JPH03161886A - Method for correcting misreading of ocr - Google Patents

Abstract

PURPOSE:To easily search a misread character by storing plural read characters correspondingly to an image and respective character codes and displaying the image of characters correspondingly to each group of character codes belonging to the same sort. CONSTITUTION:Plural characters read out by an optical character reader (OCR) 1 are stored as an image, character codes corresponding to respective characters are also correspondingly stored, and at the time of displaying the image on a display device such as a CRT, the image of characters is displayed so as to be aligned in each group of character codes belonging to the same sort. Namely, characters recognized as the same character by the OCR 1 are displayed as a group independently of their correct reading/misreading. Since the image of characters is displayed on the display device 4 as the same character group, a misread character can easily be detected.

Description

DETAILED DESCRIPTION OF THE INVENTION (Industrial Application Field) The present invention is an OCR (optical character reading)
In particular, the present invention relates to a method for correcting misread characters and recognized characters read and recognized by OCR.

(Background of the Invention) When inputting character data such as letters, numbers, and symbols to a computer, the wrong characters may be input (misreading). Searching for and correcting these misread characters is a very troublesome task, and in the end, the only option is to visually compare the original data and the computer output to find and correct the characters.

In order to easily detect misreadings, the following measures are taken.

First, there is verification, which is inefficient because it requires three times as many people and three times as many machines to process the same data.

Even with a calculation check, Hatsushitotal can only tell which pages have misreadings, and even with methods such as total checks, check digit methods such as modulus ten and modulus eleven, range checks and master checks, there are misreadings in which parts of the material. I know if there is, but I don't know which character it is. Therefore, in the end, that part of the original data and that part of the computer output must be visually checked.

Furthermore, there are some areas where it is impossible to search for misreading using methods other than verify and hash total. In short, there are parts where you can't even look for misreadings.

Misreading also exists in the case of data input using OCR,
The problem is how to deal with misread characters.

In OCR, characters whose character data cannot be read correctly are divided into the following two types.

One is a rejected character, which is a character to which the arithmetic unit could not determine which character the image read by OCR corresponds to.

The second is a misread character, which is a character read incorrectly.
In other words, there is a discrepancy between the character specified by the original and the character recognized by OCR.

Among these characters, since the OCR side can tell which characters have been rejected, it is easy to correct them by extracting and showing them.

However, since the OCR recognizes that the misread characters are correctly read, the OC recognizes which characters are misread characters.
It cannot be determined on the R side. Therefore, it is necessary to find and correct the data using the method described above, but this is a cumbersome task similar to key input, and even if OCR is used, data input is extremely time-consuming.

Therefore, an object of the present invention is to provide a method for easily finding and correcting misread characters among characters read by OCR.

(Structure of the invention) In order to achieve the above object, the invention according to claim 1
A process of storing a plurality of characters read by the CR as images, storing them by applying them to corresponding character codes, and listing and displaying the corresponding images of the characters on a display device for each group of character codes of the same type. A method for correcting OCR misreading was constructed.

The invention according to claim 2 specifies the images listed and displayed on the display device, and inputs the correct character to input the OC.
The OCR misreading correction method according to claim 1, which corrects the misreading of R, has been constructed.

(Effects) Since the present invention has the above-described configuration, the following effects are achieved.

According to the first aspect of the invention, first, a plurality of characters read by OCR are stored as images, and corresponding character codes are applied and stored. Next, transfer this image to a CRT or LCD. When displayed on a display device such as a PDP, images of the characters are displayed in correspondence with each group of character codes of the same type. In other words, characters that the OCR recognizes as the same character are displayed as a group on the display device, regardless of whether the characters are misread or not.

Therefore, according to the invention set forth in claim 1, images are arranged as the same character group on the display device, so it is difficult for a person who corrects misreading while looking at the display device to search for different characters among the same character group. It's okay, so
This method has the advantage that misread characters can be detected very easily.

Because misread characters are easy to find and correct, even when characters are written on forms read by OCR, the formal writing style of "rOcR characters," which are different for each type of OCR, is forced. There is no need to do this, and the burden on the person filling out the form is reduced, so the degree of use and application of OCR can be greatly improved.

According to the second aspect of the present invention, the misreading can be corrected immediately by specifying the image bit with a cursor or the like on the display device and keying in the correct character, so that the correction work can be done very quickly and easily.

(Example) The illustrated embodiment will be described below.

FIG. 1 is a block diagram embodying an OCR misreading correcting method according to the present invention.

In FIG. 1, a plurality of characters read by OCR 1 are read by CPU. As in the past, the data is inputted to the control/arithmetic unit 2 of the computer including RAM and ROM, assigned a corresponding character code and stored in the RAM, and then subjected to various data processing.

Note that the explanation here assumes that the OCR 1 is a mere input device that does not include a calculation section, and that the control calculation section 2 performs character recognition. This is a data processing device that includes a calculation section, an output section, etc.

Furthermore, according to the misreading correction method according to this embodiment, OCR
The image read in step 1 is stored in memory 3 as image bits, which are assigned to corresponding character codes. There are ways to store images, such as storing them in bits or using the vector method, but here we will assume that they are stored in bits. This image bit is converted into a digital signal, but it is almost in the form of a handwritten document read by the OCR 1, or a form that is printed by a printer, printed, and stamped with a rubber stamp. The image bits converted into digital signals are transferred to the CRT4 through the control calculation unit 2.
For example, it is displayed with 16 bits per square millimeter.

FIG. 2 is an algorithm of the misreading correcting method of the present invention.

In step 1, the control calculation section 2 calls out a specific character group to which a certain character code has been applied from the memory 3, and lists and displays them on the CRT 4, which is a display device. That is,
Regardless of whether it is a misread or not, each image bit to which a certain character code is applied is displayed on the CRT 4.

FIG. 3 shows an example of an image bit group displayed on the CRT4. The squared characters O, 1, and 2 at the left end of the CRT screen are specific character types, and the handwritten image bits listed on the right side of each character type are processed by the control calculation unit 2 regardless of whether they are misread or not. This is a group of image bits that are recognized as corresponding to a character type and have their character codes applied. Here, it is easy to see that there is a misreading indicated by e in the image bit group of "1".

Although FIG. 3 shows a state in which the L screen is displayed, the method of list display can be selected to make it easier for the operator to check. Although it is possible to display one screen at a time, for example, by displaying one character at a time at regular intervals, or by displaying one line at a time and checking each time, you can more carefully and easily look for misreadings. can.

In step 2, the operator visually searches for misreading, and if so, points to the misreading character with the cursor and keys in the correct character. In the example of FIG. 3, move the cursor to position e and key in "7". The modification is now complete.

Further, if the misreading is found and corrected, the correction of the displayed character group is completed and the process waits for input in step 3. For example, enter return.

In step 4, if there is another character group, the process returns to step 1 to display the next character group; if not, all correction work is completed.

Furthermore, it is also possible to modify rejected characters.

In Figure 4, rR is displayed at the top of the leftmost specific character.
There is a letter J, and the one listed to the right is OCR.
1 is an image bit that cannot be read or is an object character. These rejected characters can also be modified by pointing them with a cursor and keying in, in the same way as the misread characters described above. In this example, from left to right: r3J r5J r6J r2
Enter J.

Note that the image bits shown in Figures 3 and 4 are
It is also possible to output it to the printer 5 shown in the figure and further check it again.

In this manner, according to the embodiment described above, it is possible to easily find and correct misreading errors or errors caused by OCR.

In addition, in practice, various checks can be performed on the input sheet data, especially regarding numbers.1 1 4. If the character data is read, the CPU
Check the calculation and check the correct character type on the CRT.
Instead of displaying it above, only the parts that have been checked are displayed as parts that may be misread. In this way, the number of characters displayed on the CRT is reduced, so the burden on the corrector in checking the screen can be reduced.

Although the embodiments of the present invention have been described above, it goes without saying that the present invention is not limited to the above embodiments, and can be modified as appropriate within the scope of the gist of the present invention.

[Brief explanation of the drawing]

Fig. 1 is a block diagram embodying an embodiment of the method for correcting OCR misreading according to the present invention, Fig. 2 is an algorithm showing each step of the same embodiment, and Fig. 3 is a diagram of the same embodiment displayed. FIG. 4 is a diagram showing a CRT screen including a display of a beam projector character. 1 2 l ・OCR, 2 ・Control calculation unit, 3 ・Memory, 4 ・CRT, misread character.

Claims (2)

[Claims] (1) A plurality of characters read by OCR are stored as images, and are stored by applying them to corresponding character codes, and for each group of character codes of the same type, the images of the characters are listed in association with each other on a display device. A method for correcting OCR misreading, the method comprising the step of displaying. (2) The OCR misreading correcting method according to claim 1, wherein the OCR misreading is corrected by specifying the image listed and displayed on the display device and keying in the correct character.