JP2004171326A - Using method of software - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce risk in system operation at a character recognition software change by enabling a change to a new version of software while ensuring the recognition precision of an old version of software at the character recognition software change. <P>SOLUTION: When an old version of character recognition software is changed for a new version of character recognition software, for a period of transition from the old version software to the new version software, a real system recognizes characters with both new and old versions of software. Information about recognition precision of both new and old versions of software is collected statistically, and both recognition precision levels are compared. If the precision of the new version is higher than the precision of the old version, the introduction of the new version software is established. If the recognition precision of the old version software is higher, on the other hand, a complete change to the new version software may be withheld to enable concurrent operation using advantages of both old and new versions of software. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a method of using software, and more particularly to a method of operating software when upgrading character recognition software in a system in which image information of a form is obtained by an image scanner and characters are recognized from image information by character recognition software.
[0002]
[Prior art]
When changing the software of the electronic information device from the old version to the new version, the new version software is installed in the memory area where the old version software is stored. Normally, the old version of software is overwritten with the new version of software, so that the old version of software does not remain in the memory, and subsequent information processing is executed by the new version of software.
[0003]
On the other hand, when the old software is changed to the new (including improved) software, the old software is left in memory to reduce the risk of transition to the new software. In some cases, a situation is created in which the software is stored in a memory, and if a problem is found in the new version when switching from the old version to the new version, the old version can be easily returned to by an administrator's command. For example, Japanese Patent Laying-Open No. 2000-36867 (Patent Document 1) discloses a method of upgrading a program in an exchange system. That is, the new program is stored without overwriting the old program, and the new program is operated on a trial basis. If a failure occurs as a result of the operation, the old program is used and the replacement with the old program is smoothly performed without returning the version upgrade procedure to the beginning.
[0004]
[Patent Document 1]
JP 2000-36867 A
[Problems to be solved by the invention]
The technique of the above-mentioned known example relates to version-up of a control program in an exchange system apparatus, and there is no suggestion that the technique of this known example can be directly applied to a system using an optical character recognition (OCR) technique. In addition, there is no description referring to issues or configurations specific to the system using the OCR.
[0006]
Generally, when an improved version of software is applied to a large-scale system including software using OCR technology (character recognition software), unlike business software or hardware control software, the operational risk associated with software switching is relatively high. May be higher.
[0007]
In other words, character recognition software can generally improve recognition accuracy by modifying the recognition algorithm or optimizing the recognition dictionary, but after these changes are made for the purpose of improving the software, the manufacturer side Evaluate the effect of improving accuracy using the evaluation data owned by the company. However, the evaluation data does not necessarily cover the data input in the operation of the target real system, and is created based on an evaluation model based on the past knowledge experience of the manufacturer. Therefore, an improvement effect is not always obtained for data other than the data to be improved focused on when the software is improved, and conversely, the recognition accuracy of the new version software may be lower than that of the old version. This is because there is a difference between the set targeted by the system and the set targeted by the evaluation model. In view of the fact that the system development cost is actually limited, it has to be admitted that the deviation due to the deviation has a considerable effect on the accuracy of character recognition.
[0008]
It is an object of the present invention to provide a method of using character recognition software that can reduce the risk of operating the system when changing the character recognition software.
Another object of the present invention is to determine whether to change to a new version of software by comparing the character recognition accuracy by operating both the old version and the new version of software in a real system when changing the software. It is to provide a method of using software.
SUMMARY OF THE INVENTION It is an object of the present invention to provide a method for operating and using software when changing from one software to another of the same kind.
[0009]
[Means for Solving the Problems]
The present invention relates to a system for recognizing characters of an image using character recognition software, in which a new version of software is stored in a memory, and an old version of software is stored in a memory in an executable state. When there is a recognition request, the same object is recognized by executing the old version of software, and the new version of software is also recognized. The information on recognition accuracy is collected from the recognition results of both software and stored in the storage device. Then, referring to information on recognition accuracy of each software, it is determined whether or not to shift to a new version of software.
As described above, an evaluation result suitable for the operation of the actual system is obtained, and as a result, the migration to the new version of software can be performed. Therefore, the system operation risk relating to software switching can be reduced.
In the above system, information on the processing frequency for each type of form and information on the recognition accuracy are collected and stored in a storage device as information on the recognition accuracy related to the new version and the old version of software. Uses information on the processing frequency and recognition accuracy for each form.
Also, depending on the recognition result by the new and old software, for example, during the period until the further new software is provided, the software is switched to the higher recognition accuracy software according to the recognition accuracy information stored in the storage device. Use while According to this method, the new version software can be used while maintaining the recognition accuracy of the old version software.
[0010]
A preferred system using the character recognition software is a center server (first server), a local server device (second server device) connected thereto via a network, and a client connected to the second server device. It is composed of devices. Normally, character recognition is performed by character recognition software stored in the client device. However, when the software is upgraded, this software is transferred to the second server device, and this software is replaced with the old version software (first software). Will be treated. On the other hand, a new version of character recognition software (second software: an improved version or new version of the first software) to be changed is installed in the client device.
In the client device, image data from a form or the like is obtained, and this image information is sent not only to the client device but also to the second server device. The client device performs recognition using the new version of character recognition software, and the second server device performs character recognition using the old version of character recognition software. Recognition results from both software are sent to the first server as statistical information and stored in the storage device. After elapse of a certain period determined at the time of software transfer, information on the recognition accuracy of the old and new versions of character recognition software stored in the storage device of the first server is output, and according to this information, switching to the new version software is performed. A decision is made as to whether or not to perform full operation using the new software. This determination may be performed by the center server device, or the determination itself may be performed by an introductory person.
[0011]
INDUSTRIAL APPLICABILITY The present invention can be broadly applied to software operation processing when changing from one software (first) to another software of the same type (second). That is, in a certain system, executing and processing first software for a certain processing purpose, and the same type as the first software while the first software remains in an executable state in the system. A step of finally installing a second software to be replaced with the first software, a step of executing and processing the second software, and a first processing when a processing request is generated in the system. A step of performing the processing by the software and the processing by the second software in parallel; a step of collecting information on an execution performance by both the software from a processing result by the first and second software; and an execution of the second software To determine whether the performance is good or the execution performance by the first software is good And a step of using the collected information.
Further, in the above-described operation processing, when it is determined based on the collection result that the first software is better than the second software, the processing by using the processing by the first software and the processing by the second software is performed in combination. Do. This allows you to take advantage of both software.
[0012]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a configuration example of a system that performs OCR processing using character recognition software. This system includes a center 107 and a sales base 101 connected via a network 106 such as a WAN. Generally, it may be assumed that a plurality of sales bases 101 exist in the system. A plurality of client devices 102 are connected to the sales base 101 via a LAN 104 and a sales base server 105 that relays data with the center. A scanner device 103 for converting a form into image data is connected to each client device 102.
The center 107 is provided with a center server 108 for centrally managing business data and statistical data transmitted from each sales base 101 via the network 106.
[0013]
FIG. 2 shows a configuration example of a client device set in the sales base 101. The client device 201 includes an I / O control unit 206 for controlling data input / output with the outside. The I / O control unit 206 includes a display 202 for prompting an operator, a keyboard 203 for inputting the operator's instruction, an image scanner 204 for converting a form into image data, and a LAN for performing data communication with the business base server 105. The board 205 is connected. In the memory 207, business software 208 for realizing business processing of the sales counter, control software 209 for supporting data exchange between each software and the I / O control unit 206, and a form of image data converted by the image scanner 204 are stored. The form identification software 210 for specifying the type and the character recognition software 211 for performing character recognition of each field of the form image data in which the form type has been found are stored. Here, the character recognition software 211 is a new version of character recognition software with improved recognition accuracy. An execution instruction of a program included in each software is executed by the CPU 212, and when data input / output is necessary, peripheral devices such as a display or LAN boards 202 to 205 are controlled via an I / O control unit 206. .
[0014]
FIG. 3 shows a configuration example of the server device 105 set in the sales base 101. The server device 301 includes an I / O control 305 for controlling data input / output with the outside. The I / O control unit 305 includes a display unit 302 for displaying the internal state of the server device 301, a LAN board 303 for performing data communication with the client device 102, and a modem 304 for performing communication with the center server. It is connected. The memory 306 includes server control software 307 for supporting data exchange between each software and the I / O control unit 305, character recognition software 308 for performing character recognition of each field of the form image data transmitted from the client device. Is stored. Here, the character recognition software 308 is an old version of character recognition software before the improvement. An execution instruction of a program included in each software is executed by the CPU 309, and when data input / output is necessary, peripheral devices such as the display 302 to the modem 304 are controlled via the I / O control unit 305.
[0015]
FIG. 4 shows a configuration example of the server device 108 set in the center 107. The server device 401 includes an I / O control unit 406 for controlling data input / output with the outside. In the I / O control unit 406, a display 402 for displaying the internal state of the server device 401 and data are stored. A hard disk device (storage device) 403 as a device for communication and a modem 405 for communicating with the sales office server 105 are connected.
[0016]
The hard disk device 403 stores a statistical information file 404 for editing and recording the result of the character recognition processing performed at each business base for each date and time / form.
FIG. 5 shows an outline of a storage table of the statistical information file 404 in the hard disk device. The tables shown in (5B) and (5C) in the figure are tables stored in the statistics file. The table shown in FIG. 5A is a table stored in the memory of the base server of each branch, and information such as a reading rate is stored for each form and for each version of software (program). The center servers 108 and 401 acquire relevant information from the branch table (5A), edit the information, create tables as shown in FIGS. 5B and 5C, and store the tables in the file 404. (5B) stores information such as the reading rate of the old version software for each form. (5C) stores information such as a reading rate by the new version of software for each form. Each is a tally in the same period.
[0017]
Returning to FIG. 4 again, the memory 407 includes server control software 408 for supporting data exchange between each software and the I / O control unit 406, recording of statistical information transmitted from the business base, and client of the business base. A statistical information recording determination module that processes a character recognition software use determination request (a processing request for determining whether to use the new version 211 or the old version 308 of character recognition software based on statistical information), which is an inquiry request from the device 102. 409 are stored. An execution instruction of a program included in each software is executed by the CPU 410, and when data input / output is necessary, peripheral devices such as the display 402 to the modem 405 are controlled via the I / O control unit 406.
[0018]
FIG. 6 is a flowchart illustrating an example of a recognition processing method in the system during the transition period of the character recognition software. Hereinafter, a specific processing flow will be described.
First, the image of the form is optically read by the image scanner 204 and digitized as image data (501). Next, the form type is identified by the form identification software 210 (502). The digitized image data is subjected to character recognition processing in parallel by both the client device 102 and the base server 105. First, in order to perform character recognition processing by the old version software 308 of the business base server 105, Then, the client device 102 requests the business base server 105 to perform a character recognition process (503). Image data to be recognized is also sent to the server 105. In parallel, the character recognition processing by the new version software 211 of the client device 102 is performed.
The business base server 105 performs a character recognition process using the old version software (505). Thereafter, the result of the character recognition processing is communicated to the client device 102 (506).
[0019]
After performing the character recognition processing (504) by the new software 211, the client device 102 does not recognize the result of the recognition by the new software and the result of the character recognition processing (505) by the old software 308 sent from the server 105. The comparison is made by paying attention to the reading rate (507). Here, the unread rate indicates the ratio of the total number of characters whose recognition target characters could not be correctly recognized by the character recognition software to the total number of characters. If the unread rate of the new version software 211 is low, the result of character recognition by the new version software is adopted as coded data for business processing (508). On the other hand, if the unread rate of the old version software 308 is low, the result of character recognition by the old version software 308 is used as coded data for business processing (509).
[0020]
In the client device 102 at the business base, business processing is performed by the business software 208 with the processing result having a low unread rate as input (510). Then, the adoption information of the form type / processing date / time / character recognition result (information indicating which software of the new version or the old version was adopted) is transmitted to the center server 108 (511). In this communication, the sales office server 105 is used as a relay server. The center server 401 processes the information transmitted from the client device 102 as statistical information and records it on the hard disk 403 (512). Thus, a series of processing ends.
[0021]
FIG. 7 illustrates a software operating method when the recognition accuracy of the new version of the character recognition software 211 is lower than that of the old version 308 after the elapse of the transition period of the character recognition software. In this example, in the period until the new improved software is provided, rather than reverting to the old software 308, both the new and old character recognition software are used in combination so that the recognition ability of the system is not reduced. It is. In other words, it is an operation method for software migration.
Hereinafter, a specific processing flow will be described. First, the image of the form is read by the image scanner 204 and digitized as image data (601). Next, the form type is identified by the form identification software 210 (602). At this point, since the form type and the processing date and time are determined, the center server 108 is inquired about which new or old character recognition software should be used under the conditions (603). This inquiry is made by the client device 102 to the center server 108, and the sales office server 105 is used as a relay server.
[0022]
The center server 108 generates identification information on the version of the character recognition software with high recognition accuracy based on the accumulated statistical information file 404, and determines whether the recognition accuracy of the new or old version of the character recognition software is high. (604). The result of the determination is transmitted to the client device 102 at the sales office (605).
On the client side of the sales office, character recognition processing is performed using character recognition software with high recognition accuracy based on the acquired determination result (606). If it is determined that the accuracy of the new version of the software 211 is high, a character recognition process is executed in the client device 102 (607). On the other hand, if the accuracy of the software of the old version 308 is determined to be high, the sales base server 105 is requested to perform character recognition processing (608).
In the latter case, the sales office server 301 performs character recognition processing on the form image data transmitted from the client device 102 by the old version of the character recognition software 308 (609). As a result, the sales base server 105 notifies the client device 102 of the character recognition result (610).
[0023]
The client device 102 at the business office determines the result of the character recognition processing by either the new software 211 or the old software 308 as coded data (611). Thereafter, business processing is performed by the business software 208 using the coded data as input (612). Thus, a series of business processes at the sales base is completed.
[0024]
According to this method, the recognition accuracy of the new version software can be calculated while maintaining the recognition accuracy of the old version software. As described above, an evaluation result suitable for the operation of the actual system is obtained, and as a result, the new version of the software can be switched. Therefore, the system operation risk related to the software switching can be reduced. In addition, during the transition period, it is possible to collect, as statistical information, the recognition accuracy of the old and new versions of the character recognition software for each date and time and for each form type as statistical information. It can also be used as a reference in such cases.
[0025]
Although the embodiment of the present invention has been described above, the present invention can be implemented with various modifications and applications other than the above-described example. The present invention can be applied not only when upgrading recognition software in a character recognition system, but also in other systems such as an information retrieval system. Even when applied to an information retrieval system, information on execution performance to be evaluated (in this case, information on retrieval accuracy or retrieval performance, retrieval speed, etc.) is statistically collected from the operation of new and old information retrieval software in a real system. It is possible to determine whether to adopt a new version of the software based on the analysis result of the information.
[0026]
【The invention's effect】
As described above, according to the present invention, when the software in the system is changed, both the old version and the new version of the software are actually operated in the actual system, and whether or not to change to the new version of the software with reference to the result is determined. Can be determined. In particular, when changing the character recognition software, the evaluation results based on the operation of the actual system, that is, the evaluation results of both the old and new software can be statistically obtained based on the character recognition accuracy characteristic of the character recognition software. Therefore, it is possible to reduce the risk of operating the system when introducing the new version software.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration of a system using an OCR technology according to an embodiment of the present invention.
FIG. 2 is a diagram showing a configuration example of a client device in a business base.
FIG. 3 is a diagram showing a configuration example of a server device at a business base.
FIG. 4 is a diagram showing a configuration example of a server device in a center.
FIG. 5 is a diagram schematically showing a storage table of a statistical information file in a hard disk device.
FIG. 6 is a flowchart showing an operation of a recognition process during a transition period of the character recognition software.
FIG. 7 is a flowchart showing the operation of the recognition process after the transition period of the character recognition software has elapsed.
[Explanation of symbols]
101 business base, 102, 201 client device 103 image scanner, 104 LAN,
105, 108, 301, 401 server device, 106 network,
107 center, 108 server device,
202, 302, 402 display,
203 keyboard, 204 image scanner,
205, 303 LAN board, 206, 305, 406 I / O control unit,
207, 306, 407 memory, 208 business software,
209 control software, 210 form identification software,
211 New version of character recognition software, 212, 309, 410 CPU
307 server control software, 308 old character recognition software,
403 hard disk drive, 404 statistical information file,
408 server control software, 409 statistical information record determination module.

Claims (8)

In a system for recognizing characters of an image using character recognition software, a new version of the software is stored in a memory, and an old version of the software is stored in a memory in an executable state. In this case, the same object is subjected to recognition processing by executing the old version of software and recognition processing is executed by executing the new version of software.Information regarding recognition accuracy is collected from the recognition results of both softwares and stored in the storage device. A method of using software, comprising determining whether or not to migrate to a new version of software by referring to information on recognition accuracy of the software. As information on the recognition accuracy related to the new and old versions of software, information on the processing frequency or recognition accuracy for each type of form is collected and stored in the storage device. 2. The method according to claim 1, wherein information on processing frequency or recognition accuracy is used. 2. The method according to claim 1, wherein the software is used while being switched to a version of software having a higher recognition accuracy according to the information on the recognition accuracy stored in the storage device. In a system including a first server at a center, a local second server connected to the first server via a network, and a client connected to the second server,
The client device includes an image scanner that optically reads an image of a form, a memory that stores a new version of software for recognizing characters of an image read by the image scanner, and a character that is executed by executing the new version software. A CPU for performing a recognition process,
The second server device has a memory for storing old-version software for recognizing characters of an image read by an image scanner of a client device, and a CPU for executing the old-version software to perform character recognition processing. And
The first server has a storage device for storing information on recognition accuracy obtained from a result of character recognition sent from the client device and the server device at the base, and a statistical device for storing information on recognition accuracy stored in the storage device. Having a CPU for processing
First processing means for recognizing an image read by the image scanner of the client device by executing the new version of software;
Second processing means for transmitting the same image read by the image scanner to a second server device and performing recognition processing by executing an old version of software;
Means for transmitting to the center server information relating to the recognition accuracy of characters from the processing results of the first processing means and the second processing means, for storing in the storage device of the first server;
Means for outputting information on recognition accuracy stored in the storage device when evaluating the operation status of the new software and the old software,
A character recognition processing system comprising: After the elapse of a certain period determined for the transition to the new software, the recognition accuracy of the new software and the recognition accuracy of the old software stored in the storage device are compared for each client device, and the recognition accuracy of the new software is better. 5. The system according to claim 4, wherein in such a case, the new version software is continuously used on the client device. Executing a first software for a certain processing purpose in a system for executing a certain process by executing the software; and executing the first software in an executable state and leaving the first software in the system. Installing in the memory second software which is of the same kind as the first software and which can eventually be replaced by the first software; and executing the second software for similar processing purposes Performing the processing by the first software and the processing by the second software in parallel when a request for a certain processing is generated in the system; Collecting information on the execution performance of both software from the processing results by the second software; To determine whether execution performance by execution performance is good or first software A good, it collected operational use of software and a step of using the information. 7. The method according to claim 6, further comprising, if it is determined that the execution performance of the second software is better than the execution performance of the first software, switching to the second software and using it thereafter. How to use and use the software. In claim 6, when it is determined that the execution performance of the first software is better than the execution performance of the second software, the processing by the first software and the processing by the second software are used together. A software operation processing method having a processing step.

Images