JP2022090443A - OCR processing device, OCR processing method, and program - Google Patents

Abstract

To provide an OCR processing device which improves efficiency of confirmation or modification of a result of OCR processing.SOLUTION: An OCR processing device includes: an OCR processing unit which specifies character strings input to input items; a grouping unit which classifies the plurality of character strings specified by the OCR processing unit into a plurality of groups; and a display unit which displays the plurality of character strings classified by the grouping unit while classifying for each of the groups. The OCR processing device preferably further includes a character string decision unit which designates decided character strings for the plurality of character strings classified by the grouping unit according to an operation of a user.SELECTED DRAWING: Figure 2

Description

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

For example, in Patent Document 1, an instruction is input in which characters having a score of a predetermined score or less are grouped by characters having the same character code, and the character code is corrected for one of the grouped character groups. If this is the case, all the character codes in the group to which this one character belongs are changed according to this instruction. Image processing method Characters with a predetermined score or less are grouped by characters having the same character code. When an instruction to correct the character code is input for one character in the grouped character group, all the character codes in the group to which this one character belongs are changed according to this instruction (the score is a predetermined score). The following characters are grouped by characters with the same character code, and when an instruction to correct the character code is input for one of the grouped character groups, this one character belongs. Change all character codes in the group according to this instruction. Characters with the same score or less are grouped by characters with the same character code, and the character code is assigned to one of the grouped character groups. When an instruction to correct is input, all the character codes in the group to which this one character belongs are changed according to this instruction. However, when an instruction to correct the character code is input for one character in the grouped character group, all the character codes in the group to which this one character belongs are changed according to this instruction. ing.

Further, in Patent Document 2, images having the same character string as the OCR recognition result are classified, image areas corresponding to the character strings are overlapped and combined, and the character string of the classification axis and the character string of the composite image are combined. A document processing device that performs comparisons and collectively corrects all the combined images. Classification of images with the same character string in the OCR recognition result is performed, and the areas of the images corresponding to the character strings are superimposed and combined. It is disclosed that the character string of the classification axis and the character string of the composite image are compared, and the correction processing is performed collectively for all the composited images.

JP 2006-178543 JP 2012-146147

It is an object of the present invention to provide an OCR processing apparatus for efficiently confirming or correcting OCR processing results.

The OCR processing apparatus according to the present invention includes an OCR processing unit that specifies a character string input to an input item, a grouping unit that classifies a plurality of character strings specified by the OCR processing unit into a plurality of groups, and a grouping unit. It has a display unit for displaying a plurality of character strings classified by the grouping unit separately for each group.

Preferably, the grouping unit groups the character strings specified by the OCR processing unit with respect to the same input item based on the similarity of the character strings.

Preferably, it further has a character string fixing unit for designating a fixed character string for a plurality of character strings classified by the grouping unit according to a user operation.

Preferably, the display unit changes the display mode of a plurality of character strings classified into the same group by the grouping unit according to the number of the same character strings.

Preferably, the display unit displays the largest number of the same character strings emphasized more than other character strings for a plurality of character strings classified into the same group by the grouping unit.

Preferably, the grouping unit does not match any of the registered character strings by the dictionary grouping unit that classifies the character strings that match the pre-registered registered character strings into groups and the dictionary grouping unit. The determined character string includes a similarity grouping unit that classifies the determined character strings based on their similarity to each other, and the display unit is classified by the character string classified by the dictionary grouping unit or the similarity grouping unit. The displayed character strings are displayed separately for each group.

Preferably, the grouping unit includes a blank grouping unit that groups input items that are determined to be blank, and an input item that is not determined to be blank by the blank grouping unit. , The dictionary grouping unit that classifies the character strings that match the pre-registered registered character strings into groups, and the character strings that are determined by the dictionary grouping unit to be inconsistent with any of the registered character strings. The display unit includes a similarity grouping unit that classifies based on the degree of similarity, and the display unit displays a character string classified by the dictionary grouping unit or a character string classified by the similarity grouping unit for each group. Display separately.

Preferably, the group moving unit that accepts the operation of moving the character string between the groups classified by the grouping unit and the character string moved by the group moving unit are determined with respect to the destination group. It further has a movement confirmation unit to which a confirmation character string determined by the unit is applied.

Preferably, it further has a group display unit that displays a list of the groups classified by the grouping unit and adds a display indicating whether or not the confirmation work for each group is completed.

Further, in the OCR processing method according to the present invention, an OCR processing step in which a computer specifies a character string input in an input item and a plurality of character strings specified by the computer in the OCR processing step are grouped in a plurality of groups. It has a grouping step for classifying into groups, and a display step in which a computer displays a plurality of character strings classified by the grouping step separately for each group.

Further, the program according to the present invention includes an OCR processing step for specifying a character string input to an input item, a grouping step for classifying a plurality of character strings specified by the OCR processing step into a plurality of groups, and a grouping step. A computer is made to execute a display step of displaying a plurality of character strings classified by the grouping step separately for each group.

Confirmation or correction of OCR processing results can be made more efficient.

It is a figure which illustrates the whole structure of the OCR processing system 1. It is a figure explaining the confirmation correction work in the OCR processing system 1. It is a figure which illustrates the hardware composition of the OCR processing apparatus 2. It is a figure which illustrates the functional structure of the OCR processing apparatus 2. It is a flowchart explaining the OCR processing (S10) by the OCR processing apparatus 2. It is a flowchart explaining the character string movement processing (S20) by the OCR processing apparatus 2. It is a figure which illustrates the display screen for confirming and correcting an OCR processing result. This is an example of a display screen when the OCR processing result is moved between groups. This is an example of a display screen when the OCR processing result is moved to a confirmed group. (A) is a diagram for explaining the grouping process by the dictionary grouping unit 314 in more detail, and (B) is a diagram for explaining the grouping process by the similar grouping unit 316 in more detail.

Embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a diagram illustrating the overall configuration of the OCR processing system 1.
As illustrated in FIG. 1, the OCR processing system 1 includes an OCR processing device 2 and a scanner device 4, and is connected to each other via a network 7. The plurality of scanner devices 4A and scanner devices 4B are collectively referred to as a scanner device 4. The OCR processing device 2 may be directly connected to the scanner device 4 by wire via a USB cable or the like.
The OCR processing device 2 is a computer terminal, performs OCR processing on the image data received from the scanner device 4, and causes the user to confirm or correct the OCR processing result. The OCR processing device 2 of this example performs OCR processing on the image data of the form received from the scanner device 4, specifies a character string for each input item of the form, and causes the user to perform confirmation work and correction work.

The scanner device 4 is an optical reading device, and transmits image data acquired from a document to the OCR processing device 2. In this example, the scanner device 4 shall scan the form.

FIG. 2 is a diagram illustrating confirmation and correction work in the OCR processing system 1.
As illustrated in FIG. 2A, there is a method of confirming the OCR processing result for each scanned form (comparative example). In this case, it is easy to check the consistency of the input items of the entire form.
Further, as illustrated in FIG. 2B, there is a method of collecting OCR processing results for each input item (“company name” in this example) and checking them collectively (comparative example). In this case, since the same input items can be confirmed collectively, it is easy to notice an error in the OCR processing result.
However, in each of the above comparative examples, it is necessary to confirm all the OCR processing results, and the work efficiency is poor.

Therefore, in the OCR processing system 1 of this example, as illustrated in FIG. 2C, the OCR processing result (character string) for each input item is classified into a plurality of groups, and the OCR processing result is classified for each group. List and confirm (character string). Classification into groups is based on the degree of similarity (match and similarity) of OCR processing results. Therefore, a plurality of character strings (character strings similar to each other) in the group can be confirmed at once, and the time required for the confirmation work can be shortened.
Further, in the OCR processing system 1 of this example, as illustrated in FIG. 2C, when any character string (OCR processing result) in the group is selected as the correct answer, the OCR processing result in the group is selected. It is corrected all at once to the character string. As a result, the time required for the correction work can be shortened. The "character string" in the present application is a concept including a case where there is only one character and a case where there are no characters (not entered).

FIG. 3 is a diagram illustrating a hardware configuration of the OCR processing device 2.
As illustrated in FIG. 3, the OCR processing device 2 includes a CPU 200, a memory 202, an HDD 204, a network interface 206 (network IF206), a display device 208, and an input device 210, and these configurations are via the bus 212. Are connected to each other.
The CPU 200 is, for example, a central arithmetic unit.
The memory 202 is, for example, a volatile memory and functions as a main storage device.
The HDD 204 is, for example, a hard disk drive device, and stores a computer program (for example, the OCR processing program 3 in FIG. 4) and other data files as a non-volatile recording device.
The network IF 206 is an interface for communicating by wire or wirelessly, and realizes communication with, for example, the scanner device 4.
The display device 208 is, for example, a liquid crystal display.
The input device 210 is, for example, a keyboard and a mouse.

FIG. 4 is a diagram illustrating the functional configuration of the OCR processing device 2.
As illustrated in FIG. 4, the OCR processing program 3 is installed in the OCR processing device 2 of this example.
The OCR processing program 3 includes an OCR processing unit 300, a grouping unit 310, a character string display unit 320, a group display unit 330, a character string confirmation unit 340, a group movement unit 350, and a movement confirmation unit 360. Further, the grouping unit 310 includes a blank grouping unit 312, a dictionary grouping unit 314, and a similar grouping unit 316.
A part or all of the OCR processing program 3 may be realized by hardware such as ASIC, or may be realized by borrowing a part of the function of the OS (Operating System).

In the OCR processing program 3, the OCR processing unit 300 performs OCR processing on the image data read by the scanner device 4, and specifies a character string included in the image data. The OCR processing unit 300 of this example performs OCR processing on the image data of the form according to the OCR definition information of the form, and specifies the character string of each input item of the form.

The grouping unit 310 classifies the plurality of character strings specified by the OCR processing unit 300 into a plurality of groups. More specifically, the grouping unit 310 classifies a plurality of character strings specified by the OCR processing unit 300 into a plurality of groups based on the similarity of the character strings with respect to the same input item. The grouping unit 310 of this example groups character strings by different logics in a multi-stage grouping process. That is, the grouping unit 310 of this example includes a blank grouping unit 312, a dictionary grouping unit 314, and a similar grouping unit 316.

The blank grouping unit 312 groups blank items for the same input item. Here, what is determined to be blank is, for example, in addition to the input item for which the character string could not be specified by the OCR processing unit 300, the number of characters in the character string specified by the OCR processing unit 300 is less than the standard. Including the input items that were there.

The dictionary grouping unit 314 classifies the character strings that match the pre-registered registered character strings into groups for the input items that are not determined to be blank by the blank grouping unit 312. For example, the dictionary grouping unit 314 classifies a character string that matches a registered character string registered in advance in the dictionary into a group corresponding to each registered character string.

The similarity grouping unit 316 classifies the character strings determined to be inconsistent with any of the registered character strings by the dictionary grouping unit 314 based on the degree of similarity with each other. For example, the similar grouping unit 316 calculates the Levenshtein distance for a character string determined by the dictionary grouping unit 314 to be inconsistent with any of the registered character strings, and a plurality of Levenshtein distances within a certain distance. Classify strings into the same group. The similar grouping unit 316 of this example classifies a plurality of character strings into the same group when there are a predetermined number or more (for example, 30% or more) of character strings whose Levenshtein distance is within a certain distance. ..

The character string display unit 320 is an example of the display unit according to the present invention, and displays a plurality of character strings classified by the grouping unit 310 separately for each group. As illustrated in FIG. 7, the character string display unit 320 of this example displays a list of character strings classified into a group selected by the user in the character string list display area 920. An input item image cut out from the form is displayed in the vicinity of the character string (OCR processing result) displayed in the list.
Further, the character string display unit 320 changes the display mode of a plurality of character strings classified into the same group by the grouping unit 310 according to the number of the same character strings. As illustrated in FIG. 2C, the character string display unit 320 of this example encloses the largest number of the same character strings in a blue frame for a plurality of character strings classified into the same group by the grouping unit 310, and others. Display with emphasis over the character string of.

The group display unit 330 displays a list of groups classified by the grouping unit 310, and attaches a display indicating whether or not the user's confirmation work for each group is completed. As illustrated in FIG. 7, the group display unit 330 of this example displays a list of the group names classified by the grouping unit 310 in the group list display area 900, associates them with the group names, and confirms each group. Place a confirmed icon indicating that the confirmation work has been completed, or an unconfirmed icon indicating that the confirmation work for each group has not been completed.

The character string confirmation unit 340 collectively designates the confirmed character strings for the plurality of character strings classified by the grouping unit 310 according to the user's operation. In the character string confirmation unit 340 of this example, when the user clicks any character string (OCR processing result) in the character string list display area 920 of FIG. 7, the clicked character string is displayed on all the character strings belonging to this group. Applies to strings. As a result, the correction work for the group is completed in a batch.

The group moving unit 350 accepts an operation of moving a character string (OCR processing result) between groups classified by the grouping unit 310. As illustrated in FIG. 8, the group moving unit 350 of this example selects a character string (OCR processing result) to be moved in the character string list display area 920, and uses the group name displayed in the group list display area 900 as the group name. When dragged, the selected character string is moved to the group corresponding to the group name of the drag destination.

The movement confirmation unit 360 applies the confirmation character string determined by the character string confirmation unit 340 with respect to the destination group to the character string moved by the group movement unit 350. As illustrated in FIG. 9, the movement confirmation unit 360 of this example drags and drops the character string of the character string list display area 920 to the group with the confirmed icon, and the character string confirmed in this group is confirmed. (Value) is applied to the dragged character string (OCR processing result).

FIG. 5 is a flowchart illustrating OCR processing (S10) by the OCR processing device 2. In this example, it is assumed that a plurality of forms are scanned by the scanner device 4 and these image data are transmitted to the OCR processing device 2.
As illustrated in FIG. 5, in step 100 (S100), the OCR processing unit 300 (FIG. 4) of the OCR processing device 2 OCRs the plurality of form image data received from the scanner device 4 according to the OCR definition information. Perform processing and specify the character string of each input item.

In step 105 (S105), the blank grouping unit 312 of the grouping unit 310 groups the input items to be processed that are not filled in.
In step 110 (S110), the dictionary grouping unit 314 sets a character string that matches the registered character string registered in advance in the dictionary with respect to the unclassified character string of the input item to be processed, in a group corresponding to each registered character string. Classify into. Note that S110 targets only those that are not grouped in S105.
In step 115 (S115), the similarity grouping unit 316 calculates the Levenshtein distance with respect to the unclassified character string of the input item to be processed, and selects a plurality of character strings in which the calculated Levenshtein distance is within a certain distance. , Classify into the same group. Note that S115 targets only those that are not grouped in S105 and S110. Further, the character strings that are not classified by the similar grouping unit 316 are finally classified into the unclassified group.

In step 120 (S120), the grouping unit 310 determines whether or not the grouping process has been completed for all the input items, and if there is an input item that has not been completed, processes the next input item. It returns to the process of S105 as a target, and when all the input items are completed, it shifts to the process of S125.

In step 125 (S125), as illustrated in FIG. 7, the group display unit 330 displays a list of group names of the groups classified by the grouping unit 310 in the group list display area 900, and allows the user to perform a selection operation. prompt.
As illustrated in FIG. 7, when the user selects any group name in the group list display area 900, the character string display unit 320 includes a character string (OCR processing result) classified into the selected group and a form. The character string image cut out from the image is displayed in a list in the character string list display area 920.

In step 130 (S130), the character string display unit 320 highlights the most character strings (OCR processing result) among the character strings displayed in the character string list display area 920 as the most character strings.

In step 135 (S135), the character string determination unit 340, as illustrated in FIG. 7, when any of the displayed character strings (OCR processing result) is clicked in the character string list display area 920, this Change all the strings classified in the group to the clicked strings and confirm.
In step 140 (S140), when the character string confirmation unit 340 performs the confirmation processing of the character string (OCR processing result), the group display unit 330 does not display the icon of this group in the group list display area 900 of FIG. Change from the confirmed icon to the confirmed icon.

In step 145 (S145), the OCR processing program 3 returns to the processing of S125 when an undetermined group exists, completes the confirmation work of the OCR processing result when all the groups are confirmed, and performs the OCR processing. (S10) is terminated.

FIG. 6 is a flowchart illustrating the character string movement process (S20) by the OCR processing device 2. The character string movement process (S20) is performed in parallel with the processes of S125 to S140 in FIG.
In step 200 (S200), as illustrated in FIG. 8, one of the character strings displayed in the character string list display area 920 is the group name of another group in the group list display area 900. It waits until it is dragged and dropped to (S200: No), and when any character string is dragged and dropped to the group name of another group, the process proceeds to S205 (S200: Yes).

In step 205 (S205), the group moving unit 350 moves the dragged and dropped character string to the group of the dropped group name, as illustrated in FIG. That is, the classification by the grouping unit 310 is modified by the user.

In step 210 (S210), in the movement confirmation unit 360, whether the movement destination of the character string moved by the group movement unit 350 is a group in which the character string is confirmed by the character string confirmation unit 340 or an undetermined group. If it is a confirmed group, it shifts to the process of S215, and if it is an undecided group, it shifts to the process of S220.
In step 215 (S215), when the movement destination of the character string is the confirmed group, the movement confirmation unit 360 is determined by the movement destination group with respect to the moved character string as illustrated in FIG. The string is automatically applied.
In step 220 (S220), when the movement destination of the character string is an undetermined group, the movement confirmation unit 360 does not determine the moved character string as illustrated in FIG. 8, and the character string list display area as it is. Display at 920.

FIG. 10A is a diagram for explaining the grouping process by the dictionary grouping unit 314 in more detail, and FIG. 10B is a diagram for explaining the grouping process by the similar grouping unit 316 in more detail.
As illustrated in FIG. 10A, the OCR processing result generally has a structure in which each character has a plurality of candidates and each has a reliability (OCR processing result in the figure). See 1st to 3rd place). Therefore, as shown in FIG. 10A, the dictionary grouping unit 314 compares the OCR processing result (combination of all candidates) with the words registered in the dictionary, and if there is a corresponding word, the target word is used. Group as the same value as a word. After comparing with all the words, if more than one is applicable, the rank of each letter is given as a score, and the one with the lowest score is adopted.

As illustrated in FIG. 10B, the similar grouping unit 316 measures the Levenshtein distance for all combinations of OCR processing results (character strings), and sets the values below a certain level as a group. deal. The Levenshtein distance is a distance indicating how different the two character strings are, and the smaller the value, the more similar they are. However, since it is grouped in all streets (order), it may belong to multiple groups. When belonging to multiple groups, it shall belong to the group with the shortest distance. Ultimately, even one that belongs to a group is the result of grouping. In the example of FIG. 10B, the one having the minimum Levenshtein distance of 1 is adopted (cell with a gray background). When a plurality of minimum values appear, they are narrowed down to either one in the order of arrangement (ascending order of No. 1). Finally, those with a large number of occurrences of the minimum value 1 (No. 1 and No. 4 surrounded by a thick frame) form a group.

As described above, according to the OCR processing system 1 of the present embodiment, the OCR processing results can be grouped based on the degree of similarity, and the confirmation work and the correction work can be performed in group units. Therefore, the confirmation work and the correction work can be performed. The time and effort required for this can be reduced.

1 ... OCR processing system 2 ... OCR processing device 3 ... OCR processing program 300 ... OCR processing unit 310 ... Grouping unit 320 ... Character string display unit 330 ... Group display unit 340 ... Character string confirmation unit 350 ... Group movement unit 360 ... Movement Confirmation part

Claims (11)

OCR processing unit that specifies the character string entered in the input item,
A grouping unit that classifies a plurality of character strings specified by the OCR processing unit into a plurality of groups, and a grouping unit.
An OCR processing apparatus having a display unit for displaying a plurality of character strings classified by the grouping unit separately for each group. The OCR processing apparatus according to claim 1, wherein the grouping unit groups character strings specified by the OCR processing unit with respect to the same input item based on the similarity of the character strings. The OCR processing apparatus according to claim 2, further comprising a character string confirmation unit that specifies a confirmed character string for a plurality of character strings classified by the grouping unit according to a user operation. The OCR processing device according to claim 3, wherein the display unit changes the display mode of a plurality of character strings classified into the same group by the grouping unit according to the number of the same character strings. The OCR processing apparatus according to claim 4, wherein the display unit displays the largest number of the same character strings in a plurality of character strings classified into the same group by the grouping unit with emphasis over other character strings. The grouping unit
A dictionary grouping unit that classifies character strings that match pre-registered registered character strings into groups,
The dictionary grouping unit includes a similar grouping unit that classifies character strings that are determined to be inconsistent with any of the registered character strings based on their similarity to each other.
The OCR processing apparatus according to claim 5, wherein the display unit displays a character string classified by the dictionary grouping unit or a character string classified by the similar grouping unit separately for each group. The grouping unit
An unfilled grouping unit that groups input items that are determined to be blank, and
A dictionary grouping unit that classifies character strings that match pre-registered registered character strings into groups for input items that are not determined to be blank by the blank grouping unit.
The dictionary grouping unit includes a similar grouping unit that classifies character strings that are determined to be inconsistent with any of the registered character strings based on their similarity to each other.
The OCR processing apparatus according to claim 5, wherein the display unit displays a character string classified by the dictionary grouping unit or a character string classified by the similar grouping unit separately for each group. A group moving unit that accepts an operation for moving a character string between groups classified by the grouping unit, and a group moving unit.
The OCR processing apparatus according to claim 3, further comprising a movement confirmation unit that applies a confirmation character string confirmed by the character string confirmation unit to the movement destination group for the character string moved by the group movement unit. The OCR processing apparatus according to claim 3, further comprising a group display unit that displays a list of groups classified by the grouping unit and indicates whether or not the confirmation work for each group has been completed. An OCR processing step in which the computer identifies the character string entered in the input item,
A grouping step in which the computer classifies the plurality of character strings identified by the OCR processing step into a plurality of groups,
An OCR processing method in which a computer has a display step for displaying a plurality of character strings classified by the grouping step separately for each group. An OCR processing step that identifies the character string entered in the input item,
A grouping step for classifying a plurality of character strings specified by the OCR processing step into a plurality of groups, and a grouping step.
A program that causes a computer to execute a display step that displays a plurality of character strings classified by the grouping step separately for each group.

Images