Image analysis – Editing – error checking – or correction – Including operator interaction
Reexamination Certificate
2000-09-28
2004-07-06
Mariam, Daniel (Department: 2621)
Image analysis
Editing, error checking, or correction
Including operator interaction
C382S175000
Reexamination Certificate
active
06760490
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to entry of data from form documents into a computer, and specifically to methods and systems for improving the accuracy and efficiency of manual key-in of such data.
BACKGROUND OF THE INVENTION
Despite progress in automation of data gathering functions, much large-scale data collection, such as census taking, is still carried out by manually filling in forms. The data are typically written by hand into designated fields on a preprinted form. Although optical character recognition (OCR) can generally be used to read at least a portion of the forms automatically by computer, there usually remain many fields that cannot be deciphered by OCR, at least not with a desired level of confidence. These fields must then be processed by a human operator, who manually keys the field contents into a computer. This manual key-in is typically the most costly part of large-scale data gathering. Manual key-in is also used in applications, such as check processing, for which OCR is not sufficiently accurate or is otherwise unsuitable.
In large-scale manual key-in, forms to be processed are generally scanned into a computer, and an image of the entire document or of each field that is to be keyed is displayed on the operator's computer screen. Manual key-in is typically about 95% accurate, which is far too low for most applications. The usual solution to this problem is to have all of the data keyed-in twice, by two different operators, which doubles the cost of processing. Visual checking of the keyed-in data against the original field images is no less time-consuming.
U.S. Pat. No. 5,455,875, whose disclosure is incorporated herein by reference, describes a system and method for quality control and correction of computer-generated OCR data by a human operator. The principles of this patent are embodied in the Intelligent Forms Processing (IFP)system produced by IBM Corporation, of Armonk, N.Y. The system can be configured to display to a human operator a full screen of images of individual characters from scanned documents, which were classified by OCR as being the same character. This type of image is referred to as a “carpet.” Errors in the OCR are manifested as character images that do not fit the displayed classification and stand out clearly against the correct images in the carpet. For example, if the OCR erroneously classifies a “P” as an “A,” the operator will see an image of a P in a screen full of A's. This type of discrepancy is very easy for the human operator to spot and mark on the screen. The image of the field that was read erroneously by the OCR is then displayed so that the operator (or another operator) can type in the correct character.
SUMMARY OF THE INVENTION
Preferred embodiments of the present invention provide improved methods and systems for visual checking and correction of character data that have been keyed in by a human operator. A computer key-in system displays images of filled-in form documents, or of fields in such documents, for viewing by an operator who keys in corresponding character codes, as is known in the art. The system then uses the character codes to segment the images, preferably so that each segment contains a single character and has an associated character code that has been keyed in by the operator. The system groups the image segments according to the associated character codes and displays the grouped segments together in a “carpet” of segments for viewing by a verifying operator. If the key-in operator were totally accurate, all of the segments would contain the same character (albeit in different forms). Practically, however, some of the segments will contain different characters, generally due to key-in errors by the operator. The verifying operator can recognize these deviant segments easily in the carpet display and marks them on the display for correction.
The present invention thus enables high accuracy to be achieved in manual key-in by improving the efficiency of quality verification. It obviates the need to key in all of the characters twice and thus saves substantially on the personnel costs of key-in. Preferred embodiments of the present invention may be used to facilitate manual key-in for applications in which OCR is unavailable or impractical, as well as for verifying fields that have been keyed-in manually due to OCR failure.
There is therefore provided, in accordance with a preferred embodiment of the present invention, a method for data key-in, including:
receiving images of documents containing characters;
receiving character codes entered by one or more key-in operators, the codes corresponding respectively to the characters in the documents;
identifying, for at least some of the character codes, respective areas of the images that contain the corresponding characters, such that each of the areas contains one of the characters; and
displaying together a group of the areas of the images responsive to the corresponding character codes, so as to facilitate verification of the codes by a verifying operator.
In a preferred embodiment, the documents include form documents having fields that are filled in with the characters, and receiving the images includes receiving an image of one of the fields containing one or more of the characters at a known location on one of the form documents, and identifying the respective areas includes finding the areas containing the characters in the field at the known location.
Preferably, receiving the character codes includes receiving an operator-generated string of the codes corresponding to a group of the characters in one of the documents, and identifying the respective areas includes finding the group of characters responsive to the string. Most preferably, finding the group of characters includes applying optical character recognition to the characters in the group so as to produce a machine-generated string of character codes, and matching the operator-generated string to the machine-generated string.
In a further preferred embodiment, the documents include form documents having fields that are filled in with the characters, and receiving the images includes receiving an image of one of the fields containing a group of the characters, and identifying the respective areas includes segmenting the image of the field responsive to the keyed-in character codes corresponding to the characters in the field. Preferably, receiving the character codes includes receiving a sequence of codes corresponding respectively to the characters in the group, and segmenting the image includes dividing the image of the field into a number of segments equal to the number of codes in the sequence. Alternatively or additionally, segmenting the image includes dividing the image of the field into the areas such that the character in each of the areas has an appearance that accords with the corresponding one of the character codes.
Preferably, displaying the group of the areas includes selecting the group for display such that all of the areas in the group are identified with a selected one of the character codes. Further preferably, the method includes receiving an input from the verifying operator indicating that the character in one of the displayed areas does not accord with the selected character code and correcting the character code corresponding to the indicated character.
There is also provided, in accordance with a preferred embodiment of the present invention, a method for data key-in, including:
receiving images of documents containing characters;
receiving character codes entered by one or more key-in operators, the codes corresponding respectively to the characters in the documents;
identifying, for a selected sequence of the character codes, respective areas of the images that contain a corresponding sequence of the characters, such that each of the areas contains an instance of the corresponding sequence; and
displaying together a group of the areas of the images that contain the corresponding sequence of the characters, so as
Darby & Darby
Mariam Daniel
LandOfFree
Efficient checking of key-in data entry does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Efficient checking of key-in data entry, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Efficient checking of key-in data entry will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3195333