, a camera may be used. Image data scanned by the scanner device 110 is stored in an image data DB 221 described later.

  The user terminal 120 is a terminal installed in a user company, and sets a reading item for image data by a user operation, and performs a visual check on text data in which character recognition of the reading item is performed. It is. The user terminal 120 includes a display unit that displays image data, and the character recognition device 200 is accessed via the network NW by operating the operation unit, and various programs are started and provided. The display unit includes a display and the operation unit includes a keyboard and a mouse.

  FIG. 2 is a functional block diagram showing an example of the configuration of the character recognition device 200 shown in FIG. The character recognition device 200 generates, as text data, character information displayed in the reading item on the image data based on the set reading item. The character recognition device 200 includes a communication unit 210, a storage unit 220, and a control unit 230.

  The communication unit 210 is a communication interface for communicating with the user system 100 via the network NW, and performs communication according to a communication protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol).

  The storage unit 220 stores programs for executing various control processes and functions in the control unit 250, input data, and the like, and includes a RAM (Random Access Memory), a ROM (Read Only Memory), and the like. The In addition, the storage unit 220 temporarily stores image data of a reading item displayed by a text data display unit 233 described later and the likelihood calculated by the likelihood calculation unit 234.

  Further, the storage unit 220 includes an image data DB 221 that stores image data converted by the scanner device 110, a reading item DB 222 that stores reading items set for the image data, and a text generated by reading the reading items. A text data DB 223 for storing data. The image data DB 221, the reading item DB 222, and the text data DB 223 are databases that can be accessed and referenced by various programs of the control unit 230.

  The control unit 230 controls the overall operation of the character recognition apparatus 200 by executing a program stored in the storage unit 220. From the CPU (Central Processing Unit), the GPU (Graphics Processing Unit), and the like. Composed. As a function of a program executed by the control unit 230, a reading item setting unit 231, a character recognition unit 232, a text data display unit 233, a likelihood calculation unit 234, and a display change unit 235 are provided. The reading item setting unit 231, the character recognition unit 232, the text data display unit 233, the likelihood calculation unit 234, and the display change unit 235 are activated and executed by a program stored in the storage unit 220.

  The reading item setting unit 231 sets a reading item to be read for the image data stored in the image data DB 221 by operating the operation unit of the user terminal 120. The setting of the reading item is performed by, for example, a range setting by so-called drag and drop by operating an operation unit (for example, a mouse) on the display unit of the user terminal 120 on which image data is displayed. The set reading item is stored in the reading item DB 222.

  FIG. 3 is a schematic diagram showing an account transfer request form which is an example of image data stored in the image data DB 221 of FIG. This account transfer request form is an application form for requesting account transfer to a financial institution for the deduction of utility bills, etc., and includes the name reading field A1, name kanji entry field A2, financial An institution name entry field A3, a branch name entry field A4, a financial institution code entry field A5, a branch code entry field A6, a deposit item entry field A7, and an account number entry field A8 are provided. As an example of these entries, “Tokkyo Taro” is in the name entry field A1, “Taro Patent” is in the name Kanji entry field A2, and “Mizuho (Bank) is in the bank name entry field A3”. ”,“ Kojimachi (branch) ”in the branch name entry field A4,“ 0001 ”in the financial institution code entry field A5,“ 021 ”in the branch code entry field A6, and the account number entry field A8. “1111111” is written by hand, and “ordinary” is marked with “◯” in the deposit item entry field A7.

  In the user company, when performing the process of converting the account transfer request form into a text, such an account transfer request form is scanned using the scanner device 110 and image data is stored in the image data DB 221.

  FIG. 4 is a schematic diagram illustrating an example of a state in which the image data of the account transfer request document in FIG. 3 is displayed on the display unit. The image data of the account transfer request form shown in FIG. 3 is displayed on the screen P1 of the display unit of the user terminal 120. The name reading field A1, the name kanji entry field A2, and the financial information similar to the entry items of the account transfer request form are displayed. An institution name entry field A3, a branch name entry field A4, a financial institution code entry field A5, a branch code entry field A6, a deposit item entry field A7, and an account number entry field A8 are displayed.

  In this state, the operation unit of the user terminal 120 is operated to perform drag and drop, and the range of the reading item is set as indicated by a broken line in FIG. For example, name reading item S1, name kanji reading item S2, financial institution name reading item S3, branch name reading item S4, financial institution code reading item S5, branch code reading item S6, and account number reading item S7 are ranges as reading items. Is set.

  Here, the method for setting the range of the read item by the user's operation has been described, but the read item may be automatically set by the character recognition device 200. For example, in the account transfer request form of FIG. 3, characters entered by handwriting other than printed characters and ruled lines may be recognized, and the range may be set for each character recognition. As a result, character recognition is possible even when it is written out of the printed ruled line.

  The character recognition unit 232 has an OCR function, reads image data stored in the image data DB 221, and reads text information displayed in the reading item stored in the reading item DB 222 by OCR to generate text data. To do. The generated text data is stored in the text data DB 223.

  In the example shown in FIG. 3, “Tokyo Taro” is generated as text data as the contents described in the name reading item S1. Similarly, “patent Taro” is written as the name Kanji reading item S2, “Mizuho (bank)” is written as the financial institution name reading item S3, and “Kashiwamachi (branch)” is written as the branch name reading item S4. ”Is generated as text data“ 0001 ”as the description contents of the financial institution code reading item S5,“ 021 ”as the description contents of the branch code reading item S6, and“ 1111111 ”as the description contents of the account number reading item S7. The

  The text data display unit 233 extracts the image data of the reading item stored in the reading item DB 222 from the image data stored in the image data DB 221. The text data display unit 233 reads the text data stored in the text data DB 223 and displays the image data and text data of the read item on the display unit of the user terminal 120 in parallel.

  FIG. 5 is a schematic diagram showing an example of a state in which the image data of the reading item extracted from the image data of FIG. 4 and the text data are displayed in parallel. On the left side of the screen P1 of the display unit of the user terminal 120, as in FIG. 4, a full name entry column A1, a full name Chinese character entry column A2, a financial institution name entry column A3, a branch name entry column A4, and a financial institution code entry column A5, branch code entry field A6, deposit type entry field A7, and account number entry field A8 are displayed.

  Further, on the right side of the screen P1, as an example of the image data of the reading items extracted from the image data, the name reading item S1, the name kanji reading item S2, the financial institution name reading item S3, the branch name reading item S4, the financial An institution code reading item S5, a branch code reading item S6, and an account number reading item S7 are displayed.

  Further, as examples of text data read from these reading items, name reading text T1, name kanji text T2, financial institution name text T3, branch name text T4, financial institution code text T5, branch code text T6, account number text T7 is displayed in parallel below the image data of each reading item. The reason for displaying in this way is to make it easier to compare the image data of each reading item and the text data that is the reading result, thereby making it easier to confirm that text data is being generated accurately. .

  The image data and text data of the read item may be displayed in parallel with the image data as shown in FIG. 5, or may be displayed as a separate screen (window) on the screen P1.

  The likelihood calculating unit 234 calculates the likelihood of the text data for each reading item displayed on the screen P1. At this time, the likelihood of the text data for each reading item is calculated by multiplying the likelihood calculated for each character constituting the text data by a predetermined calculation, for example, the likelihood for each character. Note that the likelihood of the text data may be calculated by other methods, for example, by calculating an average value of the likelihood for each character.

  Here, the relationship between character identification and likelihood in the OCR function performed by the character recognition unit 232 will be described. FIG. 6 is a schematic diagram showing the feature extraction and vector conversion by the character recognition unit 232 of FIG. FIG. 7 is a schematic diagram showing a state of character type determination by the character recognition unit 232 of FIG.

As shown in FIG. 6, the character recognition unit 232 performs feature extraction on the image data of one cut character pattern. Then, the direction component of the character stroke is extracted, and the image data is converted into one vector on the feature space. In the example shown in FIG. 6, a state in which image data X is input to a multilayer neural network and features are extracted by capturing features such as directions and positions is schematically shown. Further, in the example illustrated in FIG. 6, the state of being converted into the vector X ₁ , the vector X _2, and the vector X ₃ is schematically illustrated.

  Next, as shown in FIG. 7, the character recognition unit 232 determines what the character type is based on the converted vector. For example, in this determination, refer to the dictionary data that holds which character type is distributed in which area on the feature space from the state of distribution using a large number of patterns in advance, with unknown input patterns A candidate for certain image data is determined. The example shown in FIG. 7 conceptually shows that the information of the character type “medium”, the character type “Sen”, and the character type “ten” is stored in the dictionary data. This indicates that the likelihood of the species is high.

  The character recognizing unit 232 acquires a plurality of text candidates (for example, medium, deity, and ten) by the above process. Then, a likelihood indicating the likelihood of each text candidate is calculated for each character. The likelihood calculation unit 234 calculates the likelihood of the read item from the likelihood for each character. The likelihood of each text candidate can be calculated by the distance between the center of each candidate in the feature space and image data that is an unknown input pattern.

  The display change unit 235 changes the display mode of the text data displayed by the text data display unit 233 based on the likelihood of the text data calculated by the likelihood calculation unit 234. For example, a color is applied to the right end of the frame of each text data displayed on the screen P2, and the color is changed based on the likelihood of the text data.

  FIG. 8 is a schematic diagram illustrating an example of a state in which color is applied to the right end of the text data from the state in which the image data and text data of the reading item in FIG. 5 are displayed in parallel. The name reading text T1, name kanji text T2, financial institution name text T3, branch name text T4, financial institution code text T5, branch code text T6, and account number text T7 displayed by the text data display unit 233 include: As shown in FIG. 8, the name name coloring portion T11, the name Chinese character coloring portion T21, the financial institution name coloring portion T31, the branch name coloring portion T41, the financial institution code coloring portion T51, the branch code coloring portion T61, and the account number coloring portion T71. Is provided.

  This name reading portion T11, name kanji coloring portion T21, financial institution name coloring portion T31, branch name coloring portion T41, financial institution code coloring portion T51, branch code coloring portion T61, account number coloring portion T71 is a likelihood calculating portion. Based on the likelihood of the name reading text T1, the name kanji text T2, the financial institution name text T3, the branch name text T4, the financial institution code text T5, the branch code text T6, and the account number text T7 calculated by H.234, for example, When the degree is 0 to 0.6, it is colored red, when it is 0.6 to 0.8, it is colored orange, and when it is 0.8 or more, it is colored yellow. Further, the higher the likelihood, the smaller the display. In this way, the reason for changing the mode is that the higher the likelihood of the text data, the more accurately the conversion process of the read item to the text data is performed, so it is possible to determine the likelihood of the read item. This is to make it possible to grasp at a glance.

<Process flow>
Hereinafter, an example of a character recognition method executed by the character recognition system 1 will be described with reference to FIG. FIG. 9 is a flowchart showing a character recognition method by the character recognition system 1 shown in FIG.

  As the processing of step S101, the reading item setting unit 231 sets reading items for reading and generating text data for the image data stored in the image data DB 221.

  At this time, for example, only katakana is described in the name reading item S1 shown in FIG. 4, and only numbers are described in the financial institution code reading item S5, the branch code reading item S6, and the account number reading item S7. It is also possible to set so as to limit the character types of the text data to be generated. The financial institution name reading item S3 is filled with the name of the financial institution corresponding to the value of the financial institution code reading item S5, and the branch name reading item S4 is filled with the name of the branch corresponding to the value of the branch code reading item S6. Therefore, for example, it is possible to set to select the corresponding text and value from the financial institution table as shown in FIG. The set reading item is stored in the reading item DB 222.

  In step S102, the character recognition unit 232 reads the image data stored in the image data DB 221, reads the character information displayed in the reading item stored in the reading item DB 222 by OCR, Text data is generated. Specifically, for example, feature extraction is performed on image data of a character pattern, a direction component of a character stroke is extracted and converted into one vector on the feature space, and which character type is on the feature space. A character type having a high likelihood is determined by referring to dictionary data that holds the distribution of which of the characters.

  As the processing in step S103, the text data display unit 233 extracts the image data of the reading item stored in the reading item DB 222 from the image data stored in the image data DB 221, and stores it in the text data DB 223. The read text data is read, and the image data and text data of the read item are displayed in parallel as shown in a screen P1 shown in FIG.

  In the example of FIG. 5, the full name reading item S1 and full name reading text T1, the full name kanji reading item S2 and full name kanji text T2, the financial institution name reading item S3, and the financial institution name text T3 are read branch names. The item S4 and the branch name text T4, the financial institution code reading item S5 and the financial institution code text T5, the branch code reading item S6 and the branch code text T6, the account number reading item S7 and the account number text T7. Each is displayed in parallel.

  As the processing of step S104, the likelihood calculating unit 234 multiplies the likelihood for each character calculated for each character constituting the text data stored in the text data DB 223, and calculates the likelihood of the read item. calculate.

  As the processing in step S105, the display changing unit 235, based on the likelihood of the text data calculated by the likelihood calculating unit 234, as shown in FIG. 8, the name reading text T1, the name kanji text T2, the financial institution name text T3, branch name text T4, financial institution code text T5, branch code text T6, and account number text T7, respectively, name full color part T11, name kanji color part T21, financial institution name color part T31, branch name color part The color and thickness of the part T41, financial institution code coloring part T51, branch code coloring part T61, and account number coloring part T71 are changed. For example, when the calculated likelihood of the text data is 0 to 0.6, it is colored red, when it is 0.6 to 0.8, it is colored orange, and when it is 0.8 or more, it is colored yellow.

  As described above, the character recognition system according to the present embodiment multiplies the likelihood calculated for each character constituting the text data by the amount of the characters constituting the text data, and determines the text data of the reading item. Calculate the likelihood. This makes it possible to evaluate the reading reliability for each reading item from the likelihood of the read text data.

  In addition, based on the likelihood of the text data for each reading item, color is applied to the right edge of the displayed text data frame, and the mode is changed by a method such as changing the color based on the likelihood of the text data. Therefore, the likelihood of text data can be grasped at a glance. Thereby, since the reliability for each reading item can be grasped at a glance, it is possible to contribute to improving the efficiency of visual confirmation, for example, checking only reading items with low reliability.

(Embodiment 2)
<Configuration>
FIG. 11 is a schematic diagram illustrating an example of a state in which image data and text data of a reading item extracted from image data are displayed in parallel in the character recognition system according to the second embodiment of the present disclosure. The character recognition system 1 has the same configuration as that of the first embodiment, but the display mode of the text data displayed by the text data display unit 233 changed by the display change unit 235 is different. In the second embodiment, when the likelihood of the text data for each reading item calculated by the likelihood calculating unit 234 is equal to or less than a predetermined threshold (first threshold) , the display changing unit 235 reads the text data of the reading item. Is different from the first embodiment in that no display is performed (the display is made blank).

  As shown in FIG. 11, on the screen P1, as in FIG. 5 of the first embodiment, the name reading item S1, the name kanji reading item S2, the financial institution name reading item S3, the branch name reading item S4, the financial institution code Reading item S5, branch code reading item S6, account number reading item S7, name reading text T1, name kanji text T2, financial institution name text T3, branch name text T4, financial institution code text T5, branch code text T6, account number Although the text T7 is displayed, the text data of the name reading text T1 and the name kanji text T2 are not displayed and are in a blank (blank) state. In this example, since the likelihood of the text data of the name reading text T1 and the name kanji text T2 is equal to or less than a predetermined threshold, the display changing unit 235 does not display.

  In this way, the reason why text data below a predetermined threshold is not displayed is because the likelihood of these text data is low and it is considered that there is no certain reliability. It is necessary to redo. Therefore, by not displaying the text data from the beginning, it is possible to omit extra visual confirmation processing and to efficiently input characters. Other configurations and processing flows are the same as those in the first embodiment.

  According to the present embodiment, in addition to the effects of the first embodiment, when the likelihood of the text data calculated by the likelihood calculating unit 234 is equal to or less than a predetermined threshold, the display changing unit 235 does not display the text data. Therefore, it is possible to omit the display of text data for reading items that do not have a certain level of reliability. As a result, unnecessary visual confirmation processing can be omitted, and character input can be performed efficiently.

(Embodiment 3)
<Configuration>
FIG. 12 is a schematic diagram illustrating an example of a state in which image data and text data of a reading item extracted from image data are displayed in parallel in the character recognition system according to the third embodiment of the present disclosure. The character recognition system 1 has the same configuration as that of the first embodiment, but the display mode of the text data displayed by the text data display unit 233 changed by the display change unit 235 is different. In the third embodiment, the display changing unit 235 colors the coloring portion provided in each text display column according to the likelihood of the text data for each reading item, and changes the character mode according to the likelihood for each character. It differs from the first embodiment in that it is changed. The character mode is changed by, for example, changing the character color, changing the character font (character font or character thickness), changing the character size, or the like.

  As shown in FIG. 12, on the screen P1, as in FIG. 5 of the first embodiment, the name reading item S1, the name kanji reading item S2, the financial institution name reading item S3, the branch name reading item S4, the financial institution code Reading item S5, branch code reading item S6, account number reading item S7, name reading text T1, name kanji text T2, financial institution name text T3, branch name text T4, financial institution code text T5, branch code text T6, account number The text T7 is displayed. Of the “Tokkyo Taro” displayed in the name reading text T1, the “G” part, the “Kyo” part, and the “Taro” part have different colors. Is displayed. In addition, among the “patent Taro” displayed in the name kanji text T2, the “patent” portion and the “taro” portion are displayed with different thicknesses. For example, according to the likelihood of each character, for example, when the likelihood is 0 to 0.6, it is bold red, and when it is 0.6 to 0.8, it is normal red, 0.8 or more In the case of, it is displayed in black of normal thickness.

  In this way, because the likelihood is different depending on the character in the text data, it is displayed in a different manner depending on the likelihood of each character, so it is possible to grasp at a glance how much the likelihood of the character was This is to make it possible. Other configurations and processing flows are the same as those in the first embodiment.

  According to the present embodiment, in addition to the effects of the first embodiment, the character mode is changed according to the likelihood of each character constituting the text data, so it is possible to determine at a glance how much the likelihood of the character was. It becomes possible to grasp. This makes it possible to evaluate the reading reliability for each character.

(Embodiment 4)
<Configuration>
FIG. 13 is a schematic diagram illustrating an example of a state in which image data and text data of a reading item extracted from image data are displayed in parallel in the character recognition system according to the fourth embodiment of the present disclosure. The character recognition system 1 has the same configuration as that of the first embodiment, but the display mode of the text data displayed by the text data display unit 233 changed by the display change unit 235 is different. In the fourth embodiment, when the likelihood of the text data calculated by the likelihood calculating unit 234 is greater than or equal to a predetermined threshold (second threshold) , the display changing unit 235 performs the processing between the image data of the reading item and the text data. This is different from the first embodiment in that no display is performed.

  As shown in FIG. 13, a full name reading item S1, a full name kanji reading item S2, an account number reading item S7, a full name reading text T1, a full name kanji text T2, and an account number text T7 are displayed on the screen P1. 5, financial institution name reading item S3, branch name reading item S4, financial institution code reading item S5, branch code reading item S6, financial institution name text T3, branch name text T4, financial institution code text T5, branch code The text T6 is not displayed. In this example, since the likelihood of the text data of the financial institution name text T3, the branch name text T4, the financial institution code text T5, and the branch code text T6 is equal to or greater than a predetermined threshold value, the display changing unit 235 does not perform display. .

  As described above, the reason why the image data and text data of the read item that is equal to or greater than the predetermined threshold is not displayed is that the likelihood of the text data is high and it is considered that there is a certain level of reliability. It can be omitted. Therefore, these items are not displayed on the screen P2, but only text data that can be confirmed by humans and individually corrected manually is displayed, thereby enabling efficient character input. Because. Other configurations and processing flows are the same as those in the first embodiment.

  According to the present embodiment, in addition to the effects of the first embodiment, when the likelihood of the text data calculated by the likelihood calculation unit 234 is equal to or greater than a predetermined threshold, the display change unit 235 displays the image data and text of the read item. Since the data is not displayed, the confirmation of the text data can be omitted for the read item having a certain reliability. This makes it possible to input characters efficiently.

  As another embodiment, it is also possible to connect the scanner device directly to the character recognition device and configure the character recognition device standalone without going through a network. This is effective when it is desired to configure the character recognition device without connecting to the network in order to prevent information leakage to the outside.

  While the embodiments according to the disclosure have been described above, these can be implemented in various other forms, and can be implemented with various omissions, substitutions, and changes. Moreover, it can also implement combining the structure described in Embodiment 1-4. These embodiments and modifications, as well as those obtained by omission, substitution, and change are included in the technical scope of the claims and their equivalents.

1 character recognition system, 100 user system, 110 scanner device, 120 user terminal, 200 character recognition device, 210 communication unit, 220 storage unit, 221 image data DB, 222 reading item DB, 223 text data DB, 230 control unit, 231 Reading item setting unit, 232 character recognition unit, 233 text data display unit, 234 likelihood calculation unit, 235 display change unit, NW network

Claims (9)

A character recognition device that reads character information from image data obtained by scanning a document as an image,
A character recognition unit that performs character recognition of character information displayed in a reading item set in the image data of the document and generates text data;
A text data display unit for displaying the image data of the reading item and the text data for each reading item in a comparable manner, and making the text data editable by inputting characters to the text data; ,
A likelihood calculating unit for calculating the likelihood of character recognition for each reading item;
Based on the likelihood of each of the read item, Bei example and a display changing unit that changes the manner of a portion displaying the text data,
The display change unit changes the setting of the text data display item in which the likelihood for each reading item is equal to or less than a predetermined first threshold to a setting that does not display a character string from the text data display column, The image data and the text data of the reading item whose likelihood for each reading item is equal to or greater than a predetermined second threshold are not displayed, and the likelihood for each reading item is larger than the first threshold and the second A character recognition device that changes a mode of a portion where the text data of the read item is displayed according to the likelihood of each read item when the value is less than the threshold . It said text data display unit, together with the image data of the document, and extracts the image data of the read item is displayed on the text data and the parallel character recognition device according to claim 1.   The character recognition device according to claim 1, further comprising a reading item setting unit that sets the reading item from image data of the document. The character recognition unit generates the text data and calculates a likelihood for each character constituting the text data,
The character recognition according to any one of claims 1 to 3 , wherein the likelihood calculating unit performs a predetermined operation on the likelihood for each character to calculate the likelihood for each reading item. apparatus. The character recognition device according to claim 4 , wherein the likelihood calculation unit calculates the likelihood for each reading item by multiplying the likelihood for each character. The character recognition device according to claim 4 or 5 , wherein the display change unit displays the character in a different manner for each character according to the likelihood for each character. The character according to any one of claims 1 to 6 , wherein the display change unit displays a portion where the text data of the reading item is displayed in a different color according to the likelihood of each reading item. Recognition device. A character recognition method for reading character information from image data obtained by scanning a document as an image,
A character recognition step for performing character recognition of the character information displayed in the reading item set in the image data of the document, and generating text data, performed by the character recognition unit;
A state in which the text data display unit displays the image data of the read item and the text data so that the read data can be compared for each read item, and the text data can be edited by inputting characters to the text data. Text data display step to
A likelihood calculating step for calculating a likelihood of character recognition for each of the reading items, performed by a likelihood calculating unit;
Based on the calculated likelihood, performed by the display change section, Bei example and a display change step of changing the mode of a portion displaying the text data,
In the display change step, the reading item of the text data whose likelihood for each reading item is equal to or less than a predetermined first threshold is changed from the text data display column to a setting not to display a character string, The image data and the text data of the reading item whose likelihood for each reading item is equal to or greater than a predetermined second threshold are not displayed, and the likelihood for each reading item is larger than the first threshold and the second The character recognition method of changing the aspect of the location which displays the said text data of the said reading item according to the likelihood for every said reading item when it is less than a threshold value . A character recognition program that reads character information from image data obtained by scanning a document as an image,
A character recognition step of performing character recognition of the character information displayed in the reading item set in the image data of the document and generating text data;
A text data display step for displaying the image data of the reading item and the text data for each reading item in a comparable manner, and making the text data editable by inputting characters to the text data; ,
A likelihood calculating step for calculating the likelihood of character recognition for each reading item;
Based on the calculated likelihood, causing the electronic computer to execute a display change step of changing the mode of the location where the text data is displayed ,
In the display change step, the reading item of the text data whose likelihood for each reading item is equal to or less than a predetermined first threshold is changed from the text data display column to a setting not to display a character string, The image data and the text data of the reading item whose likelihood for each reading item is equal to or greater than a predetermined second threshold are not displayed, and the likelihood for each reading item is larger than the first threshold and the second The character recognition program which changes the aspect of the location which displays the said text data of the said reading item according to the likelihood for every said reading item when it is less than a threshold value .

Images