JPS62274381A - Automatic deciding system for misread of ocr - Google Patents

Abstract

PURPOSE:To estimate the item of a misread by sending an account number recognized by an OCR, to a central processing unit, retrieving the corresponding full name from a file and collating it with a full name recognized by the OCR, and comparing whether the number of dissidence characters and the reference value are large or small, in case of dissidence. CONSTITUTION:A slip 1 is read by an OCR 2, and recognized information and image information are stored in the file 4 of a terminal control equipment 3. By a correcting request from a terminal 5, the recognized information is fetched, and with regard to an entry item, a misreading is decided automatically on the equipment 3. First of all, an account number is sent to a central processing unit 6, and the corresponding full name by KANA (Japanese syllabary) name is fetched from a file 7 and compared with a full name recognized by the OCR. In case of coincidence, it is displayed to the terminal 5 that the correction is not required, and if the number of dissidence characters is <=2, the full name is emphasized, and if said number is >=3, the account number is emphasized, and the account number, the full name by KANA, and the image information of the slip are displayed in parallel. An operator executes a correcting input, sends the corrected information to the central processor 6 and executes a transaction processing. According to such constitution, the wrong entry of a customer can also be detected, and the man-hour for checking a misreading is remarkably decreased.

Description

[Detailed Description of the Invention] 3. Detailed Description of the Invention [Field of Industrial Application] The present invention relates to the work of inputting slips using an OCR device.
In particular, the present invention relates to a method for automatically determining OCR misreading, which is suitable for reducing bank transaction input.

[Conventional technology]

Conventional OCR misreading judgment methods include mutual checking between items (equivalence check, size check, total value check, etc.) as described in Information Processing Exam Course Hardware Knowledge (Ohmsha), page 165. ), it was possible to determine whether there was a misreading. However, even for items that are mutually related, such as an account number and a name, no consideration was given to automatic determination of misreading when a relationship cannot be derived using four arithmetic operations.

[Problem that the invention seeks to solve]

The above-mentioned conventional technology does not take into account the fact that when inputting a slip using OCR, it automatically determines whether the account number and name are misread.
There was a problem in that it was necessary to visually check for misreading after inputting the information.The purpose of the present invention is to check the relationship between an account number (or a number similar to an account number to identify an individual) and a name. The aim is to automate checking for misreading of both items and reduce the input burden on the operator. In other words, if it is determined that there are no misreadings, there is no need to manually check for misreadings, and even if it is determined that there are misreadings, the misreadings are checked for either account number or name so that the work to correct the misreading items is minimized. The purpose is to infer and guide whether there was a

[Means for solving problems]

The above purpose is to transmit the account number recognized by OCR to the central processing unit, search the customer file on the central processing unit for the name corresponding to that account number, and search from the name recognized by OCR and the customer file. This is achieved by comparing and collating the name and, if there is a mismatch, by further comparing the magnitude relationship between the number of mismatched characters and a reference value.

[Effect]

Assuming that one account number and name are correctly registered in the customer file on the central processing unit, the name searched from the customer file using the account number recognized by OCR as a key matches the name recognized by OCR. Then, since the correspondence between the account number and name recognized by OCR matches the one in the customer file, it can be determined that both items have been correctly recognized.

Furthermore, if there is a mismatch, it means that the same correspondence relationship does not exist in the customer file, so it can be determined that there has been a misreading.

At this time, if the account number is correctly recognized and the name is misread, the number of mismatched characters is often small (about 2 characters per character), although it depends on the recognition rate of the OCR device.

Conversely, if there is a misreading of the account number, the customer file will be searched for a completely different person's name, so there is a high probability that most of the characters will not match.

Therefore, a standard value is set for the number of unmatched characters (usually around 3 or 4 characters, depending on the recognition rate of the OCR device), and if the number of unmatched characters is less than the standard value, it is assumed that there was a misreading of the name. If it is greater than the standard value, it can be inferred that the account number was misread.

〔Example〕

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

The slip 1 is read by the ○CR device 2, and the ocR recognition data and the image data of the slip are stored in the slip file 4 connected to the terminal control device 3.

In response to a request from the correction input terminal 5, OCR recognition data is extracted from the slip file 4, and automatic misreading determination is performed on the terminal control device 3 regarding the entries in the slip.

First, the OCR-recognized account number is transmitted to the central processing unit 6, and the central processing unit 6 extracts the corresponding kana name from the customer file 7 and transmits it to the terminal control device 3. The terminal control device 3 compares the received kana name with the kana name recognized by OCR and automatically determines whether the name is misread.

If the two match, go to the correction input terminal 5.

A message indicating that no correction is necessary is displayed, and if there is a discrepancy, the account number and kana name are displayed in parallel with the image data of the slip to guide correction input. At this time, if the number of mismatched characters is 2 or less, the name is highlighted, and if the number of mismatched characters is 3 or more, the account number is highlighted.

Accordingly, when the operator makes a correction input, the operator sends the slip input data (corrected data) to the central processing unit 6 to process the transaction.

In the conventional method, all account numbers and kana names had to be manually checked for misreading and corrected, but according to this embodiment, misreading can be automatically determined, greatly reducing the need for manual checking. It has the effect of reducing

In addition, in the conventional method, if a customer incorrectly enters the account number and name in kana, if OCR correctly recognizes the name,
Although it cannot be checked manually, according to this first embodiment, it is also possible to detect incorrect entries by the customer.

〔Effect of the invention〕

According to the present invention, in the business of inputting account numbers and names using OCR, it is possible to use the system to instruct the presence or absence of misreading and to infer the misread items, thereby significantly reducing the need for manual checks for misreading. effective.

[Brief explanation of the drawing]

The figure is a system configuration diagram of an embodiment of the present invention. 1... Slip, 2... OCR, 3... Terminal control device, 4... Slip file, 5... Correction input terminal, 6...
...Central processing unit, 7...Customer file.

Claims (1)

[Claims] 1. When inputting the contents of a slip using an OCR device, from the account number recognized by the OCR or a number to identify an individual similar to the account number, the corresponding name from the customer file on the central processing unit. index and OCR
If the comparison results match, it is determined that both the account number and the name have been correctly recognized, and if the comparison results do not match, either the account number or the name is selected. determining that there was a misreading of the
In the case of a mismatch, a standard value is set, and if the number of mismatched characters is greater than the standard value, it is assumed that the account number was misread, and if it is less than the standard value, it is assumed that the name was misread. Automatic OCR misreading detection method.