Learning device, learning method, accounting data conversion device, accounting data conversion method, and program

The learning device updates its neural network model with new layers using post-event data to maintain accuracy in accounting item inference, addressing the challenge of changes in business type or legal regulations.

JP2024170067APending Publication Date: 2024-12-06YAYOI CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2023087026
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-26
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Existing accounting systems face challenges in maintaining accuracy of journal entries due to changes in business type or legal regulations, leading to decreased accuracy rates in accounting items inferred by learning models.

Method used

A learning device that updates its neural network model by adding a new output layer and intermediate layer using post-event data to adapt to changes in business type or legal regulations, ensuring the model can accurately infer accounting items even after such changes occur.

Benefits of technology

The solution enhances the accuracy of journal entries by allowing the model to respond effectively to changes in business operations and legal environments, maintaining high accuracy rates despite shifts in journal entry trends.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024170067000001_ABST
    Figure 2024170067000001_ABST
Patent Text Reader

Abstract

To provide a learning device, a learning method, an accounting data conversion device, an accounting data conversion method, and a program.SOLUTION: In an accounting management system 10 in which a learning device 100, an accounting data conversion device 200, a business service providing device 20, an accounting service providing device 30, and a communication terminal 80 of a user 82 communicate with each other via a network 50, the learning device 100 comprises a learning execution unit which uses learning data including a summary of a transaction of a business operator after the occurrence of an event and an account title corresponding to the transaction to update a new output layer added in place of an output layer of a learning model which is generated or updated by using learning data including a summary of the transaction before the occurrence of the event and the account title corresponding to the transaction and includes an input layer, an intermediate layer, and the output layer to receive summaries and output account titles.SELECTED DRAWING: Figure 1
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a learning device, a learning method, an accounting data conversion device, an accounting data conversion method, and a program. [Background technology]

[0002] Patent Document 1 describes an accounting processing form change method / device that automatically changes the accounting processing form associated with a change in industry or corporate form by using accounting file data before the change. [Prior art document] [Patent documents] [Patent Document 1] Japanese Patent Application Laid-Open No. 10-091694 Summary of the Invention

[0003] According to one embodiment of the present invention, there is provided a learning device, which may include a learning execution unit that updates a new output layer added to replace an output layer of a learning model including an input layer, an intermediate layer, and an output layer, the output layer being generated or updated using learning data including summaries of transactions of an enterprise before an event occurs and account items corresponding to the transactions, using learning data including summaries of transactions and account items corresponding to the transactions after the event occurs.

[0004] The above summary of the invention does not list all of the necessary features of the present invention. Also, subcombinations of these features may also be inventions. [Brief description of the drawings]

[0005] [Figure 1] 1 illustrates a schematic diagram of an example of an accounting management system 10. [Diagram 2] 2 illustrates an example of a functional configuration of the learning device 100. [Diagram 3] 1 illustrates an example of a processing flow by the learning device 100. [Figure 4] FIG. 2 is an explanatory diagram for explaining generation of a learning model by the learning device 100. [Diagram 5] FIG. 2 is an explanatory diagram for explaining generation of a learning model by the learning device 100. [Figure 6] FIG. 2 is an explanatory diagram for explaining generation of a learning model by the learning device 100. [Figure 7] 2 illustrates an example of a functional configuration of an accounting data conversion device 200. [Figure 8] 13 shows an example of a processing flow by the accounting data conversion device 200. [Figure 9] An example of the hardware configuration of a computer 1200 that functions as the learning device 100 or the accounting data conversion device 200 is shown in schematic form. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0006] The present invention will be described below through embodiments of the invention, but the following embodiments do not limit the scope of the invention according to the claims. Furthermore, not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

[0007] FIG. 1 illustrates an example of an accounting management system 10. The accounting management system 10 includes a learning device 100 and an accounting data converting device 200. The learning device 100 and the accounting data converting device 200 may be different devices. The learning device 100 and the accounting data converting device 200 may be realized by a single device. The learning device 100 may be realized by multiple devices. The accounting data converting device 200 may be realized by multiple devices.

[0008] The accounting management system 10 may further include a business service providing device 20. The business service providing device 20 may be a device different from the learning device 100 and the accounting data converting device 200. The business service providing device 20 and the learning device 100 may be realized by a single device. The business service providing device 20 and the accounting data converting device 200 may be realized by a single device. The business service providing device 20, the learning device 100, and the accounting data converting device 200 may be realized by a single device.

[0009] The accounting management system 10 may further include an accounting service providing device 30. The accounting service providing device 30 may be a device different from the learning device 100 and the accounting data converting device 200. The accounting service providing device 30 and the learning device 100 may be realized by a single device. The accounting service providing device 30 and the accounting data converting device 200 may be realized by a single device. The accounting service providing device 30, the learning device 100, and the accounting data converting device 200 may be realized by a single device.

[0010] The learning device 100, the accounting data converting device 200, the business service providing device 20, and the accounting service providing device 30 may be realized by a single device. Here, a case will be mainly given as an example in which the learning device 100, the accounting data converting device 200, the business service providing device 20, and the accounting service providing device 30 are different devices.

[0011] The learning device 100, the accounting data conversion device 200, the business service providing device 20, and the accounting service providing device 30 communicate via a network 50. The network 50 may include a cloud network. The network 50 may include the Internet. The network 50 may include a mobile communication network. The network 50 may include a dedicated network such as a LAN (Local Area Network). The learning device 100 and the accounting data conversion device 200 may communicate directly without going through the network 50.

[0012] The business service providing device 20 manages transaction data of a user 82, who is the subject of a transaction. The user 82 is a business owner. The user 82 may be a sole proprietor or a corporation. The transaction data may include the transaction date. The transaction data may include the business partner. The transaction data may include the tax category. The transaction data may include the amount. The transaction data may include income / expense identification data indicating whether it is income or expenditure. The transaction data may include the amount of consumption tax. The transaction data may include a summary. The business service providing device 20 receives the transaction data of the user 82, for example, from the communication terminal 80 of the user 82.

[0013] The business service providing device 20 may communicate with a communication terminal 80 via a network 50. The communication terminal 80 may be any terminal capable of communication. For example, the communication terminal 80 is a mobile phone such as a smartphone, a tablet terminal, a PC (Personal Computer), etc. The business service providing device 20 may obtain transaction data of a user 82 from an accounting firm or the like that manages accounting information of the user 82.

[0014] The business service providing device 20 may provide the transaction data to the learning device 100. The learning device 100 may receive the transaction data of the user 82 from the business service providing device 20. The learning device 100 further acquires accounting data corresponding to the transaction data, which is generated by actually journalizing the transaction data of the user 82. The accounting data includes a summary of the transaction and an account item corresponding to the transaction.

[0015] The learning device 100 generates a learning model for performing accounting entries by executing machine learning using the acquired data. Conventionally, a learning model for inferring an account item corresponding to a transaction from a transaction summary has been generated by adjusting parameters of a natural language processing model using a neural network such as BERT, using a transaction summary and an account item corresponding to the transaction as learning data. A neural network is a combination of mathematical models of neurons in the human brain and nervous system, and is composed of three layers: an input layer, an output layer, and an intermediate layer (also called a hidden layer, a layer between the input layer and the output layer).

[0016] By generating a learning model using learning data including a summary of a business's transactions and the account items corresponding to the transactions, it is possible to reduce the burden of journal entry on the business and automate journal entry. However, if the business's industry changes or laws and regulations are revised after the learning model is generated, the tendency of journal entry may change, in which case the accuracy rate of the account items inferred by the learning model may decrease. Note that the correct account item is the appropriate account item for the transaction, and the accuracy rate is the probability of selecting the appropriate account item.

[0017] As a specific example, if a business operator changes its industry and an item that was purchased for its own use before the change needs to be purchased for sale after the change, the correct account item will change from "consumables expenses" or "supplies" to "purchases," and the accuracy rate of the learning model's account items will decrease. Also, for example, if a reduced tax rate is applied due to a revision of laws and regulations, the correct tax rate category before and after the application may differ.

[0018] On the other hand, even if the business sector changes or laws and regulations are revised, accounting entry trends may not change. Therefore, if training data from before these times is thinned out or the learning rate is lowered, the accuracy rate of accounting entries may decrease.

[0019] In response to this, the learning device 100 adds a new output layer to replace the output layer of the natural language processing model configured with a neural network that was trained with training data before an event such as a change in the company's industry or a revision of laws and regulations occurred, and trains with training data after the event occurred. This makes it easier to respond to changes in trend, since if the accounting trend changes due to an event, the output layer can be trained using only training data that reflects the trend change. Also, if the accounting trend does not change, inference is performed mainly on the layer trained with the training data before the event occurred, so the accuracy rate is less likely to be low compared to the original learning model.

[0020] Furthermore, the learning device 100 adds a new intermediate layer and a new output layer to replace the output layer of the natural language processing model configured with a neural network that has been trained with training data before the event occurs, and performs training with training data after the event occurs. This allows the added intermediate layer and output layer to be trained using only training data that reflects the trend change when an event causes a change in accounting trend, making it easier to respond to the trend change.

[0021] The accounting data conversion device 200 receives the learning model generated by the learning device 100 from the learning device 100 and stores it. The accounting data conversion device 200 receives transaction data of the user 82 from the business service providing device 20.

[0022] The accounting data conversion device 200 generates accounting data corresponding to the transaction data received from the business service providing device 20, using the learning model received from and stored in the learning device 100. The accounting data conversion device 200 may input summaries included in the transaction data to the learning model, and generate accounting data including the account items output from the learning model.

[0023] The accounting data conversion device 200 stores a learning model (sometimes referred to as a pre-event learning model) including an input layer, an intermediate layer, and an output layer, which is generated or updated using learning data including a summary of a transaction of an enterprise before a certain event occurs and an account item corresponding to the transaction, and which inputs a summary and outputs an account item. The accounting data conversion device 200 also stores a learning model (sometimes referred to as a post-event learning model) generated by updating a new layer (output layer, or intermediate layer and output layer) added in place of the output layer of the pre-event learning model using learning data including a summary of a transaction after the event occurs and an account item corresponding to the transaction. Then, the accounting data conversion device 200 inputs the summary included in the transaction data into the pre-event learning model for transaction data before the event occurs, and inputs the summary included in the transaction data into the post-event learning model for transaction data after the event occurs, selects the output account item, and generates accounting data including the selected account item. The accounting data conversion device 200 may output the generated accounting data to the accounting service providing device 30.

[0024] The accounting service providing device 30 manages the accounting data obtained from the accounting data conversion device 200. The accounting data conversion device 200 provides accounting services to the user 82 using the user's 82 accounting data.

[0025] 2 illustrates an example of the functional configuration of the learning device 100. The learning device 100 includes a data acquisition unit 102, a data storage unit 104, a learning execution unit 106, an event detection unit 108, and a learning model transmission unit 110. Note that it is not essential that the learning device 100 includes all of these units.

[0026] The data acquiring unit 102 acquires various data. The data acquiring unit 102 acquires transaction data of the user 82. The data acquiring unit 102 may acquire the transaction data from the business service providing device 20.

[0027] The data acquiring unit 102 acquires accounting data corresponding to the transaction data, including account items corresponding to the transactions, which is generated by actually journalizing the transaction data of the user 82. The data acquiring unit 102 may acquire the accounting data from the business service providing device 20. The data acquiring unit 102 may also acquire the accounting data from a source other than the business service providing device 20.

[0028] The data acquiring unit 102 stores the learning data generated from the acquired data in the data storage unit 104. The learning data includes a summary of the transaction and an account item corresponding to the transaction.

[0029] The learning execution unit 106 uses the learning data stored in the data storage unit 104 to generate a learning model including an input layer, an intermediate layer, and an output layer, in which abstracts are input and account items are output. The learning execution unit 106 may generate a learning model including an input layer, an intermediate layer, and an output layer by using the learning data to update parameters of a natural language processing model including the input layer, the intermediate layer, and the output layer through machine learning. The number of intermediate layers of the learning model generated by the learning execution unit 106 may be one or more.

[0030] In a state where no particular events that may affect the business operator's accounting entries have occurred (sometimes referred to as the normal state), the learning execution unit 106 may generate a learning model using all of the learning data stored in the data storage unit 104.

[0031] Examples of events that may affect accounting entries by a business operator include a change in the business operator's industry, amendments to laws and regulations, a change in status of a business operator that adopts the tax-excluded accounting method from a tax-exempt business operator to a taxable business operator, and a change in status of a business operator that adopts the tax-excluded accounting method from a taxable business operator to a tax-exempt business operator, etc. Note that these are examples and are not limited to these.

[0032] After an event that may affect the accounting by the business operator occurs, the learning execution unit 106 may update the new output layer added in place of the output layer of the learning model including the input layer, intermediate layer, and output layer generated or updated using the learning data before the event occurs, using the learning data after the event occurs. The learning execution unit 106 may generate or update the learning model using the learning data before the event occurs. Another device may generate or update the learning model using the learning data before the event occurs. The learning execution unit 106 may add a new output layer in place of the output layer of the learning model generated or updated using the learning data before the event occurs. That is, the learning execution unit 106 may add a new output layer in place of the output layer of the learning model, and then update the new output layer using the learning data after the event occurs. Another device may add a new output layer in place of the output layer of the learning model generated or updated using the learning data before the event occurs.

[0033] After an event that may affect the accounting entry by the business operator occurs, the learning execution unit 106 may update the new intermediate layer and the new output layer added in place of the output layer of the learning model including the input layer, intermediate layer, and output layer generated or updated using the learning data before the event occurs, using the learning data after the event occurs. It may be the learning execution unit 106 that adds the new intermediate layer and the new output layer in place of the output layer of the learning model generated or updated using the learning data before the event occurs. That is, the learning execution unit 106 may add the new intermediate layer and the new output layer in place of the output layer of the learning model, and then update the new intermediate layer and the new output layer using the learning data after the event occurs. It may be another device that adds the new intermediate layer and the new output layer in place of the output layer of the learning model generated or updated using the learning data before the event occurs.

[0034] If an event that may affect the business operator's accounting entries occurs multiple times, the learning execution unit 106 may generate a learning model to match them.

[0035] For each of a number of events that occur at different times, the learning execution unit 106 may update a new output layer that has been added in place of the output layer of the learning model that was generated or updated using the learning data before the event occurred, using the learning data after the event occurred.

[0036] Here, an example will be described in which a first event occurs, and then a second event occurs some time later.

[0037] The learning execution unit 106 generates a first learning model including a first input layer, a first hidden layer, and a first output layer using first learning data including a summary of the business operator's transaction before the occurrence of the first event and an account item corresponding to the transaction. The timing of generating the first learning model may be any timing after the occurrence of the first event. The timing of generating the first learning model may be before the occurrence of the first event. Note that the first learning model may be generated by another device. The first learning model may be generated by updating an existing learning model, or may be generated without using an existing learning model.

[0038] The learning execution unit 106 may add a new second intermediate layer and a new second output layer in place of the first output layer of the first learning model, and then update the second intermediate layer and the second output layer of the first learning model using second learning data including transaction summaries and account items corresponding to the transactions from the occurrence of the first event to the occurrence of the second event (the updated first learning model may be referred to as the second learning model). The learning execution unit 106 may add a new third intermediate layer and a new third output layer in place of the second output layer of the second learning model, and then update the third intermediate layer and the third output layer of the second learning model using third learning data including transaction summaries and account items corresponding to the transactions after the occurrence of the second event (the updated second learning model may be referred to as the third learning model). Note that adding a new intermediate layer and a new output layer in place of the output layer of the learning model may be performed by another device.

[0039] The learning execution unit 106 may execute learning as follows. The learning execution unit 106 adds a new second output layer in place of the first output layer of the first learning model, and then updates the second output layer of the first learning model using the second learning data. The learning execution unit 106 adds a new third output layer in place of the second output layer of the second learning model, and then updates the third output layer of the second learning model using the third learning data. The addition of a new output layer in place of the output layer of the learning model may be performed by another device.

[0040] Next, an example will be described in which a first event occurs, a second event occurs some time later, and a third event occurs some time later.

[0041] The learning execution unit 106 generates a first learning model including a first input layer, a first hidden layer, and a first output layer using first learning data including a summary of the business operator's transaction before the occurrence of the first event and an account item corresponding to the transaction. The timing of generating the first learning model may be any timing after the occurrence of the first event. The timing of generating the first learning model may be before the occurrence of the first event. Note that the first learning model may be generated by another device. The first learning model may be generated by updating an existing learning model, or may be generated without using an existing learning model.

[0042] The learning execution unit 106 may add a new second hidden layer and a new second output layer in place of the first output layer of the first learning model, and then update the second hidden layer and the second output layer of the first learning model using second learning data including transaction summaries and account items corresponding to the transactions from the occurrence of the first event to the occurrence of the second event (the updated first learning model may be referred to as the second learning model). The learning execution unit 106 may add a new third hidden layer and a new third output layer in place of the second output layer of the second learning model, and then update the third hidden layer and the third output layer of the second learning model using third learning data including transaction summaries and account items corresponding to the transactions after the occurrence of the second event (the updated second learning model may be referred to as the third learning model). The learning execution unit 106 may add a new fourth intermediate layer and a new fourth output layer in place of the third output layer of the third learning model, and then update the fourth intermediate layer and the fourth output layer of the third learning model using fourth learning data including a transaction summary and an account item corresponding to the transaction after the third event occurs (the updated third learning model may be referred to as a fourth learning model). Note that adding a new intermediate layer and a new output layer in place of the output layer of the learning model may be performed by another device.

[0043] The learning execution unit 106 may execute learning as follows. The learning execution unit 106 adds a new second output layer in place of the first output layer of the first learning model, and then updates the second output layer of the first learning model using the second learning data. The learning execution unit 106 adds a new third output layer in place of the second output layer of the second learning model, and then updates the third output layer of the second learning model using the third learning data. The learning execution unit 106 adds a new fourth output layer in place of the third output layer of the third learning model, and then updates the fourth output layer of the third learning model using the fourth learning data. The addition of a new output layer in place of the output layer of the learning model may be performed by another device.

[0044] The learning execution unit 106 may execute learning in the same manner even when an event occurs four or more times.

[0045] As described above, when an event occurs, a new layer (an intermediate layer and an output layer, or only an output layer) is added in place of the output layer of the learning model generated using learning data before the event occurred, and the new layer is updated using learning data after the event occurred. This makes it possible to generate a learning model in which the original layer corresponds to transaction contents that are not affected by the event, and the new layer corresponds to transaction contents that are affected by the event, which can contribute to achieving a high accuracy rate.

[0046] When an event occurs, the learning execution unit 106 may generate a new learning model by updating all layers of the learning model generated using the learning data before the event occurs using the learning data after the event occurs.Which method is more appropriate, updating only the new layers or updating all layers, may vary depending on the changes in the accounting pattern and the amount and quality of the actually accumulated learning data.

[0047] The learning execution unit 106 may receive instructions from the user 82, an administrator of the accounting management system 10, or the like regarding whether to use a method of updating only new layers or a method of updating all layers. The learning execution unit 106 may select either a method of updating only new layers or a method of updating all layers in accordance with instructions from the user 82, an administrator of the accounting management system 10, or the like. In this way, if the user 82 or an administrator of the accounting management system 10, or the like, has knowledge about which method is more appropriate to use, a learning model can be generated in accordance with that knowledge.

[0048] The learning execution unit 106 may generate a learning model using a method of updating only a new layer and generate a learning model using a method of updating all layers, and select which one to adopt by evaluating both. For example, when an event occurs, the learning execution unit 106 generates a first learning model by adding a new layer in place of the output layer of the learning model generated or updated using the learning data before the event occurs and updating the new layer using the learning data after the event occurs, and generates a second learning model by updating all layers of the learning model generated or updated using the learning data before the event occurs using the learning data after the event occurs.

[0049] Then, the learning execution unit 106 evaluates the first learning model and the second learning model using evaluation data including transaction summaries and account items corresponding to the transactions. Specifically, the learning execution unit 106 inputs the transaction summaries included in the evaluation data into the first learning model and the second learning model, and determines whether the account items output from the first learning model and the second learning model match the account items included in the evaluation data. The learning execution unit 106 performs such a determination multiple times to calculate the accuracy rate of each of the first learning model and the second learning model, and selects the one with the higher accuracy rate. This allows the method that can actually achieve a high accuracy rate to be adopted from the method of updating only the new layer and the method of updating all layers.

[0050] The learning execution unit 106 may use data that is not used in learning as the evaluation data. This allows the generalization performance of the first learning model and the second learning model to be evaluated. Note that it is not essential to use data that is not used in learning, and the learning execution unit 106 may use data that is used in learning as the evaluation data.

[0051] The learning execution unit 106 may use data after the timing at which the event occurred as the evaluation data. What is particularly desired to be determined is whether correct classification can be performed using data after the timing at which the event occurred, and it can be said that by performing evaluation using only data after the timing at which the event occurred, it is highly likely that a determination that meets the purpose will be made. Note that it is not essential to use data after the timing at which the event occurred, and the learning execution unit 106 may use data before and after the timing at which the event occurred as the evaluation data.

[0052] The event detection unit 108 detects the occurrence of an event that may affect the accounting entry by the business operator. Before the event detection unit 108 detects an event, the learning execution unit 106 may generate a learning model using all the learning data stored in the data storage unit 104. After the event detection unit 108 detects an event, the learning execution unit 106 may generate a post-event learning model by adding a new intermediate layer and a new output layer in place of the output layer of the pre-event learning model generated or updated using the learning data before the event occurs, and updating the new intermediate layer and the new output layer of the pre-event learning model using the learning data after the event occurs. After the event detection unit 108 detects an event, the learning execution unit 106 adds a new output layer in place of the output layer of the pre-event learning model generated or updated using the learning data before the event occurs, and updates the new output layer of the pre-event learning model using the learning data after the event occurs.

[0053] The event detection unit 108 detects that an event has occurred, for example, based on a change in the registered information in an accounting management system that manages the accounting data of the business. This makes it possible to automatically detect the occurrence of an event that may affect accounting entries made by the business. The accounting service providing device 30 may be an example of an accounting management system. In other words, the event detection unit 108 may detect that an event has occurred based on a change in the registered information in the accounting service providing device 30.

[0054] For example, if the accounting management system registers law information and that law information is changed, the event detection unit 108 detects that a law change event has occurred.

[0055] For example, if the accounting management system registers information on the industry of a user 82 as basic information for the user 82, and the industry information is changed, the event detection unit 108 detects that an industry change event has occurred for the user 82.

[0056] For example, in the case where the accounting management system registers the business classification of user 82 as basic information of user 82, when the business classification is changed from a tax-exempt business to a taxable business, or from a taxable business to a tax-exempt business, the event detection unit 108 detects that an event of change from a tax-exempt business to a taxable business or from a taxable business to a tax-exempt business has occurred. In addition, the event detection unit 108 may detect that an event of change from a tax-exempt business to a taxable business or from a taxable business to a tax-exempt business has occurred depending on whether the total sales during the accounting period exceeds 10 million yen.

[0057] The learning execution unit 106 may execute learning according to the type of the event that has occurred. For example, the learning execution unit 106 adds new intermediate layers and a new output layer in the number corresponding to the type of event, in place of the output layer of the pre-event learning model that has been generated or updated using learning data including summaries of transactions of the business entity before the event occurs and account items corresponding to the transactions, and then updates the new intermediate layers and new output layers of the pre-event learning model using learning data including summaries of transactions after the event occurs and account items corresponding to the transactions.

[0058] Depending on the type of event, the accounting may change simply or may change in a complicated manner. As a specific example, taking a change in the business type of a business operator, a revision of laws and regulations, a change from a taxable business operator to a tax-exempt business operator, and a change from a tax-exempt business operator to a taxable business operator, the accounting may change in a more complicated manner in the order of a change from a taxable business operator to a tax-exempt business operator, a change in business type, and a revision of laws and regulations. Therefore, the learning execution unit 106 adds more new intermediate layers in the order of the event that occurs, that is, a change from a taxable business operator to a tax-exempt business operator, a change in business type, and a revision of laws and regulations. As a specific example, when the event that occurs is a change from a taxable business operator to a tax-exempt business operator and a change from a tax-exempt business operator to a taxable business operator, the learning execution unit 106 adds one new intermediate layer and one new output layer to the pre-event learning model from which the output layer has been removed. When the event that occurs is a change in business type, the learning execution unit 106 adds two new intermediate layers and one new output layer to the pre-event learning model from which the output layer has been removed. When the event that has occurred is a revision of laws and regulations, the learning execution unit 106 adds three new hidden layers and one new output layer to the pre-event learning model from which the output layer has been removed. This makes it possible to prevent a situation in which learning is insufficient due to a lack of hidden layers even though the accounting entry has changed in a complex manner, or to prevent unnecessary learning from being performed due to a large number of hidden layers even though the accounting entry has only changed simply.

[0059] The learning execution unit 106 may execute learning according to the content of the event that has occurred. For example, the learning execution unit 106 adds new intermediate layers and a new output layer in the number corresponding to the content of the event, in place of the output layer of the pre-event learning model that has been generated or updated using learning data including summaries of transactions of the business entity and account items corresponding to the transactions before the event occurs, and then updates the new intermediate layers and new output layers of the pre-event learning model using learning data including summaries of transactions and account items corresponding to the transactions after the event occurs.

[0060] Depending on the contents of the event, the journal entry may change simply or may change in a complicated manner. As a specific example, when the event that has occurred is a change in industry, the journal entry may change more complicatedly when the change is complicated, compared to when the change is simple. Therefore, the learning execution unit 106 stores in advance correspondence data that associates the contents of the industry change with the number of intermediate layers, and adds an intermediate layer of a number corresponding to the contents of the industry change. Also, as a specific example, when the event that has occurred is a revision of laws and regulations, the journal entry may change more complicatedly as the impact of the revision of laws and regulations is greater. Therefore, when the event that has occurred is a revision of laws and regulations, the learning execution unit 106 further acquires information on the impact of the revision of laws and regulations, and adds a larger number of intermediate layers as the impact is greater. Information on the impact of the revision of laws and regulations may be registered by the administrator of the accounting management system, the administrator of the accounting management system 10, the user 82, and the like.

[0061] 3 shows an example of a processing flow by the learning execution unit 106. Here, an example is described in which an event that may affect accounting by a business entity has occurred once, a certain period of time has passed since the event occurred, and learning data after the event occurrence has been stored in the data storage unit 104 to a certain extent.

[0062] In step (sometimes abbreviated to S) 102, the learning execution unit 106 acquires learning data before the occurrence of an event from the data storage unit 104. In S104, the learning execution unit 106 updates parameters of the learning model using the learning data acquired in S102.

[0063] In S106, the learning execution unit 106 acquires learning data after the occurrence of the event from the data storage unit 104. In S108, the learning execution unit 106 adds an intermediate layer and an output layer in place of the output layer of the learning model whose parameters have been updated in S104, and updates the parameters of the intermediate layer and the output layer added to the learning model using the learning data acquired in S106.

[0064] 4 is an explanatory diagram for explaining generation of a learning model by the learning device 100. Here, generation of a learning model in a case where an event 41, an event 42, and an event 43 occur in this order from a normal state will be explained. Note that, although an example is given here in which a new intermediate layer and a new output layer are added in place of the output layer, a new output layer may be added in place of the output layer.

[0065] The learning execution unit 106 generates a second learning model by adding a new second intermediate layer and a new second output layer in place of the first output layer of the first learning model including the first input layer, the first intermediate layer, and the first output layer generated or updated using the learning data 501 of the period 401 before the occurrence of the event 41, and then updating the second intermediate layer and the second output layer of the first learning model using the learning data 502 of the period 402 from the occurrence of the event 41 to the occurrence of the event 42. Note that the generation or update of the first learning model may be performed by another device. The learning execution unit 106 generates a third learning model by adding a new third intermediate layer and a new third output layer in place of the second output layer of the second learning model, and then updating the third intermediate layer and the third output layer of the second learning model using the learning data 503 of the period 403 from the occurrence of the event 42 to the occurrence of the event 43. The learning execution unit 106 generates a fourth learning model by adding a new fourth hidden layer and a new fourth output layer in place of the third output layer of the third learning model, and then updating the fourth hidden layer and the fourth output layer of the third learning model using learning data 504 for a period 404 after the occurrence of event 43. Note that the process of adding a new hidden layer and a new output layer in place of the output layer of the learning model may be executed by another device.

[0066] As described above, the learning device 100 may generate a learning model by adding an intermediate layer and an output layer in place of the output layer of the learning model generated or updated using the learning data before the event occurs, and updating the added intermediate layer and output layer using the learning data after the event occurs. This can contribute to solving problems when there is still little training data after the tendency of accounting changes. In contrast, when a long period of time has passed since the event occurred and sufficient data has been accumulated, the learning execution unit 106 may generate one learning model using all the learning data. If such a response is not taken, the number of layers of the learning model increases every time an event occurs, which increases the cost of learning, whereas by taking such a response by the learning execution unit 106, it may be possible to appropriately reduce the cost of learning without decreasing the accuracy rate.

[0067] For example, after generating the first learning model, the second learning model, the third learning model, and the fourth learning model, the learning execution unit 106 generates a learning model by updating the parameters of a new learning model having an input layer, an intermediate layer, and an output layer using all the learning data. This makes it possible to eliminate all the added layers and reduce the learning cost thereafter. Note that, without being limited to this, the learning execution unit 106 may eliminate some of the added layers. For example, the learning execution unit 106 may update a new intermediate layer and a new output layer added in place of the output layer of the learning model generated or updated using the learning data before the event occurs for each of a plurality of events that occurred at different timings, using the learning data after the event occurs, and then update a new intermediate layer and the new output layer added in place of the output layer of the learning model generated or updated using the learning data before the event occurs for each of some of the plurality of events, using the learning data after the event occurs.

[0068] Fig. 5 is an explanatory diagram for explaining generation of a learning model by the learning device 100. Here, a process will be described in which an added layer corresponding to an event 41 is eliminated after the first learning model, the second learning model, the third learning model, and the fourth learning model are generated by the process shown in Fig. 4. Note that, although an example is given here in which a new intermediate layer and a new output layer are added in place of the output layer, a new output layer may also be added in place of the output layer.

[0069] The learning execution unit 106 generates a sixth learning model by adding a new sixth intermediate layer and a new sixth output layer in place of the fifth output layer of the fifth learning model including the fifth input layer, the fifth intermediate layer, and the fifth output layer generated using the learning data 501 and the learning data 502, and then updating the sixth intermediate layer and the sixth output layer of the fifth learning model using the learning data 503. The learning execution unit 106 generates a seventh learning model by adding a new seventh intermediate layer and a new seventh output layer in place of the sixth output layer of the sixth learning model, and then updating the seventh intermediate layer and the seventh output layer of the sixth learning model using the learning data 504. In this way, for each of the events 42 and 43, which are part of the events 41, 42, and 43, the learning execution unit 106 may add a new intermediate layer and a new output layer in place of the output layer of the learning model generated using the learning data before the event occurs, and update the added new intermediate layer and new output layer using the learning data after the event occurs.

[0070] Fig. 6 is an explanatory diagram for explaining the generation of a learning model by the learning device 100. Here, a process is described for the case where the additional layer corresponding to the event 42 is abolished after the first learning model, the second learning model, the third learning model, and the fourth learning model are generated by the process shown in Fig. 4.

[0071] The learning execution unit 106 generates a 9th learning model by adding a new 9th intermediate layer and a new 9th output layer in place of the 8th output layer of the 8th learning model including the 8th input layer, the 8th intermediate layer, and the 8th output layer generated using the learning data 501, and then updating the 9th intermediate layer and the 9th output layer of the 8th learning model using the learning data 502 and the learning data 503. The learning execution unit 106 generates a 10th learning model by adding a new 10th intermediate layer and a new 10th output layer in place of the 9th output layer of the 9th learning model, and then updating the 10th intermediate layer and the 10th output layer of the 9th learning model using the learning data 504. In this way, for each event 41 and event 43, which are part of the events 41, 42, and 43, the learning execution unit 106 may add a new intermediate layer and a new output layer in place of the output layer of the learning model generated using the learning data before the event occurs, and update the added new intermediate layer and new output layer using the learning data after the event occurs.

[0072] 7 illustrates an example of the functional configuration of the accounting data conversion device 200. The accounting data conversion device 200 includes a learning model acquisition unit 202, a learning model storage unit 204, a transaction data acquisition unit 206, an item selection unit 208, an accounting data generation unit 210, and an accounting data output unit 212.

[0073] The learning model acquisition unit 202 acquires a learning model that takes a summary as input and an account title as output. The learning model storage unit 204 stores the learning model acquired by the learning model acquisition unit 202.

[0074] The learning model acquisition unit 202 may acquire the learning model from the learning device 100. When the pre-event learning model and the post-event learning model are generated by the learning device 100, the learning model acquisition unit 202 may acquire the pre-event learning model and the post-event learning model from the learning device 100.

[0075] When the learning device 100 generates learning models corresponding to multiple events, the learning model acquisition unit 202 may acquire a learning model corresponding to a normal state and a learning model corresponding to a final event. As a specific example, when the learning device 100 generates a first learning model, a second learning model, and a third learning model, the learning model acquisition unit 202 may acquire the first learning model as a pre-event learning model and the third learning model as a post-event learning model. When the learning device 100 generates a first learning model, a second learning model, a third learning model, and a fourth learning model, the learning model acquisition unit 202 may acquire the first learning model as a pre-event learning model and the fourth learning model as a post-event learning model. When the learning device 100 generates one learning model using all the learning data, the learning model acquisition unit 202 may acquire the learning model. Note that the learning model acquisition unit 202 may acquire a learning model from a source other than the learning device 100.

[0076] The transaction data acquisition unit 206 acquires transaction data. The transaction data acquisition unit 206 may acquire transaction data from which an account item is to be inferred. The transaction data acquisition unit 206 acquires transaction data, for example, from the business service providing device 20. The transaction data acquisition unit 206 acquires transaction data, for example, from the communication terminal 80.

[0077] The account selection unit 208 selects an account item corresponding to the transaction indicated by the transaction data acquired by the transaction data acquisition unit 206. The account selection unit 208 may select an account item corresponding to the transaction data based on a summary included in the transaction data acquired by the transaction data acquisition unit 206. The account selection unit 208 may input the summary included in the transaction data acquired by the transaction data acquisition unit 206 into a learning model stored in the learning model storage unit 204, and select an account item output from the learning model.

[0078] When an event occurs that may affect the business's accounting entries and the learning device 100 generates a post-event learning model, the account selection unit 208 may input the summary included in the transaction data, if the transaction data acquisition unit 206 acquires transaction data before the event occurs, into the pre-event learning model and select the account item output by the pre-event learning model, and when the transaction data acquisition unit 206 acquires transaction data after the event occurs, input the summary included in the transaction data into the post-event learning model and select the account item output by the post-event learning model.

[0079] When the learning device 100 generates a single learning model using all the learning data, the account selection unit 208 may input the summary contained in the transaction data acquired by the transaction data acquisition unit 206 into the learning model and select the account item output by the learning model.

[0080] The accounting data generation unit 210 generates accounting data corresponding to the transaction data acquired by the transaction data acquisition unit 206. The accounting data generation unit 210 generates accounting data including the account item selected by the item selection unit 208.

[0081] The accounting data output unit 212 outputs the accounting data generated by the accounting data generation unit 210. The accounting data output unit 212 may output the accounting data to the accounting service providing device 30. The accounting data output unit 212 may output the accounting data to the communications terminal 80.

[0082] As described above, the learning device 100 and the accounting data conversion device 200 may be realized by a single device. In this case, the learning device 100 may further include a learning model acquisition unit 202, a learning model storage unit 204, a transaction data acquisition unit 206, an item selection unit 208, an accounting data generation unit 210, and an accounting data output unit 212.

[0083] 8 shows an example of the flow of processing by the accounting data conversion device 200. Here, the flow of processing will be described when generating accounting data corresponding to each of a plurality of transaction data.

[0084] In S202, the transaction data acquisition unit 206 acquires transaction data. If the transaction data acquired in S202 is transaction data after the occurrence of the event (YES in S204), the process proceeds to S206, and if the transaction data is transaction data before the occurrence of the event (NO in S204), the process proceeds to S208.

[0085] In S206, the account selection unit 208 selects an account item using the post-event learning model. The account selection unit 208 inputs the summary included in the transaction data into the post-event learning model and selects the account item output by the post-event learning model.

[0086] In S208, the item selection unit 208 selects an account item using the pre-event learning model. The item selection unit 208 inputs the summary included in the transaction data into the pre-event learning model and selects the account item output by the pre-event learning model.

[0087] In S210, the accounting data generation unit 210 generates accounting data corresponding to the transaction data acquired by the transaction data acquisition unit 206 in S202. The accounting data generation unit 210 generates accounting data including the account item selected by the account selection unit 208 in S206 or S208.

[0088] If generation of accounting data for all target transaction data has been completed (YES in S212), proceed to S214, otherwise return to S202. In S214, the accounting data output unit 212 outputs the generated accounting data to the accounting service providing device 30.

[0089] 9 is a schematic diagram showing an example of a hardware configuration of a computer 1200 functioning as the learning device 100 or the accounting data conversion device 200. A program installed in the computer 1200 can cause the computer 1200 to function as one or more "parts" of the device according to the present embodiment, or cause the computer 1200 to execute operations or one or more "parts" associated with the device according to the present embodiment, and / or cause the computer 1200 to execute a process or a step of the process according to the present embodiment. Such a program may be executed by the CPU 1212 to cause the computer 1200 to execute specific operations associated with some or all of the blocks of the flowcharts and block diagrams described herein.

[0090] The computer 1200 according to this embodiment includes a CPU 1212, a RAM 1214, and a graphics controller 1216, which are connected to each other by a host controller 1210. The computer 1200 also includes input / output units such as a communication interface 1222, a storage device 1224, a DVD drive 1226, and an IC card drive, which are connected to the host controller 1210 via an input / output controller 1220. The DVD drive 1226 may be a DVD-ROM drive, a DVD-RAM drive, or the like. The storage device 1224 may be a hard disk drive, a solid state drive, or the like. The computer 1200 also includes a legacy input / output unit such as a ROM 1230 and a keyboard, which are connected to the input / output controller 1220 via an input / output chip 1240.

[0091] The CPU 1212 operates according to a program stored in the ROM 1230 and the RAM 1214, thereby controlling each unit. The graphic controller 1216 acquires image data generated by the CPU 1212 into a frame buffer or the like provided in the RAM 1214 or into itself, and causes the image data to be displayed on the display device 1218.

[0092] The communication interface 1222 communicates with other electronic devices via a network. The storage device 1224 stores programs and data used by the CPU 1212 in the computer 1200. The DVD drive 1226 reads programs or data from a DVD-ROM 1227 or the like and provides the programs or data to the storage device 1224. The IC card drive reads programs and data from an IC card and / or writes programs and data to an IC card.

[0093] The ROM 1230 stores therein a boot program or the like that is executed by the computer 1200 upon activation, and / or a program that depends on the hardware of the computer 1200. The input / output chip 1240 may also connect various input / output units to the input / output controller 1220 via a USB port, a parallel port, a serial port, a keyboard port, a mouse port, and the like.

[0094] The programs are provided by a computer-readable storage medium such as a DVD-ROM 1227 or an IC card. The programs are read from the computer-readable storage medium, installed in the storage device 1224, the RAM 1214, or the ROM 1230, which are also examples of computer-readable storage media, and executed by the CPU 1212. Information processing described in these programs is read by the computer 1200, and brings about cooperation between the programs and the various types of hardware resources described above. An apparatus or method may be constructed by implementing operations or processing of information according to the use of the computer 1200.

[0095] For example, when communication is performed between the computer 1200 and an external device, the CPU 1212 may execute a communication program loaded in the RAM 1214 and instruct the communication interface 1222 to perform communication processing based on the processing described in the communication program. Under the control of the CPU 1212, the communication interface 1222 reads transmission data stored in a transmission buffer area provided in the RAM 1214, the storage device 1224, the DVD-ROM 1227, or a recording medium such as an IC card, and transmits the read transmission data to a network, or writes reception data received from the network to a reception buffer area or the like provided on the recording medium.

[0096] Furthermore, the CPU 1212 may cause all or a necessary portion of a file or database stored in an external recording medium such as the storage device 1224, the DVD drive 1226 (DVD-ROM 1227), an IC card, etc. to be read into the RAM 1214, and perform various types of processing on the data on the RAM 1214. The CPU 1212 may then write back the processed data to the external recording medium.

[0097] Various types of information, such as various types of programs, data, tables, and databases, may be stored in the recording medium and undergo information processing. The CPU 1212 may perform various types of processing on the data read from the RAM 1214, including various types of operations, information processing, conditional judgment, conditional branching, unconditional branching, information search / replacement, etc., as described throughout this disclosure and specified by the instruction sequence of the program, and write back the results to the RAM 1214. The CPU 1212 may also search for information in a file, database, etc. in the recording medium. For example, when a plurality of entries each having an attribute value of a first attribute associated with an attribute value of a second attribute are stored in the recording medium, the CPU 1212 may search for an entry whose attribute value of the first attribute matches a specified condition from among the plurality of entries, read the attribute value of the second attribute stored in the entry, and thereby obtain the attribute value of the second attribute associated with the first attribute that satisfies a predetermined condition.

[0098] The above-described programs or software modules may be stored in a computer-readable storage medium on the computer 1200 or in the vicinity of the computer 1200. In addition, a recording medium such as a hard disk or a RAM provided in a server system connected to a dedicated communication network or the Internet can be used as a computer-readable storage medium, thereby providing the programs to the computer 1200 via the network.

[0099] The blocks in the flowcharts and block diagrams in the present embodiment may represent stages of a process in which an operation is performed or "parts" of an apparatus responsible for performing the operation. Particular stages and "parts" may be implemented by dedicated circuitry, programmable circuitry provided with computer-readable instructions stored on a computer-readable storage medium, and / or a processor provided with computer-readable instructions stored on a computer-readable storage medium. The dedicated circuitry may include digital and / or analog hardware circuits, and may include integrated circuits (ICs) and / or discrete circuits. The programmable circuitry may include reconfigurable hardware circuits, such as, for example, field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), and the like, including AND, OR, XOR, NAND, NOR, and other logical operations, flip-flops, registers, and memory elements.

[0100] A computer-readable storage medium may include any tangible device capable of storing instructions that are executed by a suitable device, such that a computer-readable storage medium having instructions stored thereon comprises an article of manufacture that includes instructions that can be executed to create means for performing the operations specified in the flowcharts or block diagrams. Examples of computer-readable storage media may include electronic storage media, magnetic storage media, optical storage media, electromagnetic storage media, semiconductor storage media, and the like. More specific examples of computer-readable storage media may include floppy disks, diskettes, hard disks, random access memories (RAMs), read-only memories (ROMs), erasable programmable read-only memories (EPROMs or flash memories), electrically erasable programmable read-only memories (EEPROMs), static random access memories (SRAMs), compact disk read-only memories (CD-ROMs), digital versatile disks (DVDs), Blu-ray disks, memory sticks, integrated circuit cards, and the like.

[0101] The computer readable instructions may include either assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk®, JAVA®, C++, etc., and conventional procedural programming languages ​​such as the “C” programming language or similar programming languages.

[0102] Computer readable instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, or to a programmable circuit, either locally or over a local area network (LAN), a wide area network (WAN), such as the Internet, etc., to cause the processor of the general purpose computer, special purpose computer, or other programmable data processing apparatus, or to a programmable circuit, to execute the computer readable instructions to generate means for performing the operations specified in the flowcharts or block diagrams. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, etc.

[0103] Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments. It is clear to those skilled in the art that various modifications and improvements can be made to the above embodiments. It is clear from the description of the claims that such modifications and improvements can also be included in the technical scope of the present invention.

[0104] It should be noted that the order of execution of each process, such as operations, procedures, steps, and stages, in the devices, systems, programs, and methods shown in the claims, specifications, and drawings is not specifically stated as "before" or "prior to," and that the process may be performed in any order unless the output of a previous process is used in a later process. Even if the operational flow in the claims, specifications, and drawings is described using "first," "next," and the like for convenience, this does not mean that the process must be performed in this order. [Explanation of symbols]

[0105] 10 accounting management system, 20 business service providing device, 30 accounting service providing device, 41, 42, 43 event, 50 network, 80 communication terminal, 82 user, 100 learning device, 102 data acquisition unit, 104 data storage unit, 106 learning execution unit, 108 event detection unit, 110 learning model transmission unit, 200 accounting data conversion device, 202 learning model acquisition unit, 204 learning model storage unit, 206 transaction data acquisition unit, 208 subject selection unit, 210 accounting data generation unit, 212 accounting data output unit, 401, 402, 403, 404 period, 501, 502, 503, 504 learning data, 1200 computer, 1210 host controller, 1212 CPU, 1214 RAM, 1216 graphic controller, 1218 display device, 1220 I / O controller, 1222 communication interface, 1224 storage device, 1226 DVD drive, 1227 DVD-ROM, 1230 ROM, 1240 I / O chip

Claims

1. A learning execution unit that updates a new output layer added in place of an output layer of a learning model including an input layer, an intermediate layer, and an output layer, the output layer being generated or updated using learning data including summaries of transactions of an enterprise before an event occurs and account items corresponding to the transactions, using learning data including summaries and account items corresponding to the transactions after the event occurs. A learning device comprising:

2. 2. The learning device according to claim 1, wherein the learning execution unit adds the new output layer in place of the output layer of the learning model, and then updates the new output layer using the learning data including a summary of a transaction after the event occurs and an account item corresponding to the transaction.

3. The learning device according to claim 1 , wherein the learning execution unit updates a new intermediate layer and the new output layer added in place of the output layer of the learning model using the learning data including a summary of a transaction after the event occurs and an account item corresponding to the transaction.

4. 2. The learning device according to claim 1, wherein the learning execution unit updates, for each of a plurality of events occurring at different timings, a new intermediate layer and the new output layer added in place of the output layer of the learning model generated or updated using the learning data before the event occurs, using the learning data after the event occurs.

5. 5. The learning device according to claim 4, wherein the learning execution unit updates, for each of the plurality of events occurring at different timings, a new intermediate layer and the new output layer added in place of the output layer of the learning model generated or updated using the learning data before the event occurs, using the learning data after the event occurs, and then, for each of some of the plurality of events, updates the new intermediate layer and the new output layer added in place of the output layer of the learning model generated or updated using the learning data before the event occurs, using the learning data after the event occurs.

6. An event detection unit that detects the occurrence of the event that may affect the accounting entry by the business entity. Further equipped with 6. The learning device according to claim 3, wherein after the event is detected by the event detection unit, the learning execution unit updates the new intermediate layer and the new output layer using the learning data including a summary of a transaction after the event occurs and an account item corresponding to the transaction.

7. The learning device according to claim 6 , wherein the event detection unit detects the occurrence of the event based on a change in registered information in an accounting management system that manages accounting data of the business.

8. 7. The learning device according to claim 6, wherein the learning execution unit adds, in place of the output layer of the learning model, the new intermediate layer and the new output layer in a number corresponding to the type or content of the event, and then updates the new intermediate layer and the new output layer using the learning data including a summary of a transaction after the event occurs and an account item corresponding to the transaction.

9. a learning model storage unit that stores the learning model generated or updated using the learning data before the event occurs and a post-event learning model in which the new output layer is updated by the learning execution unit; a transaction data acquisition unit that acquires transaction data including a summary of the transaction; an account selection unit that, when the transaction data acquisition unit acquires the transaction data after the occurrence of the event, inputs the summary included in the transaction data into the post-event learning model and selects an account item output by the post-event learning model; an accounting data generation unit that generates accounting data corresponding to the transaction data, including the account item selected by the item selection unit; The learning device according to claim 1 , further comprising:

10. Computer, A learning execution unit that updates a new output layer added in place of an output layer of a learning model including an input layer, an intermediate layer, and an output layer, the output layer being generated or updated using learning data including summaries of transactions of an enterprise before an event occurs and account items corresponding to the transactions, using learning data including summaries and account items corresponding to the transactions after the event occurs. A program to function as a

11. 1. A computer-implemented method of learning, comprising: A learning execution step of updating a new output layer added in place of the output layer of the learning model including an input layer, an intermediate layer, and an output layer, which is generated or updated using learning data including summaries of transactions of the business operator before the event occurs and account items corresponding to the transactions, using learning data including summaries and account items corresponding to the transactions after the event occurs. A learning method that provides:

12. a learning model storage unit that stores a pre-event learning model including an input layer, an intermediate layer, and an output layer, which is generated or updated using learning data including summaries of transactions of business operators before an event occurs and account items corresponding to the transactions, and which takes summaries as input and account items as output, and a post-event learning model generated by updating a new output layer added in place of the output layer of the pre-event learning model using learning data including summaries of transactions after the event occurs and account items corresponding to the transactions; a transaction data acquisition unit that acquires transaction data including a summary of the transaction; an account selection unit that, when the transaction data acquisition unit acquires the transaction data after the occurrence of the event, inputs the summary included in the transaction data into the post-event learning model and selects an account item output by the post-event learning model; an accounting data generation unit that generates accounting data corresponding to the transaction data, including the account item selected by the item selection unit; An accounting data conversion device comprising:

13. Computer, a learning model storage unit that stores a pre-event learning model including an input layer, an intermediate layer, and an output layer, which is generated or updated using learning data including summaries of transactions of business operators before an event occurs and account items corresponding to the transactions, and which takes summaries as input and account items as output, and a post-event learning model generated by updating a new output layer added in place of the output layer of the pre-event learning model using learning data including summaries of transactions after the event occurs and account items corresponding to the transactions; a transaction data acquisition unit for acquiring transaction data including a summary of the transaction; an account selection unit that, when the transaction data acquisition unit acquires the transaction data after the occurrence of the event, inputs the summary included in the transaction data into the post-event learning model and selects an account item output by the post-event learning model; and an accounting data generation unit that generates accounting data corresponding to the transaction data, including the account item selected by the item selection unit; A program to function as a

14. 1. A computer-implemented method for converting accounting data, comprising: a storage step of storing in a learning model storage unit a pre-event learning model including an input layer, an intermediate layer, and an output layer, which is generated or updated using learning data including summaries of transactions of business operators before an event occurs and account items corresponding to the transactions, and which takes summaries as input and account items as output, and a post-event learning model generated by updating a new output layer added in place of the output layer of the pre-event learning model using learning data including summaries of transactions after the event occurs and account items corresponding to the transactions; A transaction data acquisition stage for acquiring transaction data including a summary of the transaction; a post-event learning model outputting an account item selected in a post-event learning model by inputting the summary included in the transaction data obtained after the event has occurred in the transaction data obtaining step; and an accounting data generation step of generating accounting data corresponding to the transaction data, including the account item selected in the item selection step; An accounting data conversion method comprising: