Information processing device, method, storage medium and computer program product

Abstract

An information processing device, a storage medium and an information processing method, characterized in that they are provided with a processor, which executes a first acquisition process of acquiring a first recognition result and a first recognition accuracy for object data from a first identifier, and executes a second acquisition process of acquiring a second recognition accuracy for the first recognition result for the object data from a second identifier, and controls which of the first process to be executed on the first recognition result and the second process requiring more human workload than the first process based on the first recognition accuracy and the second recognition accuracy.

Description

Technical Field

[0001] The present invention relates to an information processing device, a storage medium and an information processing method. Background Art

[0002] A data input system is provided for digitizing character strings written by hand or printed on paper forms. The data input system digitizes character strings on forms by a method in which a person reads the form and inputs the characters by keystrokes, by a method in which a character recognizer using optical character recognition (OCR) technology performs character recognition processing on an image on the form, or by a combination of these methods.

[0003] The character recognizer generally has a function of outputting a degree of reliability indicating the processing result of the recognition processing performed, which is referred to as the accuracy or confidence of the recognition (hereinafter referred to as recognition accuracy or simply accuracy).

[0004] Patent document 1 describes a device that displays a character code if the recognition accuracy output by a character recognizer together with the character code of the recognition result for a character image is above a threshold, and displays a character image and requires manual input from the user if the recognition accuracy is less than the threshold.

[0005] In the device disclosed in Patent Document 2, if the recognition accuracy is equal to or higher than a threshold value, an output is performed that does not require manual verification, and otherwise manual verification is performed.

[0006] Patent Document 1: Japanese Patent Application Publication No. 2000-259847

[0007] Patent Document 2: Japanese Patent Application Publication No. 2003-346080 Summary of the invention

[0008] An object of the present invention is to reduce the required human workload compared to a method of controlling which of the first processing and the second processing requiring more human workload than the first processing is applied to the recognition result of an identifier for object data based only on the recognition accuracy of the recognition result of the identifier.

[0009] The invention involved in Scheme 1 is an information processing device, characterized in that it includes a processor, which executes a first acquisition process of acquiring a first recognition result and a first recognition accuracy for object data from a first identifier, and executes a second acquisition process of acquiring a second recognition accuracy for the first recognition result for the object data from a second identifier, and controls which of the first process to be executed on the first recognition result and the second process requiring more human workload than the first process based on the first recognition accuracy and the second recognition accuracy.

[0010] The invention involved in Scheme 2 is the information processing device described in Scheme 1, characterized in that, in the control, when the first recognition accuracy is higher than the first threshold and the second recognition accuracy is higher than the second threshold, the first processing is executed, otherwise the second processing is executed.

[0011] The invention involved in Scheme 3 is the information processing device described in Scheme 1, characterized in that, in the first acquisition process, a plurality of candidates for the first recognition result of the object data and the recognition accuracy of each candidate, i.e., the first candidate recognition accuracy, are acquired from the first identifier, and in the second acquisition process, for each of the candidates, the recognition accuracy of the candidate for the object data, i.e., the second candidate recognition accuracy, is acquired from the second identifier, and in the control, if there is a candidate among the plurality of candidates whose first candidate recognition accuracy is higher than the first threshold and whose second candidate recognition accuracy is higher than the second threshold, the first process is performed on the candidate.

[0012] The invention involved in Scheme 4 is the information processing device described in Scheme 1, characterized in that the first recognizer recognizes the character string represented by the character string image as the object data, and obtains the recognized character string as the first recognition result, and the second recognizer recognizes the character represented by each character image contained in the character string image, and obtains the recognition accuracy of the recognition result for the character corresponding to the character image in the character string represented by the first recognition result among multiple recognition results regarding the character image as the second recognition accuracy.

[0013] The invention involved in Scheme 5 is the information processing device described in Scheme 4, characterized in that, in the control, when the first recognition accuracy is higher than the first threshold and the second recognition accuracy of each character in the character string represented by the first recognition result is all higher than the second threshold, the first processing is executed, otherwise the second processing is executed.

[0014] The invention involved in Scheme 6 is an information processing device as described in Scheme 4, wherein, in the first acquisition process, a plurality of candidates for the first recognition result of the object data and the recognition accuracy of each candidate, i.e., the first candidate recognition accuracy, are acquired from the first identifier, and in the second acquisition process, for each of the candidates, the recognition accuracy of each character in the character string of the candidate of the object data, i.e., the second candidate recognition accuracy, is acquired from the second identifier, and in the control, if there is a candidate among the plurality of candidates whose first candidate recognition accuracy is higher than the first threshold and whose second candidate recognition accuracy of each character is all higher than the second threshold, the first process is performed on the candidate.

[0015] The invention involved in Scheme 7 is a storage medium, which stores a program for causing a computer to execute the following processing: executing a first acquisition process of acquiring a first recognition result and a first recognition accuracy for object data from a first identifier; executing a second acquisition process of acquiring a second recognition accuracy for the first recognition result for the object data from a second identifier; and controlling which of the following processes to execute on the first recognition result: a first process or a second process requiring more human workload than the first process, based on the first recognition accuracy and the second recognition accuracy.

[0016] The invention involved in Scheme 8 is an information processing method, characterized in that it includes the following steps: executing a first acquisition process of acquiring a first recognition result and a first recognition accuracy for object data from a first identifier; executing a second acquisition process of acquiring a second recognition accuracy for the first recognition result for the object data from a second identifier; and controlling which of the following processes is executed on the first recognition result: a first process or a second process requiring more human workload than the first process, based on the first recognition accuracy and the second recognition accuracy.

[0017] Effects of the Invention

[0018] According to the first, second, fourth, fifth, seventh or eighth scheme of the present invention, compared with a method of controlling which of the first processing and the second processing requiring more human workload than the first processing is applied to the recognition result of the identifier for the object data based only on the recognition accuracy of the recognition result of the identifier, the required human workload can be reduced.

[0019] According to the third or sixth aspect of the present invention, the required human workload can be reduced compared to a method that targets only a single first recognition result. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Embodiments of the present invention will be described in detail with reference to the following drawings.

[0021] Figure 1 is a diagram illustrating a system configuration of an embodiment of character string recognition;

[0022] Figure 2 is a diagram illustrating the hardware configuration of a computer that performs the main information processing of the system;

[0023] Figure 3 is a diagram illustrating a conventional processing flow;

[0024] Figure 4 is a diagram illustrating a process flow of a system according to an embodiment;

[0025] Figure 5 is a diagram used to illustrate the effects obtained through the system;

[0026] Figure 6 is a diagram schematically showing an example of a UI screen for setting a threshold value;

[0027] Figure 7 is a diagram showing another example of the processing flow of the system according to the embodiment;

[0028] Figure 8 is a diagram illustrating a system configuration of a generalized embodiment;

[0029] Fig. 9 It is a diagram illustrating the processing flow of a generalized implementation method.

[0030] Explanation of symbols

[0031] 10 - OCR, 50 - input image data, 100 - information processing device, 102 - processor, 104 - memory, 106 - auxiliary storage device, 108 - input and output device, 110 - network interface, 112 - bus. DETAILED DESCRIPTION

[0032] <Example of the overall system>

[0033] exist Figure 1 An implementation of a data input system (hereinafter referred to as the "system") is described below.

[0034] In this system, input image data 50 as a target of character recognition is input to an OCR (character recognizer) 10. The OCR 10 performs character recognition processing on the input image data 50. The input image data 50 is image data of an image including a character string. For example, an image of a certain input field cut out from a form image by preprocessing becomes the input image data 50.

[0035] OCR10 cuts the image of each character from the input image data 50, and performs character recognition processing on the image of each character (hereinafter referred to as "character image"), thereby obtaining the recognition result of each character image and the accuracy of the recognition result. The recognition result is the character code of the character represented by the character image. OCR10 obtains more than one candidate for the recognition result for a character image, and obtains the accuracy for each of these candidates. The multiple candidates obtained for the same character image can be sorted in descending order of the corresponding accuracy. Hereinafter, the candidate for the recognition result obtained for the character image and the accuracy corresponding to the candidate are referred to as single character candidates and single character accuracy.

[0036] OCR10 recognizes the character string represented by the input image data 50. In the recognition of the character string, for example, a character string is generated by arranging the single character candidates as the recognition results of each character image in the arrangement order of each character image in the input image data 50, and the character string is evaluated. Since there are usually a plurality of single character candidates for each character image, it is possible to form different character strings of a number equivalent to the number of combinations of the single character candidates of each character image. OCR10 evaluates each of these character strings. In the evaluation, the evaluation value of each character string is calculated, for example, by referring to information on the ease of connection between characters or a vocabulary in which words or phrases are registered. And, OCR10 outputs the character string with the highest evaluation value as the recognition result corresponding to the input image data 50. The recognition result about the character string is called a character string recognition result. And, OCR10 calculates the accuracy corresponding to the character string recognition result and outputs it. This accuracy is called character string accuracy.

[0037] OCR10 uses the conventionally known technology to perform the above processing. Conventionally, OCR outputs the string recognition result and string accuracy, but Figure 1 The OCR 10 illustrated in FIG. 1 also outputs the single-character accuracy for each of the character codes included in the character string recognition result.

[0038] The OCR 10 can also be considered as a device having a first recognizer for recognizing individual characters and a second recognizer for recognizing the entire character string using the recognition results of the individual characters by the first recognizer.

[0039] The information processing apparatus 100 uses the information input from the OCR 10 to execute a process for obtaining a final recognition result as a system for the input image data 50 .

[0040] <Example of hardware configuration>

[0041] For example, the information processing device 100 is constructed based on a general-purpose computer. Figure 2As shown, the information processing device 100 has the following circuit structure: a processor 102 as a hardware control, a memory (main storage device) 104 such as a random access memory (RAM), a controller of an auxiliary storage device 106 such as a flash memory or SSD (solid state drive), HDD (hard disk drive), etc., an interface with various input and output devices 108, and a network interface 110 for controlling connection with a network such as a local area network, etc. are connected through a data transmission path such as a bus 112. The program group representing the function of the information processing device 100 described below is installed in a computer through a network or the like, and is stored in the auxiliary storage device 106. The functions of the information processing device 100 are realized by the processor 102 executing these program groups stored in the auxiliary storage device 106 using the memory 104.

[0042] Here, the processor 102 refers to a processor in a broad sense, including a general-purpose processor (such as a CPU: Central Processing Unit, etc.) or a dedicated processor (such as a GPU: Graphics Processing Unit, ASIC: Application Specific Integrated Circuit, FPGA: Field Programmable Gate Array, programmable logic device, etc.).

[0043] Furthermore, the actions of the processor 102 may be performed not only by one processor 102 but also by cooperation of a plurality of processors 102 located in physically separate locations. Furthermore, the actions of the processor 102 during program execution are not limited to the order described in the following embodiments and may be changed as appropriate.

[0044] exist Figure 1 In the example shown, OCR 10 is an external device of information processing device 100. OCR 10 may be, for example, a dedicated OCR device connected to information processing device 100 via a communication cable or the like, or an OCR service located on the Internet. Furthermore, information processing device 100 may have OCR 10 built in. For example, information processing device 100 executes a program indicating the function of OCR 10.

[0045] <Example of the previous process flow>

[0046] If the information processing device 100 performs character string recognition in the conventional manner, the processing flow is as follows: Figure 3 exemplified.

[0047] In this process, the processor 102 obtains the string recognition result and the string accuracy from the OCR 10 (S1), and determines whether the string accuracy is higher than a specified (i.e., predetermined) threshold value A (S2). When the determination result is "yes", the processor 102 outputs the string recognition result as the final recognition result of the system (S3). And, when the determination result of S2 is "no", the processor 102 rejects the string recognition result (S4). Rejection means not adopting the character recognition result that becomes the object as the final recognition result of the system. When the string recognition result of the OCR 10 is rejected, the processor 102 requires the worker to intervene in S4. In this case, the processor 102 displays the input image data 50 to the terminal of the predetermined worker, and receives, for example, the input of the string represented by the input image data 50 from the worker. Alternatively, the processor 102 displays the string recognition result together with the input image data 50 on the worker's terminal, and receives confirmation and correction of the string recognition result from the worker. In this case, the worker compares the input image data 50 displayed on the terminal with the character string recognition result, and when it is determined that the character string of the character string recognition result is correct, the worker performs a confirmation operation on the character string, and when it is determined to be wrong, the worker corrects the character string. The processor 102 outputs the character string that has been confirmed or corrected by the worker as the final recognition result of the system.

[0048] Thus, in the conventional method, when the character string accuracy is high enough (ie, higher than the threshold A), the character string recognition result is used as the final recognition result without human intervention.

[0049] <Processing flow of implementation method>

[0050] The processing flow of this embodiment relative to the above-mentioned conventional flow is shown as an example. Figure 4 middle.

[0051] In this process, the processor 102 obtains the string recognition result, the string accuracy, and the single character accuracy of each character included in the string recognition result from the OCR 10 (S10). Then, the processor 102 determines whether the obtained string accuracy is higher than a predetermined threshold value 1 (S12). The threshold value 1 used in this determination is lower than Figure 3 The value of threshold A used in S2 of the previous process can be used.

[0052] When the determination result of S12 is "yes", the processor 102 further compares the single character accuracy of the single character results of each character included in the string recognition result with the specified threshold value 2. The threshold value 2 is, for example, a value common to all characters. Moreover, the processor 102 determines whether the single character accuracy of all characters included in the string recognition result is higher than the threshold value 2 (S14). As long as there is a character whose single character accuracy is below the threshold value 2 among the characters in the string recognition result, the determination result of S14 will become "no".

[0053] When the determination result of S14 is "Yes", the processor 102 executes the first process (S16). On the other hand, when the determination result of S12 or S14 is "No", the processor 102 executes the second process (S18).

[0054] The first process is a process that requires relatively less work for the worker than the second process. Figure 2 As in S3 of the process, the process of directly outputting the character string recognition result as the final processing result without the intervention of the worker is an example of the first process, and the process of rejecting the character string recognition result and accepting the worker's confirmation / correction is an example of the second process.

[0055] As another example of the first process, there is a process of only asking the worker to confirm the character string recognition result. In this process, the processor 102 displays the input image data 50 and the character string recognition result on the screen of the worker's terminal, and asks the worker to input a confirmation result of whether the character string recognition result is correct. In addition, as another example of the second process that replaces the worker's confirmation / correction, there is a process of prompting the worker to input the input image data 50 and asking the worker to input the text data of the character string included in the input image data 50.

[0056] exist Figure 4 In the process, the execution order of S12 and S14 can also be reversed.

[0057] Compared with the second process, the first process requires less human effort, so the higher the probability of selecting the first process, the less human effort required for the system as a whole. In this embodiment, in addition to the string accuracy used in the past, the single character accuracy is also used, thereby increasing the probability of selecting the first process. Figure 5 This is explained.

[0058] Figure 5A scatter diagram 500 showing the recognition results when OCR 10 is used to recognize a plurality of sample input images, wherein the vertical axis is the string accuracy and the horizontal axis is the representative value of the single character accuracy (e.g., the lowest value among the single character accuracy of each character in the string recognition result). The string accuracy is a higher value as it goes up, and the single character accuracy is a higher value as it goes to the left. The dark diamond plot 502 represents the sample that was recognized accurately, and the light square and triangle plots 504 and 506 represent the sample that was misrecognized.

[0059] exist Figure 5 In the distribution shown, in order to set the accuracy rate to a sufficiently high value (i.e., a value close to a predetermined threshold value of 100%) using only the string accuracy, the string accuracy needs to be higher than the threshold value A shown in the figure. In contrast, when the single character accuracy is also used together, if the string accuracy is higher than the threshold value 1 shown in the figure and the single character accuracy is higher than the threshold value 2, the accuracy rate becomes a sufficiently high value. When comparing the range where the string accuracy is higher than the threshold value A and the range where the string accuracy is higher than the threshold value 1 and the single character accuracy is higher than the threshold value 2, the latter range contains a larger number of accurately recognized plots 502. Therefore, the latter has a higher ratio of string recognition results for the first process in which the workload of the user is relatively small while maintaining the accuracy rate.

[0060] <Threshold setting>

[0061] The system of the present embodiment can provide a UI (user interface) screen for receiving settings of a threshold used for determination from the user.

[0062] exist Figure 6 The UI screen 600 is illustrated in FIG. The UI screen 600 is used to set Figure 4 Threshold 1 and threshold 2 are used in the process illustrated in .

[0063] The UI screen 600 displays Figure 5 . The scatter diagram 500 is the same as the scatter diagram illustrated in . The setting of the threshold 1 for the string accuracy and the threshold 2 for the single character accuracy is received on the scatter diagram 500. The threshold 1 is shown by a horizontal line crossing the scatter diagram 500, and the threshold 2 is shown by a vertical line intersecting the scatter diagram 500. The user inputs the value of the threshold in the input field 602 of the threshold 1 and the input field 604 of the threshold 2. The lines of each threshold shown on the scatter diagram 500 are displayed at the position corresponding to the threshold in these input fields 602 to 604.

[0064] In addition, a recognition rate display column 610 and a ratio display column 620 are shown in the UI screen 600. The recognition rate display column 610 shows the recognition rate achieved by the threshold group set by the user. Figure 4The recognition rate or accuracy rate when the first process (S16) is executed in the process. In this example, the string recognition result is output in the first process as the final recognition result of the system. In this case, the recognition rate displayed is the ratio of the number of corresponding string recognition results that are accurately recognized to the total number of sample image groups whose judgment results of S12 and S14 are both "yes" based on the set threshold group. In addition, the ratio display column 620 displays the ratio of the number of sample images to which the first process (S16) is applied based on the set threshold group in the total number of sample images. The higher the ratio, the less human workload is required. The user changes each threshold while confirming the values ​​displayed in the recognition rate display column 610 and the ratio display column 620 to determine the combination of thresholds that can obtain the desired performance.

[0065] < Figure 4 Modification of the processing flow of

[0066] refer to Figure 7 A modification example of the process executed by the processor 102 according to the present embodiment will be described.

[0067] exist Figure 7 In the process of , the processor 102 obtains information of a plurality of candidate recognition results from the OCR 10 (S20). Figure 4 In the process, the processor 102 obtains a string recognition result with the highest string accuracy from OCR10. In contrast, in S20, multiple string recognition results are obtained in descending order of string accuracy as candidates. The number of candidates obtained is, for example, a predetermined number. In addition, the number of candidates obtained may be changed according to the recognition object, for example, a string recognition result with the highest string accuracy and a string recognition result with a string accuracy within a specified difference or above a specified ratio relative to the highest value may be selected as candidates. In addition, a string recognition result with a string accuracy higher than a threshold value 1 may be selected as a candidate. The threshold value 1 is greater than the threshold value 1. Figure 4 In S20, the processor 102 obtains the character string recognition result and the character string accuracy corresponding thereto and the single character accuracy of each character included in the character string recognition result for each candidate.

[0068] Next, the processor 102 takes the candidate with the highest string accuracy as the candidate of interest (S22), and determines whether the string accuracy of the candidate of interest is higher than threshold 1 (S24) and whether the single character accuracy of each character of the candidate of interest is all higher than threshold 2 (S26). If the determination results of S24 and S26 are both "yes", the processor 102 takes the string recognition result of the candidate of interest as the object and performs the first processing (S28). The first processing is, for example, a processing of outputting the string recognition result of the candidate of interest as the final recognition result of the system.

[0069] When at least one of the judgment results of S24 or S26 is "No", the processor 102 determines whether the judgments of S24 and S26 have been completed for all candidates obtained in S20 (S30). When the judgment result of S30 is "No", the processor 102 takes the second candidate, that is, the candidate with a higher string accuracy than the focus candidate, as a new focus candidate (S32), and repeats the judgments of S24 and S26. When the judgment result of S30 becomes "Yes" during the repetition of the above-described processing, this means that there is no candidate in the candidate for which both the judgment results of S24 and S26 become "Yes". In this case, the processor 102 executes the second processing (S34). The second processing is a processing that requires more human work than the first processing, for example, a processing for accepting confirmation / correction of the string recognition result of the highest-ranked candidate by a worker is an example.

[0070] According to the above description Figure 7 The process is different from the process of determining only the string recognition result with the highest string accuracy. Figure 4 Compared with the process, the number of cases where the first processing (S28) is applied increases, and the workload of people is reduced accordingly.

[0071] exist Figure 7 In the process, the candidates with high string accuracy are evaluated in order, and the candidate that is both "yes" in S24 and S26 for the first time is the object of the first process (S28), but this process is just an example. Alternatively, for example, S24 and S26 can be determined for all candidates, and the candidate with the highest evaluation value when comprehensively evaluating the string accuracy and the single character accuracy among the candidates that are both "yes" in S24 and S26 is the object of the first process.

[0072] <Example of generalized system structure>

[0073] The system described above uses a string accuracy and a single character accuracy calculated by an OCR 10.

[0074] In contrast, Figure 8The system illustrated in the example has two OCR1 and OCR2. OCR1 and OCR2 perform different character recognition processes. That is, in OCR1 and OCR2, for example, the character recognition programs used are different, or the learning data sets used in the character recognition learning are different. When OCR1 and 2 process a sufficiently large number of sample image groups, the distribution of correct answers (i.e., accurate recognition) and incorrect answers (i.e., erroneous recognition) for the OCR results of the sample image groups is different in OCR1 and 2. OCR1 and 2 recognize the same input image data 50. In this example, OCR1 is a benchmark recognizer, and OCR2 is an auxiliary recognizer.

[0075] The information processing device 100 obtains the recognition result (referred to as "recognition result 1") and the recognition accuracy (referred to as "accuracy 1") of the recognition result 1 for the input image data 50 from OCR1. The recognition result 1 is the recognition result with the highest accuracy among the recognition results of the input image data 50 recognized by OCR1. Furthermore, the information processing device 100 obtains the recognition accuracy (referred to as "accuracy 2") of the recognition result of the input image data 50 that is the same as the recognition result 1 of OCR1 from OCR2. For example, consider the case where the recognition result 1 of OCR1 for a certain input image data 50 is "ABC". OCR2 obtains several candidates of the recognition result and the accuracy of the candidates for the input image data 50, but the information processing device 100 obtains the accuracy of the candidate with the value "ABC" among these candidates as accuracy 2. Furthermore, the information processing device 100 uses the information obtained from OCR1 and 2 to perform processing for obtaining the final recognition result of the input image data 50 as a system.

[0076] exist Fig. 9 The process flow of this process is illustrated in FIG. In this process, the processor 102 obtains the recognition result 1 and the accuracy 1 for the input image data 50 from OCR1, and obtains the accuracy 2 for the recognition result 1 for the same input image data from OCR2 (S40). Next, the processor 102 determines whether the accuracy 1 is higher than a predetermined threshold value T1 (S42), and whether the accuracy 2 is higher than a predetermined threshold value T2 (S44). The threshold values ​​T1 and T2 are based on the reference value T1. Figure 5 The method for determining the thresholds 1 and 2 described above is the same as that for determining the thresholds 1 and 2. That is, if the accuracy 1 of the recognition result 1 of OCR1 is higher than the threshold A, the recognition result 1 is adopted as the final recognition result, and the accuracy of the system becomes higher than the specified value. Since the accuracy 2 is also considered in this example, the threshold T1 can be a value lower than the threshold A.

[0077] If the determination results of both S42 and S44 are "yes", the processor 102 performs the first process (S16) with the recognition result 1 as the object. In the first process, for example, the recognition result 1 is output as the final recognition result of the system. When the determination result of at least one of S42 or S44 is "no", the processor 102 performs the second process (S48). The second process is a process that requires more human work than the first process, for example, a process of accepting confirmation / correction of the recognition result 1 by a worker is an example.

[0078] In use Figure 8 and Fig. 9 In the example described, two OCRs 1 and 2 are used, but of course, one recognition device having the functions of both OCRs 1 and 2 may be used instead.

[0079] Above, in reference Figure 8 and Fig. 9 Description of the system, with reference to Figure 5 Explanatory Figure 4 The effect of the processing flow is similar, which is to maintain accuracy while increasing the proportion of application of the first processing.

[0080] Fig. 9 The process corresponds to Figure 4 process, but as Fig. 9 The process variation can also be considered relative to Figure 4 The process has Figure 7 In this modified process, the information processing device 100 obtains a plurality of recognition result candidates and their accuracy in descending order of accuracy from OCR1, and obtains the accuracy of the recognition result with the same value as each of these candidates from OCR2. The subsequent processing is the same as Figure 7 The process shown is the same.

[0081] The above-described embodiments are merely examples, and various modifications are possible within the scope of the present invention.

[0082] The above-mentioned embodiments of the present invention are provided for the purpose of illustration and description. In addition, the embodiments of the present invention do not fully and exhaustively include the present invention, and do not limit the present invention to the disclosed methods. Obviously, various modifications and changes are self-evident to those skilled in the art to which the present invention belongs. The present embodiment is selected and described in order to most easily understand the principles of the present invention and its application. Thus, other technicians in the technical field can understand the present invention by using various modified examples optimized for the determination of the assumed various embodiments. The scope of the present invention is defined by the above claims and their equivalents.

Claims

1. An information processing device, characterized in that: With processor, The processor performs the following processing: executing a first acquisition process of acquiring a first recognition result and a first recognition accuracy for the object data from a first recognizer; executing a second acquisition process of acquiring, from a second identifier, a second recognition accuracy with respect to the first recognition result for the object data; and According to the first recognition accuracy and the second recognition accuracy, control is performed on whether to perform a first process or a second process requiring more human effort than the first process on the first recognition result. The first recognizer recognizes a character string represented by the character string image as the object data, and obtains the recognized character string as the first recognition result. The second recognizer recognizes the character represented by each character image contained in the string image, and obtains the recognition accuracy of the recognition result of the character corresponding to the character image in the string represented by the first recognition result as the second recognition accuracy, and In the control, when the first recognition accuracy is higher than a first threshold and the second recognition accuracy of each character in the character string represented by the first recognition result is all higher than a second threshold, the first process is executed, otherwise the second process is executed.

2. The information processing device according to claim 1, characterized in that In the first acquisition process, a plurality of candidates of the first recognition result for the object data and a first candidate recognition accuracy which is a recognition accuracy for each candidate are acquired from the first identifier. In the second acquisition process, for each of the candidates, a second candidate recognition accuracy, which is a recognition accuracy for each character in the candidate character string of the target data, is acquired from the second recognizer. In the control, if there is a candidate among the plurality of candidates whose first candidate recognition accuracy is higher than a first threshold and whose second candidate recognition accuracy for each character is all higher than a second threshold, the first process is performed on the candidate.

3. A storage medium storing a program for causing a computer to execute the following processing: executing a first acquisition process of acquiring a first recognition result and a first recognition accuracy for the object data from a first recognizer; executing a second acquisition process of acquiring, from a second identifier, a second recognition accuracy level regarding the first recognition result for the object data; and According to the first recognition accuracy and the second recognition accuracy, control is performed on whether to perform a first process or a second process requiring more human effort than the first process on the first recognition result. The first recognizer recognizes a character string represented by the character string image as the object data, and obtains the recognized character string as the first recognition result. The second recognizer recognizes the character represented by each character image contained in the string image, and obtains the recognition accuracy of the recognition result of the character corresponding to the character image in the string represented by the first recognition result as the second recognition accuracy, and In the control, when the first recognition accuracy is higher than a first threshold and the second recognition accuracy of each character in the character string represented by the first recognition result is all higher than a second threshold, the first process is executed, otherwise the second process is executed.

4. An information processing method, characterized in that: The steps include: executing a first acquisition process of acquiring a first recognition result and a first recognition accuracy for the object data from a first recognizer; executing a second acquisition process of acquiring, from a second identifier, a second recognition accuracy level regarding the first recognition result for the object data; and According to the first recognition accuracy and the second recognition accuracy, control is performed on whether to perform a first process or a second process requiring more human effort than the first process on the first recognition result. The first recognizer recognizes a character string represented by the character string image as the object data, and obtains the recognized character string as the first recognition result. The second recognizer recognizes the character represented by each character image contained in the string image, and obtains the recognition accuracy of the recognition result of the character corresponding to the character image in the string represented by the first recognition result as the second recognition accuracy, and In the control, when the first recognition accuracy is higher than a first threshold and the second recognition accuracy of each character in the character string represented by the first recognition result is all higher than a second threshold, the first process is executed, otherwise the second process is executed.

5. A computer program product for causing a computer to execute the process as claimed in claim 4.

Images