JP7704045B2 - Document correction support program, document correction support method, and information processing device - Google Patents

Description

Embodiments of the present invention relate to a document correction assistance program, a document correction assistance method, and an information processing device.

Conventionally, various documents (hereafter referred to as documents) submitted at counters etc. are checked for consistency between documents, and employees correct any inconsistencies. For example, many tax return forms are submitted at tax counters every year. Employees check the submitted documents for errors by comparing them with basic information about residents and documents submitted by their employers.

Figure 8 is an explanatory diagram that explains an example of correcting document deficiencies. As shown in Figure 8, resident H1 submits a final tax return D1 and a resident tax return D2 to city hall. In addition, resident H1's employers K1 and K2 submit salary payment reports D3 and D4 related to resident H1. In addition, the pension service K3 submits a pension payment report D5 related to resident H1. City hall employee H2 compares the details of each item in the submitted final tax return D1, resident tax return D2, salary payment reports D3 and D4, and pension payment report D5. Employee H2 then detects any items that are inconsistent and corrects the data for those items.

To support such document correction work, there is a conventional technique that checks the contents of invoices and detailed statements received from each medical institution server according to existing rules. There is also a conventional technique that judges the validity of the item contents based on whether the semantic vector of each item content of a document deviates from the center of gravity of the semantic vector of the corresponding item in a pre-registered normative document. There is also a conventional technique that displays the evaluation results obtained by inputting the document data to be evaluated into a trained model obtained by machine learning the corrected document data. There is also a conventional technique that updates the model by using a new set of documents as training data.

JP 2007-241986 A JP 2020-140442 A JP 2018-147280 A JP 2021-89473 A

There are cases where the submitted documents contain multiple documents of the same type, such as salary payment reports D3 and D4. In such cases, city hall employee H2 considers the relationships between the documents and decides which of the documents of the same type to revise.

However, the above-mentioned conventional technology does not take into account the relationships between the multiple documents submitted, and even if it is possible to identify inconsistent items, it is difficult to identify which documents need to be revised.

In one aspect, the objective is to provide a document correction assistance program, a document correction assistance method, and an information processing device that can easily identify documents to be corrected.

In one proposal, the document correction assistance program causes a computer to execute an input process and an output process. The input process is based on learning data of a plurality of cases including the correction history of each document, and inputs explanatory variables generated based on each document to a machine-learned model based on explanatory variables including the difference between a value calculated from the value of an item included in a specific first document and the value of an item included in a specific second document based on a predefined restoration calculation for each case, and the value of the item included in the first document, and a target variable including a value corresponding to the correction history of each document of the cases. The output process is based on the output from the model and outputs whether or not each document to be determined is corrected.

Documents to be corrected can be easily identified.

FIG. 1 is a block diagram illustrating an example of a functional configuration of an information processing apparatus according to an embodiment. FIG. 2 is an explanatory diagram illustrating an example of the restoration computation definition information. FIG. 3 is a flowchart illustrating an example of the operation of the information processing device according to the embodiment. FIG. 4 is a flowchart illustrating an example of a learning process of the information processing device according to the embodiment. FIG. 5 is an explanatory diagram illustrating an example of learning data. FIG. 6 is a flowchart illustrating an example of a determination process of the information processing device according to the embodiment. FIG. 7 is a block diagram showing an example of a computer configuration. FIG. 8 is an explanatory diagram for explaining an example of correcting a document defect.

The document correction support program, document correction support method, and information processing device according to the embodiments will be described below with reference to the drawings. Configurations having the same functions in the embodiments will be given the same reference numerals, and duplicated descriptions will be omitted. Note that the document correction support program, document correction support method, and information processing device described in the following embodiments are merely examples, and do not limit the embodiments. In addition, the following embodiments may be combined as appropriate within a range that does not cause contradictions.

FIG. 1 is a block diagram showing an example of the functional configuration of an information processing device according to an embodiment. As shown in FIG. 1, the information processing device 1 has an input unit 10, a target variable generation unit 20, an explanatory variable generation unit 30, a storage unit 40, a learning unit 50, and a determination unit 60.

The information processing device 1 is an example of a learning device that generates a learning model 71 by machine learning the corrected material data 11 as learning data. The information processing device 1 is also an example of a determination device that determines whether or not each document included in the uncorrected material data 12 to be determined has been corrected using the generated learning model 71 and outputs the result.

The above-mentioned learning device and judgment device may be realized by a single information processing device 1, or may be realized separately. For example, the information processing device 1 may be a learning device having an input unit 10, a target variable generation unit 20, an explanatory variable generation unit 30, a memory unit 40, and a learning unit 50. The information processing device 1 may also be a judgment device having an input unit 10, a target variable generation unit 20, an explanatory variable generation unit 30, a memory unit 40, and a judgment unit 60.

The input unit 10 is a processing unit that accepts input of various data such as corrected material data 11 and uncorrected material data 12.

Corrected document data 11 is data that describes the corrected content of each document in each case. For example, corrected document data 11 is data that includes the content of each document (document) in each case (e.g., the value entered in each item) and whether or not each document has been corrected (e.g., an item with an error and the corrected value for that item). As an example, corrected document data 11 includes the values of the items of each document submitted by resident H1 and corrected by employee H2 in tax procedures, and whether or not it has been corrected.

In machine learning, based on the corrected document data 11, for each case, feature quantities corresponding to the contents of each document are used as explanatory variables, and a value corresponding to the presence or absence of corrections is generated as a target variable. Next, in machine learning, when the generated explanatory variables are input into the input layer of a model such as a neural network, the model parameters are calculated so that the target variable can be obtained from the output layer of the model.

Uncorrected document data 12 is data that describes the contents of each document related to the case being assessed. For example, uncorrected document data 12 is data that includes the contents of each document (document) in the case being assessed (e.g., the values entered in each item). As an example, uncorrected document data 12 includes the values of the items in each document submitted by resident H1 for tax purposes.

The objective variable generation unit 20 is a processing unit that generates an objective variable based on the corrected material data 11. Specifically, the objective variable generation unit 20 generates an objective variable that quantifies the presence or absence of corrections in each document of the case for each case included in the corrected material data 11 during machine learning.

The explanatory variable generation unit 30 is a processing unit that generates explanatory variables based on the corrected document data 11 or the uncorrected document data 12. Specifically, the explanatory variable generation unit 30 generates explanatory variables that quantify the contents of each document (document) (e.g., the entered values in each item) for each case of the corrected document data 11 or the uncorrected document data 12.

During machine learning, the information processing device 1 combines the objective variables generated by the objective variable generation unit 20 for each document and the explanatory variables generated by the explanatory variable generation unit 30 for each case of the corrected material data 11, for example, in an array with each document as a row, to obtain learning data 70 for machine learning. Furthermore, during judgment using the learning model 71, the information processing device 1 obtains evaluation data 80 by combining the explanatory variables generated by the explanatory variable generation unit 30 for each document based on the uncorrected material data 12, for example, in an array with each document as a row.

The explanatory variables generated by the explanatory variable generation unit 30 include a difference between a value calculated from the values of items in a specific document and a specific document (called the other document) other than the specific document based on a predefined restoration calculation, and the value of the item in the specific document, i.e., a value corresponding to the relationship between the specific document and the other document.

Here, the specific document and other documents are specified in advance in the definition content of the reconstruction operation. For example, by specifying a document type in the definition content, documents of the same type are specified. For example, by specifying a payroll payment report in the definition content of the reconstruction operation, multiple payroll payment reports submitted in each case (e.g., payroll payment reports D3 and D4) become the specific document and other documents.

The reconstruction calculation is a calculation that defines a typical "way of dividing documents" between a specific document and other documents and restores it. As an example, the reconstruction calculation is a calculation that employee H2 performs in his head when estimating the correct values for the items of multiple documents (e.g., salary payment reports D3, D4) submitted by resident H1 for each document type. In other words, the results of the reconstruction calculation are candidates for the correct values for each document type. The results of the reconstruction calculation represent the characteristics of the entire document of that document type (e.g., salary payment reports D3, D4). Therefore, the difference between the calculation result and the items of a specific document indirectly represents the relationship between the specific document and other documents.

To generate the explanatory variables, the explanatory variable generation unit 30 has a reconstruction calculation unit 31 and a difference calculation unit 32.

The restoration calculation unit 31 is a processing unit that performs restoration calculations based on the restoration calculation definition information 41 stored in the memory unit 40, using the values of items of a specific document included in each case of the modified material data 11 or the unmodified material data 12 and other documents separate from that document.

The difference calculation unit 32 is a processing unit that calculates the difference between the result of the restoration calculation performed by the restoration calculation calculation unit 31 and the value of an item of a specific document based on the restoration calculation definition information 41.

Figure 2 is an explanatory diagram explaining an example of restoration calculation definition information 41. As shown in Figure 2, the restoration calculation definition information 41 defines a typical "way of separating documents" between a specific document and other documents, and shows a "restoration calculation" that restores it. Note that the restoration calculation definition information 41 is prepared for each type of document, such as tax returns, D2, D3, etc., and the contents of the restoration calculation for each item between documents are defined for each document type. Note that in the following example, the values of documents targeted by the restoration calculation are limited to 0 and positive, but the values of documents targeted by the restoration calculation are not limited to these.

For example, the "way of dividing data" in the "reconstruction calculation" defined in the reconstruction calculation definition information 41 may be a case where the value of one data (e.g., salary payment report D3) is included in the value of another data (e.g., salary payment report D4). In the "reconstruction calculation" in such a case, the maximum value of the values of the two data is found. Therefore, the meaning of the difference between the result of the reconstruction calculation in this case and the value of a data is that if it is 0, then this data is the maximum value, and if it is non-zero, then one of the other data is the maximum value.

In addition, the "way of dividing materials" in the "restoration calculation" defined in the restoration calculation definition information 41 may mean that values are recorded in separate materials, with no overlap between the materials. In the "restoration calculation" in such a case, the sum of the values of all materials is calculated. Therefore, the meaning of the difference between the result of the restoration calculation in this case and the value of a certain material is that if it is 0, only this material has a positive value, and if it is non-zero, the other materials also have positive values.

In addition, the "way of dividing materials" in the "restoration calculation" defined in the restoration calculation definition information 41 can be a case where the value of a specific material item is mistakenly entered in the item of another material. In the "restoration calculation" in such a case, the value of the relevant material is subtracted from the sum of the values of all materials. Therefore, the meaning of the difference between the result of the restoration calculation in this case and the value of a certain material is that if it is 0, then only this material and the relevant material have positive values, and if it is non-zero, then either the other material also has a positive value, or this material is the relevant material or its value is 0.

The storage unit 40 is a storage device such as a memory or a hard disk drive (HDD), and stores the restoration calculation definition information 41.

The learning unit 50 is a processing unit that generates a learning model 71 by performing known machine learning processing for each example of the learning data 70 based on the explanatory variables generated by the explanatory variable generation unit 30 and the objective variables generated by the objective variable generation unit 20. The machine learning processing performed by the learning unit 50 includes decision trees, random forests, deep learning, and the like. For example, in the case of deep learning, the learning unit 50 generates the learning model 71 by determining hidden layer parameters so that, when an explanatory variable is input, an output corresponding to the objective variable is produced.

The judgment unit 60 is a processing unit that inputs evaluation data 80 including explanatory variables generated based on the uncorrected material data 12 for the case to be judged to the learning model 71 and obtains a judgment result for the case to be judged. Specifically, the judgment unit 60 reads out parameters obtained by machine learning in the learning unit 50 to construct the learning model 71. Next, the judgment unit 60 inputs the explanatory variables generated based on the uncorrected material data 12, i.e., the feature quantities of each document to be judged, to the learning model 71. Next, the judgment unit 60 obtains a degree of accuracy (evaluation value) indicating the presence or absence of corrections in each document to be judged from the output of the learning model 71. The judgment unit 60 outputs a judgment result 81 indicating the presence or absence of corrections in each document based on the obtained degree of accuracy, for example, by comparing it with a predetermined threshold value.

Fig. 3 is a flowchart showing an example of the operation of the information processing device according to the embodiment. As shown in Fig. 3, the information processing device 1 performs a learning process (S1) using the corrected material data 11 as learning data and including the results of the restoration calculation based on the restoration calculation definition information 41 as explanatory variables, thereby generating a learning model 71.

Figure 4 is a flowchart showing an example of the learning process of the information processing device 1 according to the embodiment. As shown in Figure 4, when the learning process is started, the restoration calculation unit 31 obtains the values of each submitted document (document) for each case (resident) included in the corrected document data 11. Next, the restoration calculation unit 31 performs a restoration calculation from the values of items of a specific document and another document separate from that document, based on the definition contents of the restoration calculation definition information 41 (S10).

Then, the difference calculation unit 32 calculates the difference between the reconstruction calculation result by the reconstruction calculation unit 31 and the value of the material (S11), and stores this calculation result in the learning data 70 as an explanatory variable (S12).

Next, the explanatory variable generation unit 30 determines whether the process of generating explanatory variables for all materials included in the case has been completed (S13). If the process has not been completed for all materials included in the case (S13: No), the explanatory variable generation unit 30 returns the process to S10.

When processing has been completed for all materials included in the case (S13: Yes), the explanatory variable generation unit 30 determines whether or not all the reconstruction operations defined in the reconstruction operation definition information 41 have been processed (S14). When not all the reconstruction operations have been processed (S14: No), the explanatory variable generation unit 30 returns the process to S10.

If all the reconstruction calculations have been processed (S14: Yes), the explanatory variable generation unit 30 determines whether the process for generating explanatory variables has been performed for all items of the document (S15). If the process has not been performed for all items (S15: No), the explanatory variable generation unit 30 returns to S10.

If all items have been processed (S15: Yes), the explanatory variable generation unit 30 determines whether all residents (cases) have been processed (S16). If not all residents have been processed (S16: No), the explanatory variable generation unit 30 returns to S10.

When all residents have been processed (S16: Yes), the objective variable generation unit 20 calculates whether each submitted document (document) has been corrected for each case (resident) included in the corrected document data 11 (S17). Next, the objective variable generation unit 20 stores the calculation result regarding whether each document has been corrected in the learning data 70 as an objective variable (S18).

Next, the objective variable generation unit 20 determines whether or not the process for generating objective variables has been performed for all materials (S19). If the process has not been performed for all materials (S19: No), the objective variable generation unit 20 returns the process to S17.

If all materials have been processed (S19: Yes), the learning unit 50 performs machine learning based on the learning data 70 to generate a learning model 71 (S20) and ends the process.

Figure 5 is an explanatory diagram illustrating an example of learning data 70. As shown in Figure 5, when performing machine learning, for each case (resident A, resident B) included in the corrected material data 11, learning data 70 is generated based on the corrected materials (1), (2), (3), etc., with the feature quantities corresponding to the content of each material as explanatory variables and the value corresponding to the presence or absence of correction as the objective variable.

Here, the explanatory variables in the learning data 70 include the difference between the value calculated from the values of items in a specific document and another document separate from that document, based on the restoration calculation defined in the restoration calculation definition information 41, and the value of the item in the specific document, i.e., the value corresponding to the relationship between the specific document and the other document.

For example, the explanatory variables for resident A's document (1) include values that correspond to the difference between the values obtained by the reconstruction calculations a, b, c, etc. for the items in document (1), i.e., values corresponding to the relationship between document (1) and other documents (e.g., document (2), document (3), etc.). Similarly, the explanatory variables for document (2) include values that correspond to the difference between the values obtained by the reconstruction calculations a, b, c, etc. for the items in document (2), i.e., values corresponding to the relationship between document (2) and other documents (e.g., document (1), document (3), etc.). Furthermore, the explanatory variables for document (3) include values that correspond to the difference between the values obtained by the reconstruction calculations a, b, c, etc. for the items in document (3), i.e., values corresponding to the relationship between document (3) and other documents (e.g., document (1), document (2), etc.). The same is true for each case (resident B, etc.).

In the information processing device 1, machine learning is performed based on the learning data 70 that includes such explanatory variables, thereby generating a learning model 71 that has learned the relationships between each document.

Returning to FIG. 3, following S1, the information processing device 1 applies the uncorrected material data 12 to the generated learning model 71, and performs a determination process to determine whether or not each document included in the uncorrected material data 12 has been corrected (S2). As a result, the information processing device 1 obtains a determination result 81 indicating whether or not each document has been corrected, and ends the process.

Figure 6 is a flowchart showing an example of the judgment process of the information processing device 1 according to the embodiment. As shown in Figure 6, when the judgment process is started, the restoration calculation unit 31 obtains the values of each document (document) submitted for the case (resident) to be judged that is included in the uncorrected document data 12. Next, the restoration calculation unit 31 performs a restoration calculation from the values of items of a specific document and another document separate from that document, based on the definition contents of the restoration calculation definition information 41 (S20).

Then, the difference calculation unit 32 calculates the difference between the reconstruction calculation result by the reconstruction calculation unit 31 and the value of the material (S31), and stores this calculation result in the evaluation data 80 as an explanatory variable (S32).

Next, the explanatory variable generation unit 30 judges whether the process of generating explanatory variables for all materials included in the case to be judged has been completed (S33). If the process has not been completed for all materials included in the case to be judged (S33: No), the explanatory variable generation unit 30 returns the process to S30.

When processing has been completed for all materials included in the case to be judged (S33: Yes), the explanatory variable generation unit 30 judges whether or not all the reconstruction operations defined in the reconstruction operation definition information 41 have been processed (S34). When not all the reconstruction operations have been processed (S34: No), the explanatory variable generation unit 30 returns the process to S30.

If all the restoration calculations have been processed (S34: Yes), the explanatory variable generation unit 30 determines whether the process for generating explanatory variables has been performed for all items of the document (S35). If the process has not been performed for all items (S35: No), the explanatory variable generation unit 30 returns the process to S30.

If all items have been processed (S35: Yes), the explanatory variable generation unit 30 determines whether all residents (cases) to be judged have been processed (S36). If not all residents have been processed (S36: No), the explanatory variable generation unit 30 returns to S30.

When all residents have been processed (S36: Yes), the judgment unit 60 inputs the evaluation data 80 into the learning model 71 to judge whether or not corrections have been made to each document of the case being judged, obtains the judgment result 81 (S37), and ends the process.

As described above, the information processing device 1 generates explanatory variables including the difference between a value calculated from the value of an item included in a specific first document and the value of an item included in a specific second document and the value of the item included in the first document, based on a predefined restoration calculation for each case, based on the corrected material data 11 of multiple cases including the revision history of each document. Next, the information processing device 1 generates a learning model 71 by machine learning based on the generated explanatory variables and a target variable including a value corresponding to the revision history of each document of the case. The information processing device 1 inputs explanatory variables generated in the same manner as during learning based on each document to be judged to this learning model 71, and outputs the presence or absence of revisions in each document to be judged based on the output from the learning model 71.

In this way, the explanatory variables input to the learning model 71 used for the judgment include values calculated from the values of items contained in a specific first document and the values of items contained in a specific second document, and the relationship between the documents based on the difference between the values and the items contained in the first document. Therefore, the information processing device 1 uses this learning model 71 to obtain the presence or absence of corrections in each document to be judged, and can identify documents that need to be corrected in accordance with the relationship between the documents.

In addition, the restoration calculation in the information processing device 1 is a calculation that finds the maximum value of the values of items between different specific documents. As a result, the information processing device 1 can include in the explanatory variables a feature that indicates the relationship between the documents, such that if the difference between the result of the restoration calculation that finds the maximum value of the values of items between different specific documents and the value of the first document is 0, the value of the first document is the maximum, and if the difference is not 0, one of the values of the other documents is the maximum. For example, one example of a typical way in which documents are separated is a case in which the value of a certain document (e.g., the first document) is contained in the value of another document (e.g., the second document). In the information processing device 1, the above restoration calculation can include in the explanatory variables a feature related to such a case.

In addition, the restoration calculation in the information processing device 1 is a calculation that finds the sum of the values of items between different specific documents. As a result, in the information processing device 1, the difference between the result of the restoration calculation that finds the sum of the values of items between different specific documents and the value of the first document can be included in the explanatory variables as a feature that indicates the relationship between the documents, such that if the value is 0, only the first document has a positive value, and if the value is non-zero, the other documents also have positive values. For example, one example of a typical way in which documents are separated is a case in which values are shaped in different documents and do not overlap. In the information processing device 1, the above restoration calculation can be used to include a feature related to such a case in the explanatory variables.

In addition, the restoration calculation in the information processing device 1 is a calculation that subtracts the value of an item of a specific third document included in either the first document or the second document from the total value of the values of items between different specific documents. As a result, in the information processing device 1, depending on the difference between the result of the restoration calculation that subtracts the value of the item of the third document from the total value of the values of items between different specific documents and the value of the first document, if the difference is 0, then only the values of the first document and the third document to be subtracted are positive, and if the difference is other than 0, then either the other document also has a positive value, or this document is a third document or the value is 0, then the explanatory variable can include a feature that indicates the relationship between the documents. For example, a typical example of how documents are divided is a case where a value of another item is erroneously written in a specific document. In the information processing device 1, the above restoration calculation can include a feature related to such a case in the explanatory variable.

In addition, the first document and the second document in the information processing device 1 are documents of the same type. This allows the information processing device 1 to identify the document to be corrected based on the relationship between documents of the same type.

(others)
Note that each component of each device shown in the figure does not necessarily have to be physically configured as shown in the figure. In other words, the specific form of distribution and integration of each device is not limited to that shown in the figure, and all or part of them can be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc. For example, for the information processing device 1, a configuration for generating a learning model 71 and a configuration for making a judgment based on the generated learning model 71 may be distributed.

Furthermore, the various processing functions of the information processing device 1 (the input unit 10, the objective variable generation unit 20, the explanatory variable generation unit 30, the learning unit 50, and the judgment unit 60) may be executed in whole or in part on a CPU (or a microcomputer such as an MPU or an MCU (Micro Controller Unit)). Needless to say, the various processing functions may be executed in whole or in part on a program that is analyzed and executed by the CPU (or a microcomputer such as an MPU or an MCU), or on hardware using wired logic. Furthermore, the various processing functions performed by the information processing device 1 may be executed by multiple computers working together using cloud computing.

(Example of computer configuration)
The various processes described in the above embodiment can be realized by executing a program prepared in advance on a computer. Therefore, an example of a computer configuration (hardware) that executes a program having the same functions as those of the above embodiment will be described below. Fig. 7 is a block diagram showing an example of the computer configuration.

As shown in FIG. 7, the computer 200 has a CPU 201 that executes various arithmetic processes, an input device 202 that accepts data input, a monitor 203, and a speaker 204. The computer 200 also has a medium reading device 205 that reads programs and the like from a storage medium, an interface device 206 for connecting to various devices, and a communication device 207 for connecting to external devices via wired or wireless communication. The information processing device 1 also has a RAM 208 that temporarily stores various information, and a hard disk device 209. Each unit (201 to 209) in the computer 200 is connected to a bus 210.

The hard disk drive 209 stores a program 211 for executing various processes in the functional configuration (e.g., the input unit 10, the objective variable generating unit 20, the explanatory variable generating unit 30, the learning unit 50, and the judgment unit 60) described in the above embodiment. The hard disk drive 209 also stores various data 212 referenced by the program 211. The input device 202, for example, accepts input of operation information from an operator. The monitor 203, for example, displays various screens operated by the operator. The interface device 206 is connected to, for example, a printing device. The communication device 207 is connected to a communication network such as a LAN (Local Area Network), and exchanges various information with external devices via the communication network.

The CPU 201 reads out the program 211 stored in the hard disk drive 209, expands it in the RAM 208, and executes it to perform various processes related to the above-mentioned functional configuration (for example, the input unit 10, the objective variable generating unit 20, the explanatory variable generating unit 30, the learning unit 50, and the judgment unit 60). The program 211 does not have to be stored in the hard disk drive 209. For example, the program 211 stored in a storage medium that the computer 200 can read out and execute may be read out and executed. Examples of storage media that the computer 200 can read out include portable storage media such as CD-ROMs, DVD disks, USB (Universal Serial Bus) memories, semiconductor memories such as flash memories, and hard disk drives. The program 211 may also be stored in a device connected to a public line, the Internet, a LAN, or the like, and the computer 200 may read out and execute the program 211 from the device.

The following notes are further provided with respect to the above embodiment.

(Supplementary Note 1) Based on learning data of a plurality of cases including the revision history of each document, a machine learning model is generated based on explanatory variables including a difference between a value calculated from a value of an item included in a specific first document and a value of an item included in a specific second document based on a predefined restoration calculation for each of the cases and the value of the item included in the first document, and a target variable including a value corresponding to the revision history of each document of the cases, and the explanatory variables generated based on each document to be judged are input to the model;
outputting whether or not each document to be judged has been modified based on the output from the model;
A document correction support program for causing a computer to execute a process.

(Note 2) The restoration operation is an operation for finding the maximum value of item values between different specific documents.
2. The document correction support program according to claim 1,

(Additional Note 3) The restoration operation is an operation for calculating the sum of values of items between different specific documents.
2. The document correction support program according to claim 1,

(Additional Note 4) The restoration operation is an operation of subtracting a value of an item of a specific third document included in either the first document or the second document from a total value of values of items between different specific documents.
2. The document correction support program according to claim 1,

(Supplementary Note 5) The first document and the second document are documents of the same type.
5. A document correction support program according to claim 1,

(Supplementary Note 6) Based on learning data of a plurality of cases including the revision history of each document, a machine learning model is generated based on explanatory variables including a difference between a value calculated from a value of an item included in a specific first document and a value of an item included in a specific second document based on a predefined restoration calculation for each of the cases and the value of the item included in the first document, and a target variable including a value corresponding to the revision history of each document of the cases. The explanatory variables generated based on each document to be judged are input to the model;
outputting whether or not each document to be judged has been modified based on the output from the model;
A document correction support method, comprising the steps of: executing a process on a computer;

(Additional Note 7) The restoration operation is an operation for finding the maximum value of item values between different specific documents.
7. The document correction support method according to claim 6,

(Additional Note 8) The restoration operation is an operation for calculating the sum of values of items between different specific documents.
7. The document correction support method according to claim 6,

(Supplementary Note 9) The restoration operation is an operation of subtracting a value of an item of a specific third document included in either the first document or the second document from a total value of values of items between different specific documents.
7. The document correction support method according to claim 6,

(Supplementary Note 10) The first document and the second document are documents of the same type.
10. The document correction support method according to claim 6,

(Supplementary Note 11) Based on learning data of a plurality of cases including the revision history of each document, a machine learning model is generated based on explanatory variables including a difference between a value calculated from a value of an item included in a specific first document and a value of an item included in a specific second document based on a predefined restoration calculation for each of the cases and the value of the item included in the first document, and a target variable including a value corresponding to the revision history of each document of the cases, and the explanatory variables generated based on each document to be judged are input to the model;
outputting whether or not each document to be judged has been modified based on the output from the model;
An information processing device characterized in that a control unit executes processing.

(Additional Note 12) The restoration operation is an operation for finding the maximum value of item values between different specific documents.
12. The information processing device according to claim 11.

(Additional Note 13) The restoration operation is an operation for calculating the sum of values of items between different specific documents.
12. The information processing device according to claim 11.

(Supplementary Note 14) The restoration operation is an operation of subtracting a value of an item of a specific third document included in either the first document or the second document from a total value of values of items between different specific documents.
12. The information processing device according to claim 11.

(Supplementary Note 15) The first document and the second document are documents of the same type.
15. The information processing device according to any one of claims 11 to 14.

1... Information processing device 10... Input unit 11... Corrected material data 12... Uncorrected material data 20... Objective variable generation unit 30... Explanatory variable generation unit 31... Restoration calculation calculation unit 32... Difference calculation unit 40... Storage unit 41... Restoration calculation definition information 50... Learning unit 60... Judgment unit 70... Learning data 71... Learning model 80... Evaluation data 81... Judgment result 200... Computer 201... CPU
202: Input device 203: Monitor 204: Speaker 205: Media reader 206: Interface device 207: Communication device 208: RAM
209...Hard disk device 210...Bus 211...Program 212...Various data D1...Final tax return D2...Resident tax return D3, D4...Salary payment report D5...Pension payment report H1...Resident H2...Employee K1, K2...Employee K3...Pension Service

Claims (7)

a machine learning model based on learning data of a plurality of cases including the revision history of each document, for each of the cases, an explanatory variable including a difference between a value calculated from a value of an item included in a specific first document and a value of an item included in a specific second document based on a predefined restoration calculation and the value of the item included in the first document, and a target variable including a value corresponding to the revision history of each document of the cases, the explanatory variables generated based on each document to be determined are input to the machine learning model;
outputting whether or not each document to be judged has been modified based on the output from the model;
A document correction support program for causing a computer to execute a process. The restoration operation is an operation for finding the maximum value of item values among different specific documents.
2. The document correction support program according to claim 1. The restoration operation is an operation for calculating a sum of values of items among different specific documents.
2. The document correction support program according to claim 1. The restoration operation is an operation of subtracting a value of an item of a specific third document included in either the first document or the second document from a total value of values of items among different specific documents.
2. The document correction support program according to claim 1. the first document and the second document are documents of the same type;
5. The document correction support program according to claim 1, a machine learning model based on learning data of a plurality of cases including the revision history of each document, for each of the cases, an explanatory variable including a difference between a value calculated from a value of an item included in a specific first document and a value of an item included in a specific second document based on a predefined restoration calculation and the value of the item included in the first document, and a target variable including a value corresponding to the revision history of each document of the cases, the explanatory variables generated based on each document to be determined are input to the machine learning model;
outputting whether or not each document to be judged has been modified based on the output from the model;
A document correction support method, comprising the steps of: executing a process on a computer; a machine learning model based on learning data of a plurality of cases including the revision history of each document, for each of the cases, an explanatory variable including a difference between a value calculated from a value of an item included in a specific first document and a value of an item included in a specific second document based on a predefined restoration calculation and the value of the item included in the first document, and a target variable including a value corresponding to the revision history of each document of the cases, the explanatory variables generated based on each document to be determined are input to the machine learning model;
outputting whether or not each document to be judged has been modified based on the output from the model;
An information processing device characterized in that a control unit executes processing.

Images