JP7392190B1 - Accounting data conversion device, accounting data conversion method, learning device, learning method, and program - Google Patents

Abstract

The present invention provides an accounting data conversion device, an accounting data conversion method, a learning device, a learning method, and a program that generate accounting data corresponding to transaction data provided by the accounting data conversion device. In an accounting management system 10, an accounting data conversion device includes a transaction data acquisition unit that acquires transaction data including a summary of a transaction with a user 82, a transaction summary included in the transaction data, and a transaction entity attribute. and attribute data including at least one of the transaction attributes; an item selection unit that selects an account item corresponding to the transaction data based on; and accounting data corresponding to the transaction data, including the account item selected by the item selection unit. and an accounting data generation unit that generates. [Selection diagram] Figure 1

Description

The present invention relates to an accounting data conversion device, an accounting data conversion method, a learning device, a learning method, and a program.

Patent Document 1 describes a method for creating a correspondence relationship for journalizing the contents of a transaction into a predetermined account using transaction data representing the contents of the transaction and past journal data including the account in which the contents of the transaction were journalized. This document describes a journal entry rule creation support device that supports the creation of journal entry rules.
[Prior art documents]
[Patent document]
[Patent Document 1] Japanese Patent Application Publication No. 2019-191749

According to a first aspect of the present invention, an accounting data conversion device is provided. The accounting data conversion device may include a transaction data acquisition unit that acquires transaction data including a transaction summary. The accounting data conversion device includes an item selection unit that selects an account item corresponding to the transaction data based on the summary included in the transaction data and at least one of a subject attribute and a transaction attribute of the transaction. good. The accounting data conversion device may include an accounting data generation unit that generates accounting data corresponding to the transaction data, including the account item selected by the item selection unit.

Note that the above summary of the invention does not list all the necessary features of the invention. Furthermore, subcombinations of these features may also constitute inventions.

An example of an accounting management system 10 is schematically shown. An example of the functional configuration of the learning device 100 is schematically shown. An example of the flow of processing by the learning device 100 is schematically shown. 3 is an explanatory diagram for conceptually explaining a learning model 300. FIG. 3 is an explanatory diagram for conceptually explaining a learning model 300. FIG. It is an explanatory diagram for conceptually explaining stage-based learning. An example of the functional configuration of the accounting data conversion device 200 is schematically shown. An example of the flow of processing by the accounting data conversion device 200 is schematically shown. An example of the hardware configuration of a computer 1200 functioning as the learning device 100 or the accounting data conversion device 200 is schematically shown.

Hereinafter, the present invention will be described through embodiments of the invention, but the following embodiments do not limit the invention according to the claims. Furthermore, not all combinations of features described in the embodiments are essential to the solution of the invention.

FIG. 1 schematically shows an example of an accounting management system 10. As shown in FIG. The accounting management system 10 includes a learning device 100 and an accounting data conversion device 200. The learning device 100 and the accounting data conversion device 200 may be different devices. The learning device 100 and the accounting data conversion device 200 may be realized by one device. Learning device 100 may be realized by multiple devices. Accounting data conversion device 200 may be realized by a plurality of devices.

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 one device. The business service providing device 20 and the accounting data converting device 200 may be realized by one device. The business service providing device 20, the learning device 100, and the accounting data converting device 200 may be realized by one device.

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 one device. The accounting service providing device 30 and the accounting data converting device 200 may be realized by one device. The accounting service providing device 30, the learning device 100, and the accounting data converting device 200 may be realized by one device.

The learning device 100, the accounting data conversion device 200, the business service providing device 20, and the accounting service providing device 30 may be realized by one device. Here, we mainly take as an example a case where the learning device 100, the accounting data conversion device 200, the business service providing device 20, and the accounting service providing device 30 are different devices.

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 the network 50. Network 50 may include a cloud network. Network 50 may include the Internet. Network 50 may include a mobile communications 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.

The business service providing device 20 manages transaction data of the user 82 who is the subject of the transaction. User 82 is a business operator. The user 82 may be a sole proprietor or a corporation.

The transaction data may include transaction date, business partner, tax classification, amount, income and expenditure identification data indicating whether the transaction is income or expenditure, consumption tax amount, and summary. If the user 82 is a corporation, the transaction data may include data indicating which department of the corporation the transaction is in.

The business service providing device 20 may further manage transaction-related data related to the user's 82 transactions. Transaction-related data may include, for example, origin data indicating the origin of the transaction. Examples of sources of transactions include POS (Point of Sale) systems, financial institutions (name, type), credit cards, electronic money, and evidence such as invoices and delivery notes. The transaction-related data may include, for example, source data indicating the source of the transaction. Examples of transaction import sources include import source services, CSV (Comma Separated Values), images, and the like. Examples of import source services include billing services, internet banking, and POS. The transaction-related data may include, for example, customer-related data related to a customer. The customer-related data may include the name of the customer, the size of the customer, the industry of the customer, and the like.

The business service providing device 20 may further manage user-related data related to the user 82. User-related data may include user 82's own data. Examples of user 82's own data include name, industry type, side business information (such as whether the user also works in the real estate business), company type (stock company, limited liability company, etc.), and business type identification data indicating whether the user is a sole proprietorship or a corporation. , in the case of a corporation, the size of the corporation (capital, number of employees, number of business partners, sales, etc.), whether or not it is a taxable business, if it is a taxable business, what is the consumption tax taxation method, and special provisions for small businesses. whether or not you are eligible for the service, whether you are using an accounting service, and if you are using an accounting service, the type of service you are using, the usage status of templates by industry, and the function of the department that manages accounting data for each department. These include whether they are using the service and the degree to which they have used the accounting service (number of transactions, period of use, frequency of use, etc.). The user-related data may include data on groups to which the user 82 belongs. Examples of the data of the group to which the user 82 belongs include the name, industry, and size of the parent company, the name and size of the accounting firm in charge of the user 82, and the like.

The business service providing device 20 may further manage transaction data related data related to the transaction data of the user 82. The transaction data related data may include the number of transaction data. The transaction data related data may include bias data indicating bias in the transaction data.

The business service providing device 20 receives transaction data of the user 82 from the communication terminal 80 of the user 82, for example. The business service providing device 20 may receive transaction-related data from the communication terminal 80. The business service providing device 20 may receive user-related data from the communication terminal 80. The business service providing device 20 may communicate with the communication terminal 80 via the network 50. The communication terminal 80 may be any terminal that can communicate. For example, the communication terminal 80 is a mobile phone such as a smartphone, a tablet terminal, a PC (Personal Computer), or the like. The business service providing device 20 may acquire transaction data, transaction-related data, and user-related data of the user 82 from an accounting firm that manages accounting information of the user 82 or the like.

The business service providing device 20 may provide various data to the learning device 100. The learning device 100 may receive transaction data of the user 82 from the business service providing device 20. The learning device 100 may receive transaction-related data of the user 82 from the business service providing device 20. The learning device 100 may receive user-related data of the user 82 from the business service providing device 20. The learning device 100 may receive transaction data related data of the user 82 from the business service providing device 20. The learning device 100 may acquire transaction-related data and user-related data of the user 82 from a source other than the business service providing device 20.

The learning device 100 further acquires accounting data corresponding to the transaction data generated by actually journalizing the transaction data of the user 82. The accounting data includes account items corresponding to transactions.

The learning device 100 generates a learning model for journalizing by executing machine learning using the acquired data. Conventionally, a learning model infers the account corresponding to a transaction from the transaction summary by adjusting the parameters of a natural language processing model such as BERT using the transaction summary and the account corresponding to the transaction as learning data. was being generated.

The learning device 100 according to the present embodiment inputs transaction summaries and attribute data through machine learning using learning data that further includes attribute data in addition to transaction summaries and account items corresponding to the transactions. , generate a learning model whose output is the account item corresponding to the transaction.

The attribute data includes at least one of a transaction subject attribute and a transaction attribute. That is, the attribute data may include subject attributes, may include transaction attributes, and may include both subject attributes and transaction attributes. The subject attribute may be an attribute indicating the subject of the transaction. The transaction attribute may be an attribute indicating a transaction.

The subject attributes may include user-related data. The subject attribute may include at least one of multiple types of user-related data.

Even if the contents of the summary are the same, the account items corresponding to the transactions may differ depending on the user 82. What kind of account item is appropriate is considered to depend on the subject attributes of the user 82, for example, who the user 82 is, the type of industry of the user 82, whether the user 82 has a side job, or if the user 82 has a side job. The type of side job of the user 82, the company type of the user 82, whether the user 82 is a sole proprietor or a corporation, the size of the corporation if it is a corporation, whether the user 82 is a taxable business owner, and the consumption tax taxation method. whether the user 82 is subject to special provisions for large-scale enterprises, whether the user 82 is using accounting services, the type of service being used, the usage status of accounting services, whether the user is using departmental functions, accounting It is thought that it depends on the level of use of the service, the name, industry, and size of the user's 82 parent company, the name and size of the accounting firm in charge of the user 82, and the like. Therefore, by learning with additional subject attributes, it is possible to improve the accuracy rate of account items compared to learning using only summaries. Note that the correct account item is an appropriate account item for a transaction, and the correct answer rate is the probability of selecting an appropriate account item.

The transaction attribute may include data other than the summary of the transaction data. The transaction attributes may include at least one of the transaction date, business partner, tax category, amount, income or expenditure, and consumption tax amount. Transaction attributes may include transaction-related data. The transaction attribute may include at least one of multiple types of transaction-related data. Transaction attributes may include transaction data related data. The transaction attribute may include at least one of multiple types of transaction data related data.

Even if the contents of the summary are the same, the corresponding account items may differ depending on the transaction. What kind of account is appropriate depends on transaction attributes, such as transaction date, business partner, tax classification, amount, income or expenditure, consumption tax amount, origin of transaction, source of transaction, transaction It is thought that it depends on the name, size, industry, etc. of the company. Therefore, by learning with additional transaction attributes, it is possible to improve the accuracy rate of account items compared to learning using only summaries.

Accounting data conversion device 200 receives the learning model generated by learning device 100 from 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. The accounting data conversion device 200 may receive attribute data corresponding to the transaction data from the business service providing device 20. For example, the accounting data conversion device 200 receives transaction-related data of the user 82 from the business service providing device 20. For example, the accounting data conversion device 200 receives user-related data of the user 82 from the business service providing device 20. For example, the accounting data conversion device 200 receives transaction data related data of the user 82 from the business service providing device 20. The accounting data conversion device 200 may acquire attribute data from a source other than the business service providing device 20.

The accounting data conversion device 200 uses the learning model received from the learning device 100 and stored to generate accounting data corresponding to the transaction data acquired from the accounting data conversion device 200. The accounting data conversion device 200 inputs a summary included in transaction data and attribute data corresponding to the transaction data into a learning model, selects an account item output from the learning model, and converts accounting data including the selected account item. may be generated. The accounting data conversion device 200 may output the generated accounting data to the accounting service providing device 30.

The accounting service providing device 30 manages accounting data acquired from the accounting data converting device 200. The accounting data conversion device 200 provides accounting services to the user 82 using the user's 82 accounting data.

FIG. 2 schematically shows an example of the functional configuration of the learning device 100. The learning device 100 includes a data acquisition section 102, a learning data storage section 104, a learning model generation section 106, and a learning model transmission section 108.

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

The data acquisition unit 102 acquires transaction-related data of the user 82. The data acquisition unit 102 may acquire transaction-related data from the business service providing device 20. The data acquisition unit 102 may acquire transaction-related data from a source other than the business service providing device 20.

The data acquisition unit 102 acquires user-related data of the user 82. The data acquisition unit 102 may acquire user-related data from the business service providing device 20. The data acquisition unit 102 may acquire user-related data from a source other than the business service providing device 20.

The data acquisition unit 102 acquires transaction data related data of the user 82. The data acquisition unit 102 may acquire transaction data related data from the business service providing device 20.

The data acquisition unit 102 acquires accounting data corresponding to the transaction data, which is generated by actually journalizing the transaction data of the user 82 and includes an account item corresponding to the transaction. The data acquisition unit 102 may acquire the accounting data from the business service providing device 20. The data acquisition unit 102 may acquire the accounting data from a source other than the business service providing device 20.

The data acquisition unit 102 causes the learning data storage unit 104 to store learning data generated from the acquired data.

The learning data includes a summary of the transaction and an account item corresponding to the transaction. The learning data further includes attribute data. The attribute data includes at least one of a transaction subject attribute and a transaction attribute.

The learning model generation unit 106 uses machine learning using the learning data stored in the learning data storage unit 104 to generate a learning model that inputs the transaction summary and attribute data and outputs the account item corresponding to the transaction. generate. That is, the learning model generation unit 106 receives as input the summary of the transaction and at least one of the transaction entity attribute and the transaction attribute, and generates a learning model that outputs the account item corresponding to the transaction.

The learning model generation unit 106 may perform machine learning using a summary of a transaction and attribute data to which a weight different from that of the summary is applied. Information that is subject to natural language processing, such as the summary, and information that is not subject to natural language processing, such as attribute data, can have a significantly different number of dimensions, but attribute data may be given a different weight than the summary. These can be balanced by applying

When applying weights to attribute data, the learning model generation unit 106 may make the weights applied to subject attributes and the weights applied to transaction attributes different. Thereby, for example, learning can be performed in which one of the entity attribute and transaction attribute is compared with the other.

The learning model generation unit 106 may apply different weights to each of multiple types of data included in the attribute data. For example, the learning model generation unit 106 may apply different weights to each of the multiple types of data included in the entity attribute and the multiple types of data included in the transaction attribute. For example, the learning model generation unit 106 multiplies the initial values before learning by different coefficients for each of the plurality of types of data included in the subject attribute and the plurality of types of data included in the transaction attribute. For example, the learning model generation unit 106 multiplies the weight matrix by different coefficients for each of the plurality of types of data included in the subject attribute and the plurality of types of data included in the transaction attribute. For example, the learning model generation unit 106 changes the learning rate for each of multiple types of data included in the subject attribute and multiple types of data included in the transaction attribute. For example, when changing each of multiple types of data included in a subject attribute and multiple types of data included in a transaction attribute to a One-hot expression, the learning model generation unit 106 uses a One-hot expression. Multiply the value when set by a different coefficient. For example, when changing each of multiple types of data included in a subject attribute and multiple types of data included in a transaction attribute to a One-hot expression, the learning model generation unit 106 uses a One-hot expression. Executes the process of aligning the number of dimensions when . The patterns of possible values of attribute data vary, for example, some are binary, such as sole proprietorship or incorporated, while others can take multiple values, such as the source of a transaction. There may be differences in the ease of learning. On the other hand, by making such adjustments, it is possible to reduce the possibility that unintended differences will occur in the ease of learning.

The learning model generation unit 106 uses learning data to update parameters of a pre-stored natural language processing model, thereby creating a learning model that takes transaction summary and attribute data as input and accounts as output. may be generated. That is, the learning model may be a natural language processing model whose parameters have been updated using learning data. The learning model generation unit 106 may use a natural language processing model such as fastText, for example. The learning model generation unit 106 inputs the transaction summary and attribute data into the natural language processing model, selects the account item corresponding to the transaction data, compares the learning data with the selection result, and performs natural language processing. You may update model parameters. The learning model generation unit 106 updates the parameters of the natural language processing model so that the selection result matches the account item included in the learning data.

The learning model generation unit 106 first updates the parameters of the natural language processing model using the transaction summary, and then updates the parameters of the natural language processing model using the attribute data. Good too. As a specific example, the learning model generation unit 106 first learns the part corresponding to the summary of the input weight matrix of the natural language processing model, fixes the weight for the word, and learns the part corresponding to the attribute data. . Thereby, it is possible to preferentially learn the weight matrix for the summary.

The learning model generation unit 106 may further input into the natural language processing model balance data indicating whether the account item to be output is a debit account item or a credit account item. The debit account and credit account corresponding to a transaction will be different, and by inputting debit and debit data into a natural language processing model, it is possible to perform learning that reflects this difference, and to determine the correct answer for the account. rate can be improved. The learning model generated by the learning model generation unit 106 may output both debit account items and credit account items. In this case, the learning model generation unit 106 may input credit and debit data indicating both debit and credit account items into the natural language processing model, or may avoid inputting credit and debit data into the natural language processing model. It's okay.

When inputting multiple types of attribute data, the learning model generation unit 106 may perform learning that reflects the correlation between the multiple types of attribute data. For example, when using the first attribute data and the second attribute data, the learning model generation unit 106 further uses data indicating the correlation between the first attribute data and the second attribute data for learning. Execute. As a specific example, regarding attribute data A representing ON or OFF and attribute data B representing ON or OFF, if only either attribute data A or attribute data B is ON, C is not the correct answer, but both In a case where C is the correct answer when ON or both are OFF, the learning model generation unit 106 inputs data indicating the XOR of attribute data A and attribute data B. Thereby, a learning model that reflects such a correlation can be generated.

The learning model generation unit 106 infers account items for transaction data using each of a plurality of types of learning models, and uses multiple types of learning models to select an account item corresponding to the transaction data from the plurality of inference results. A model may also be generated. For example, the learning model generation unit 106 uses machine learning using learning data including transaction summaries and account items to generate a learning model (sometimes referred to as a summary learning model) that uses transaction summaries as input and account items as output. ), and by machine learning using learning data including attribute data and account items, a learning model (sometimes referred to as an attribute learning model) that uses attribute data as input and account items as output is created. generate. The learning model generation unit 106 generates a learning model (sometimes referred to as a subject attribute learning model) that uses subject attributes as input and accounts as output by machine learning using learning data including subject attributes and account subjects. A learning model (sometimes referred to as a transaction attribute learning model) that uses transaction attributes as input and account items as output is generated by machine learning using learning data including transaction attributes and account items. It's okay. The learning model generation unit 106 may generate multiple types of learning models for the attribute data in even smaller units.

The learning model transmitter 108 transmits the learning model generated by the learning model generator 106 to the accounting data conversion device 200. When the summary learning model and the attribute learning model are generated by the learning model generation unit 106, the learning model transmitting unit 108 transmits the summary learning model and the attribute learning model together to the accounting data conversion device 200. When the learning model generation unit 106 generates a summary learning model, a subject attribute learning model, and a transaction attribute learning model, the learning model transmitting unit 108 generates a summary learning model, a subject attribute learning model, and a transaction attribute learning model. It is transmitted to the accounting data conversion device 200.

FIG. 3 schematically shows an example of the flow of processing by the learning model generation unit 106.

In step (step may be abbreviated as S) 102, the learning model generation unit 106 acquires learning data from the learning data storage unit 104. The learning model generation unit 106 acquires learning data used for learning from among a large number of learning data stored in the learning data storage unit 104.

In S104, the learning model generation unit 106 updates the parameters of the natural language processing model using the learning data acquired in S102. The learning model generation unit 106 inputs the transaction summary and attribute data into the natural language processing model, selects the account item corresponding to the transaction data, compares the learning data with the selection result, and performs natural language processing. Update model parameters. The learning model generation unit 106 updates the parameters of the natural language processing model so that the selection result matches the account item included in the learning data.

FIG. 4 is an explanatory diagram for conceptually explaining the learning model 300. In this example, two attribute data (attribute data A304, attribute data B306) are illustrated, but the number of attribute data is not limited to this.

In this example, the learning model generation unit 106 first converts the summary 302, attribute data A 304, and attribute data B 306 into an input layer x 310 of a One-hot vector. The learning model generation unit 106 multiplies the input layer x310 by the input weight matrix W _in 312 to calculate the intermediate layer y320 (distributed representation of words). The learning model generation unit 106 calculates an output layer z330 (accuracy for each subject) by multiplying the intermediate layer y320 by the output weight matrix W _out 322.

In this example, the learning model generation unit 106 adds attribute data to the input layer x310, but the invention is not limited to this. The learning model generation unit 106 may add attribute data to the intermediate layer y320. The learning model generation unit 106 may add subject attributes to the input layer x310 and transaction attributes to the intermediate layer y320. The learning model generation unit 106 may add subject attributes to the intermediate layer y320 and transaction attributes to the input layer x310. The learning model generation unit 106 may select whether to add to the input layer x310 or to the intermediate layer y320 for each of multiple types of attribute data. Depending on the type of attribute data, adding it to the input layer x310 may improve the accuracy rate of account items, and adding it to the intermediate layer y320 may improve the accuracy rate of account items. For each of multiple types of attribute data, a learning model when added to the input layer x 310 and a learning model when added to the intermediate layer y 320 are generated, and which learning model actually has a higher accuracy rate for account items. It may be determined whether to add each of the plurality of types of attribute data to the input layer x310 or the intermediate layer y320 by verifying whether the value becomes high. The learning model generation unit 106 may add each of the plurality of types of attribute data to the input layer x310 or the intermediate layer y320 according to the determination.

FIG. 5 is an explanatory diagram for conceptually explaining the learning model 300. Mainly the points different from FIG. 4 will be explained. In this example, the learning model generation unit 106 applies a different weight to the attribute data than the summary.

In the example shown in FIG. 5, the weight of the abstract is 1, while the weight α is applied to the attribute data. By setting α smaller than 1, the influence of attribute data can be made smaller than that of the abstract, and by setting α larger than 1, the influence of attribute data can be made larger than that of the summary.

The learning model generation unit 106 may apply such weights to both the input layer x310 and the hidden layer y320. The learning model generation unit 106 may apply such a weight to only one of the input layer x310 and the intermediate layer y320. The learning model generation unit 106 may use different weights for each type of attribute data.

FIG. 6 is an explanatory diagram for conceptually explaining stage-based learning. In the first learning, the learning model generation unit 106 inputs only the summary 302 to the input layer x310 and learns the input weight matrix W _in 314. In the second learning, the learning model generation unit 106 fixes the weight matrix W _in 314 and learns the weight matrix W _in 316 corresponding to the attribute data. In this way, by learning the part corresponding to the summary 302 first, the effect of the summary 302 can be prioritized.

FIG. 7 schematically shows an example of the functional configuration of the accounting data conversion device 200. The accounting data conversion device 200 includes a learning model acquisition section 202, a learning model storage section 204, a transaction data acquisition section 206, an attribute data acquisition section 208, a subject selection section 210, an accounting data generation section 212, and an accounting data output section 214. .

The learning model acquisition unit 202 receives a summary and attribute data as input, and acquires a learning model that outputs account items. The learning model acquisition unit 202 may acquire a learning model from the learning device 100. The learning model acquisition unit 202 may acquire the learning model from a source other than the learning device 100. The learning model storage unit 204 stores the learning model acquired by the learning model acquisition unit 202.

Transaction data acquisition unit 206 acquires transaction data. The transaction data acquisition unit 206 may acquire transaction data from which account items are to be inferred. The transaction data acquisition unit 206 acquires transaction data from the business service providing device 20, for example. The transaction data acquisition unit 206 acquires transaction data from the communication terminal 80, for example.

The attribute data acquisition unit 208 acquires attribute data corresponding to the transaction data acquired by the transaction data acquisition unit 206. The attribute data acquisition unit 208 may acquire multiple types of attribute data. The attribute data acquisition unit 208 acquires attribute data from the business service providing device 20, for example. The attribute data acquisition unit 208 acquires attribute data from the communication terminal 80, for example. The attribute data acquisition unit 208 acquires attribute data from the transaction data acquired by the transaction data acquisition unit 206, for example.

As a specific example, the attribute data acquisition unit 208 may acquire at least one of the transaction date, business partner, tax classification, amount, income or expenditure, and consumption tax amount from the transaction data. Further, the attribute data acquisition unit 208 may acquire transaction-related data from the business service providing device 20 or the communication terminal 80. Further, the attribute data acquisition unit 208 acquires source data from the business service providing device 20 or the communication terminal 80 . Further, the attribute data acquisition unit 208 acquires transaction-related data from the business service providing device 20 or the communication terminal 80. Further, the attribute data acquisition unit 208 acquires user-related data from the business service providing device 20 or the communication terminal 80. Further, the attribute data acquisition unit 208 may acquire transaction data related data from the business service providing device 20.

The item selection unit 210 selects the account item corresponding to the transaction indicated by the transaction data acquired by the transaction data acquisition unit 206. The item selection unit 210 may select the account item corresponding to the transaction data based on the summary included in the transaction data acquired by the transaction data acquisition unit 206 and the attribute data acquired by the attribute data acquisition unit 208.

The subject selection unit 210 may select account subjects using the learning model stored in the learning model storage unit 204. The subject selection unit 210 inputs the summary included in the transaction data acquired by the transaction data acquisition unit 206 and the attribute data acquired by the attribute data acquisition unit 208 into the learning model stored in the learning model storage unit 204. , the account items output from the learning model may be selected. The subject selection unit 210 adds the summary included in the transaction data acquired by the transaction data acquisition unit 206 and the attribute data acquisition unit 208 to a learning model generated by machine learning using the summary, attribute data, and account items. You may select the account item output from the learning model by inputting the attribute data that has been created. The subject selection unit 210 may input the summary, subject attributes, and transaction attributes to the learning model. The subject selection unit 210 may input the summary and subject attributes to the learning model. The subject selection unit 210 may input the summary and transaction attributes to the learning model.

The subject selection unit 210 may input designated attribute data from among the plurality of types of attribute data into the learning model. The subject selection unit 210 may select attribute data to be input to the learning model from a plurality of types of attribute data, and input the selected attribute data to the learning model. The item selection unit 210 selects, for example, at least one of the industry of the subject of the transaction, business type identification data, income and expenditure identification data, the number of digits of the amount of the transaction, the origin of the transaction, and the tax classification of the transaction. These data have a relatively large influence on the selection of account items, and by using these data, it is possible to improve the accuracy rate of account items.

When using multiple types of attribute data, the subject selection unit 210 may set the priority of some attribute data higher than the priority of other attribute data. Examples of attribute data to be prioritized include the industry of the subject of the transaction, business type identification data, income and expenditure identification data, the number of digits of the amount of the transaction, the origin of the transaction, and the tax classification of the transaction. These data have a relatively large influence on the selection of account items, and by giving these data a high priority, it is possible to improve the accuracy rate of account items.

The subject selection unit 210 may apply a different weight to the attribute data than the summary and input it into the learning model. Information that is subject to natural language processing, such as the summary, and information that is not subject to natural language processing, such as attribute data, can have a significantly different number of dimensions, but attribute data may be given a different weight than the summary. These can be balanced by applying

When applying weights to attribute data, the subject selection unit 210 may make the weights applied to subject attributes and the weights applied to transaction attributes different. Thereby, for example, it is possible to perform inferences that emphasize one of the entity attributes and transaction attributes by comparing them with the other.

The subject selection unit 210 may apply different weights to each of multiple types of data included in the attribute data. For example, the subject selection unit 210 may apply different weights to each of multiple types of data included in the subject attribute and multiple types of data included in the transaction attribute.

The item selection unit 210 may select the account item corresponding to the transaction data further based on whether a debit account item or a credit account item is selected. In this case, the subject selection unit 210 may input loan data to the learning model in addition to the summary and attribute data. The item selection unit 210 may select whether to select a debit account item or a credit account item, for example, according to an instruction from the user 82. The item selection unit 210 may select whether to select a debit account item or a credit account item based on the attribute data acquired by the attribute data acquisition unit 208. For example, the item selection unit 210 selects whether to select a debit account item or a credit account item, depending on the import source data. As a specific example, when the import source data indicates Internet banking, the item selection unit 210 selects a credit account item in the case of an income transaction, and selects a debit account item in the case of an expenditure transaction. I decided to. The subject of inference is mainly PL (Profit and Loss statement) subjects. Further, for example, when the transaction data indicates POS, the item selection unit 210 may select a debit account item in the case of an income transaction, and select a credit account item in the case of an expenditure transaction. do. The subject of inference is mainly BS (Balance Sheet) subjects.

When inputting multiple types of attribute data to the learning model, the subject selection unit 210 may perform input that reflects the correlation between the multiple types of attribute data. For example, the subject selection unit 210 inputs data indicating an XOR between attribute data A and attribute data B to the learning model, regarding attribute data A indicating ON or OFF and attribute data B indicating ON or OFF.

If the learning model storage unit 204 stores a summary learning model and an attribute learning model, the subject selection unit 210 may use the summary learning model and the attribute learning model. For example, the subject selection unit 210 inputs the summary included in the transaction data acquired by the transaction data acquisition unit 206 into the applied learning model, and selects the account items output from the summary learning model and the attributes acquired by the attribute data acquisition unit 208. Data is input into the attribute learning model, and an account item corresponding to the transaction data is selected based on the account item output from the attribute learning model. The subject selection unit 210 selects, for example, an account subject with higher accuracy from among the account subjects output from the summary learning model and the account subjects output from the attribute learning model. For example, the subject selection unit 210 weights differently the accuracy of the account output from the summary learning model and the accuracy of the account output from the attribute learning model, and then selects the account with higher accuracy. You can do it like this. This makes it possible to perform inference that emphasizes either the summary or the attribute data.

The accounting data generation unit 212 generates accounting data corresponding to the transaction data acquired by the transaction data acquisition unit 206. The accounting data generation unit 212 generates accounting data including the account items selected by the item selection unit 210.

Accounting data output section 214 outputs accounting data generated by accounting data generation section 212. The accounting data output unit 214 may output accounting data to the accounting service providing device 30. The accounting data output unit 214 may output accounting data to the communication terminal 80.

FIG. 8 schematically shows an example of the flow of processing by the accounting data conversion device 200. Here, the flow of processing when generating accounting data corresponding to each of a plurality of transaction data will be explained.

In S202, the transaction data acquisition unit 206 acquires transaction data. In S204, the attribute data acquisition unit 208 acquires attribute data corresponding to the transaction data acquired by the transaction data acquisition unit 206 in S202.

In S206, the subject selection unit 210 inputs the summary included in the transaction data acquired by the transaction data acquisition unit 206 in S202 and the attribute data acquired by the attribute data acquisition unit 208 in S204 to the learning model, and Select the account output from . In S208, the accounting data generation unit 212 generates accounting data including the account item selected by the item selection unit 210 in S208, which corresponds to the transaction data acquired by the transaction data acquisition unit 206 in S202.

If generation of accounting data has been completed for all target transaction data (YES in S210), the process proceeds to S212, and if it has not been completed (NO in S210), the process returns to S202. In S212, the accounting data output unit 214 outputs the generated accounting data to the accounting service providing device 30.

FIG. 9 schematically shows an example of the hardware configuration of a computer 1200 that functions as the learning device 100 or the accounting data conversion device 200. The program installed on the computer 1200 causes the computer 1200 to function as one or more "parts" of the apparatus according to the present embodiment, or causes the computer 1200 to perform operations associated with the apparatus according to the present embodiment or the one or more "parts" of the apparatus according to the present embodiment. Multiple units may be executed and/or the computer 1200 may execute a process or a stage of a process according to the present embodiments. Such programs may be executed by CPU 1212 to cause computer 1200 to perform certain operations associated with some or all of the blocks in the flowcharts and block diagrams described herein.

The computer 1200 according to this embodiment includes a CPU 1212, a RAM 1214, and a graphics controller 1216, which are interconnected 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. DVD drive 1226 may be a DVD-ROM drive, a DVD-RAM drive, or the like. Storage device 1224 may be a hard disk drive, solid state drive, or the like. Computer 1200 also includes legacy input/output units, such as ROM 1230 and a keyboard, which are connected to input/output controller 1220 via input/output chips 1240.

The CPU 1212 operates according to programs stored in the ROM 1230 and RAM 1214, thereby controlling each unit. Graphics controller 1216 obtains image data generated by CPU 1212, such as in a frame buffer provided in RAM 1214 or itself, and causes the image data to be displayed on display device 1218.

Communication interface 1222 communicates with other electronic devices via a network. Storage device 1224 stores programs and data used by CPU 1212 within computer 1200. The DVD drive 1226 reads a program or data from a DVD-ROM 1227 or the like and provides it to the storage device 1224. The IC card drive reads programs and data from and/or writes programs and data to the IC card.

ROM 1230 stores therein programs that are dependent on the computer 1200 hardware, such as a boot program that is executed by the computer 1200 upon activation. I/O chip 1240 may also connect various I/O units to I/O controller 1220 via USB ports, parallel ports, serial ports, keyboard ports, mouse ports, etc.

The program is provided by a computer readable storage medium such as a DVD-ROM 1227 or an IC card. The program is read from a computer-readable storage medium, installed in storage device 1224, RAM 1214, or ROM 1230, which are also examples of computer-readable storage media, and executed by CPU 1212. The information processing described in these programs is read by the computer 1200 and provides coordination between the programs and the various types of hardware resources described above. An apparatus or method may be configured to implement the operation or processing of information according to the use of computer 1200.

For example, when communication is performed between the computer 1200 and an external device, the CPU 1212 executes a communication program loaded into the RAM 1214 and sends communication processing to the communication interface 1222 based on the processing written in the communication program. You may give orders. The communication interface 1222 reads transmission data stored in a transmission buffer area provided in a recording medium such as a RAM 1214, a storage device 1224, a DVD-ROM 1227, or an IC card under the control of the CPU 1212, and transmits the read transmission data. Data is transmitted to the network, or received data received from the network is written to a reception buffer area provided on the recording medium.

Further, the CPU 1212 causes the RAM 1214 to read all or a necessary part of a file or database stored in an external recording medium such as a storage device 1224, a DVD drive 1226 (DVD-ROM 1227), or an IC card. Various types of processing may be performed on the data. CPU 1212 may then write the processed data back to an external storage medium.

Various types of information, such as various types of programs, data, tables, and databases, may be stored on a recording medium and subjected to information processing. CPU 1212 performs various types of operations, information processing, conditional determination, conditional branching, unconditional branching, and information retrieval on data read from RAM 1214 as described elsewhere in this disclosure and specified by the program's instruction sequence. Various types of processing may be performed, including /substitutions, etc., and the results are written back to RAM 1214. Further, the CPU 1212 may search for information in a file in a recording medium, a database, or the like. For example, when a plurality of entries are stored in a recording medium, each having an attribute value of a first attribute associated with an attribute value of a second attribute, the CPU 1212 selects the first entry from among the plurality of entries. Search for an entry whose attribute value matches the specified condition, read the attribute value of the second attribute stored in the entry, and then set the attribute value to the first attribute that satisfies the predetermined condition. An attribute value of the associated second attribute may be obtained.

The programs or software modules described above may be stored in a computer-readable storage medium on or near computer 1200. Also, a storage medium such as a hard disk or 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 allowing the program to be transferred to the computer 1200 via the network. provide.

Blocks in the flowcharts and block diagrams of the present embodiments may represent stages in a process in which an operation is performed or a "part" of a device responsible for performing the operation. Certain steps and units may be provided with dedicated circuitry, programmable circuitry provided with computer readable instructions stored on a computer readable storage medium, and/or provided with computer readable instructions stored on a computer readable storage medium. May be implemented by a processor. Dedicated circuitry may include digital and/or analog hardware circuits, and may include integrated circuits (ICs) and/or discrete circuits. Programmable circuits can perform AND, OR, EXCLUSIVE OR, NAND, NOR, and other logical operations, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), etc. , flip-flops, registers, and memory elements.

A computer-readable storage medium may include any tangible device capable of storing instructions for execution by a suitable device such that a computer-readable storage medium with instructions stored therein may be illustrated in a flowchart or block diagram. A product will be provided that includes instructions that can be executed to create a means for performing specified operations. 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 include floppy disks, diskettes, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory). , Electrically Erasable Programmable Read Only Memory (EEPROM), Static Random Access Memory (SRAM), Compact Disc Read Only Memory (CD-ROM), Digital Versatile Disc (DVD), Blu-ray Disc, Memory Stick , integrated circuit cards, and the like.

Computer-readable instructions may include assembler instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state configuration data, or instructions such as Smalltalk®, JAVA®, C++, etc. any source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as may include.

The computer-readable instructions are for producing means for a processor of a general purpose computer, special purpose computer, or other programmable data processing device, or programmable circuit to perform the operations specified in the flowchart or block diagrams. A general purpose computer, special purpose computer, or other programmable data processor, locally or over a local area network (LAN), wide area network (WAN), such as the Internet, to execute the computer readable instructions. It may be provided in a processor or programmable circuit of the device. Examples of processors include computer processors, processing units, microprocessors, digital signal processors, controllers, microcontrollers, and the like.

Although the present invention has been described above using the embodiments, the technical scope of the present invention is not limited to the range described in the above embodiments. It will be apparent to those skilled in the art that various changes or improvements can be made to the embodiments described above. It is clear from the claims that such modifications or improvements may be included within the technical scope of the present invention.

The order of execution of each operation, procedure, step, stage, etc. in the apparatus, system, program, and method shown in the claims, specification, and drawings is specifically defined as "before" or "before". It should be noted that they can be implemented in any order unless the output of the previous process is used in the subsequent process. Even if the claims, specifications, and operational flows in the drawings are explained using ``first'', ``next'', etc. for convenience, this does not mean that it is essential to carry out the operations in this order. It's not a thing.

10 accounting management system, 20 business service providing device, 30 accounting service providing device, 50 network, 80 communication terminal, 82 user, 100 learning device, 102 data acquisition section, 104 learning data storage section, 106 learning model generation section, 108 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 Attribute data acquisition unit, 210 Subject selection unit, 212 Accounting data generation unit, 214 Accounting data output unit , 300 learning model, 310 input layer, 320 intermediate layer, 330 output layer, 1200 computer, 1210 host controller, 1212 CPU, 1214 RAM, 1216 graphic controller, 1218 display device, 1220 input/output controller, 1222 communication interface, 1224 storage device , 1226 DVD drive, 1227 DVD-ROM, 1230 ROM, 1240 input/output chip

Claims (20)

a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation unit that generates a learning model that outputs account items;
a learning model storage unit that stores the learning model generated by the learning model generation unit;
a transaction data acquisition unit that acquires transaction data including a summary of the transaction;
The summary included in the transaction data acquired by the transaction data acquisition unit and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection unit that selects an item as an account item corresponding to the transaction data;
an accounting data generation unit that generates accounting data including information included in the transaction data and the account item selected by the item selection unit ;
The learning model generation unit includes the learning model that is included in the learning data in the input layer of the learning model, which is a neural network constituted by an input layer, an intermediate layer, and an output layer that indicates accuracy for each account item. After inputting only the summary of the transaction and learning the weight matrix to be applied to the input layer and the weight matrix to be applied to the intermediate layer, input the summary and the attribute data to the input layer, and then input the summary and the attribute data to the input layer. An accounting data conversion device that fixes the weight matrix to be applied and learns a weight matrix to be applied to the attribute data and a weight matrix to be applied to the intermediate layer . a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation unit that generates a learning model that outputs account items;
a learning model storage unit that stores the learning model generated by the learning model generation unit;
a transaction data acquisition unit that acquires transaction data including a summary of the transaction;
The summary included in the transaction data acquired by the transaction data acquisition unit and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection unit that selects an item as an account item corresponding to the transaction data;
an accounting data generation unit that generates accounting data including information included in the transaction data and the account item selected by the item selection unit;
Equipped with
The learning model generation unit inputs a summary of the transaction included in the learning data into the input layer of the learning model, which is a neural network composed of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. and the attribute data are input as a one-hot expression, the weight of the summary is set to 1, the weight of the attribute data is set to a value smaller than 1 or larger than 1, and a weight matrix to be applied to the input layer and a weight matrix to be applied to the middle layer. An accounting data conversion device that learns the weight matrix to be applied . a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation unit that generates a learning model that outputs account items;
a learning model storage unit that stores the learning model generated by the learning model generation unit;
a transaction data acquisition unit that acquires transaction data including a summary of the transaction;
The summary included in the transaction data acquired by the transaction data acquisition unit and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection unit that selects an item as an account item corresponding to the transaction data;
an accounting data generation unit that generates accounting data including information included in the transaction data and the account item selected by the item selection unit;
Equipped with
The learning model generation unit inputs a summary of the transaction included in the learning data into the input layer of the learning model, which is a neural network composed of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. , first attribute data indicating ON or OFF, second attribute data indicating ON or OFF, and correlation data indicating XOR of the first attribute data and the second attribute data are input. and learning a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer . Any one of claims 1 to 3 , wherein the subject attribute includes at least one of the business type of the subject of the transaction and business type identification data indicating whether the subject of the transaction is a sole proprietorship or a corporation. The accounting data conversion device according to item 1. The transaction attributes include at least one of income and expenditure identification data indicating whether the transaction is income or expenditure, the number of digits of the amount of the transaction, the origin of the transaction, and the tax classification of the transaction. The accounting data conversion device according to any one of Items 1 to 3 . computer,
a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation unit that generates a learning model that outputs account items;
a learning model storage unit that stores the learning model generated by the learning model generation unit;
a transaction data acquisition unit that acquires transaction data including a summary of the transaction;
The summary included in the transaction data acquired by the transaction data acquisition unit and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection unit that selects an item as an account item corresponding to the transaction data;
an accounting data generation unit that generates accounting data including information included in the transaction data and the account item selected by the item selection unit;
Equipped with
The learning model generation unit includes the learning model that is included in the learning data in the input layer of the learning model, which is a neural network constituted by an input layer, an intermediate layer, and an output layer that indicates accuracy for each account item. After inputting only the summary of the transaction and learning the weight matrix to be applied to the input layer and the weight matrix to be applied to the intermediate layer, input the summary and the attribute data to the input layer, and then input the summary and the attribute data to the input layer. An accounting data conversion device that fixes the weight matrix to be multiplied and learns the weight matrix to be multiplied by the attribute data and the weight matrix to be multiplied by the intermediate layer.
A program to function as computer,
a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation unit that generates a learning model that outputs account items;
a learning model storage unit that stores the learning model generated by the learning model generation unit;
a transaction data acquisition unit that acquires transaction data including a summary of the transaction;
The summary included in the transaction data acquired by the transaction data acquisition unit and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection unit that selects an item as an account item corresponding to the transaction data;
an accounting data generation unit that generates accounting data including information included in the transaction data and the account item selected by the item selection unit;
Equipped with
The learning model generation unit inputs a summary of the transaction included in the learning data into the input layer of the learning model, which is a neural network composed of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. and the attribute data are input as a one-hot expression, the weight of the summary is set to 1, the weight of the attribute data is set to a value smaller than 1 or larger than 1, and a weight matrix to be applied to the input layer and a weight matrix to be applied to the middle layer. Accounting data conversion device that learns the weight matrix to be multiplied by
A program to function as computer,
a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation unit that generates a learning model that outputs account items;
a learning model storage unit that stores the learning model generated by the learning model generation unit;
a transaction data acquisition unit that acquires transaction data including a summary of the transaction;
The summary included in the transaction data acquired by the transaction data acquisition unit and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection unit that selects an item as an account item corresponding to the transaction data;
an accounting data generation unit that generates accounting data including information included in the transaction data and the account item selected by the item selection unit;
Equipped with
The learning model generation unit inputs a summary of the transaction included in the learning data into the input layer of the learning model, which is a neural network composed of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. , first attribute data indicating ON or OFF, second attribute data indicating ON or OFF, and correlation data indicating XOR of the first attribute data and the second attribute data are input. and learns a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer.
A program to function as A computer-implemented accounting data conversion method, comprising:
a learning data storage step of storing learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction in a learning data storage unit;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation stage that generates a learning model that outputs account items;
a learning model storage step of storing the learning model generated in the learning model generation step in a learning model storage unit;
a transaction data acquisition stage of acquiring transaction data including a description of the transaction;
The summary included in the transaction data acquired in the transaction data acquisition step and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection step of selecting an item as an account item corresponding to the transaction data;
an accounting data generation step of generating accounting data including information included in the transaction data and the account item selected in the item selection step;
In the learning model generation step, the input layer of the learning model, which is a neural network constituted by an input layer, an intermediate layer, and an output layer indicating accuracy for each account item, includes the information contained in the learning data. After inputting only the summary of the transaction and learning the weight matrix to be applied to the input layer and the weight matrix to be applied to the intermediate layer, input the summary and the attribute data to the input layer, and then input the summary and the attribute data to the input layer. An accounting data conversion method , wherein the weighting matrix to be applied is fixed and the weighting matrix to be applied to the attribute data and the weighting matrix to be applied to the intermediate layer are learned . A computer-implemented accounting data conversion method, comprising:
a learning data storage step of storing learning data having a summary of the transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction in a learning data storage unit;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation stage that generates a learning model that outputs account items;
a learning model storage step of storing the learning model generated in the learning model generation step in a learning model storage unit;
a transaction data acquisition stage of acquiring transaction data including a description of the transaction;
The summary included in the transaction data acquired in the transaction data acquisition step and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection step of selecting an item as an account item corresponding to the transaction data;
an accounting data generation step of generating accounting data including information included in the transaction data and the account item selected in the item selection step;
Equipped with
The learning model generation step includes adding a summary of the transaction included in the learning data to the input layer of the learning model, which is a neural network composed of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. and the attribute data are input as a one-hot expression, the weight of the summary is set to 1, the weight of the attribute data is set to a value smaller than 1 or larger than 1, and a weight matrix to be applied to the input layer and a weight matrix to be applied to the middle layer. An accounting data conversion method that learns the weight matrix to be applied. A computer-implemented accounting data conversion method, comprising:
a learning data storage step of storing learning data having a summary of the transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction in a learning data storage unit;
By machine learning using the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and the transaction corresponding to the transaction is performed. a learning model generation stage that generates a learning model that outputs account items;
a learning model storage step of storing the learning model generated in the learning model generation step in a learning model storage unit;
a transaction data acquisition stage of acquiring transaction data including a description of the transaction;
The summary included in the transaction data acquired in the transaction data acquisition step and the attribute data of the transaction are input into the learning model stored in the learning model storage unit, and the account output from the learning model is calculated. an item selection step of selecting an item as an account item corresponding to the transaction data;
an accounting data generation step of generating accounting data including information included in the transaction data and the account item selected in the item selection step;
Equipped with
The learning model generation step includes adding a summary of the transaction included in the learning data to the input layer of the learning model, which is a neural network composed of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. , first attribute data indicating ON or OFF, second attribute data indicating ON or OFF, and correlation data indicating XOR of the first attribute data and the second attribute data are input. and learning a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer. a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. and a model generation unit that generates a learning model to be output ,
The model generation unit is configured to generate the transactions included in the learning data in the input layer of the learning model, which is a neural network constituted by an input layer, an intermediate layer, and an output layer that indicates accuracy for each account item. After inputting only the abstract and learning a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer, input the abstract and the attribute data to the input layer and apply the weight to the abstract. A learning device that fixes the weight matrix and learns a weight matrix applied to the attribute data and a weight matrix applied to the intermediate layer . a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. a model generation unit that generates a learning model to output;
Equipped with
The model generation unit inputs a summary of the transaction included in the learning data and the input layer of the learning model, which is a neural network configured of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. The attribute data is input as a one-hot expression, the weight of the summary is set to 1, the weight of the attribute data is set to a value smaller than 1 or larger than 1, and the weight matrix applied to the input layer and the weight matrix applied to the intermediate layer are input. A learning device that learns weight matrices. a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute of the transaction and a transaction attribute;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. a model generation unit that generates a learning model to output;
Equipped with
The model generation unit includes a summary of the transaction included in the learning data and a summary of the transaction included in the learning data in the input layer of the learning model, which is a neural network composed of an input layer, a middle layer, and an output layer indicating accuracy for each account item. , input first attribute data indicating ON or OFF, second attribute data indicating ON or OFF, and correlation data indicating XOR of the first attribute data and the second attribute data. , a learning device that learns a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer. computer,
a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction; Based on the stored learning data, a learning model is generated that takes as input a transaction summary and attribute data including at least one of a subject attribute and a transaction attribute of the transaction and outputs an account item corresponding to the transaction. model generator
Equipped with
The model generation unit is configured to generate the transactions included in the learning data in the input layer of the learning model, which is a neural network constituted by an input layer, an intermediate layer, and an output layer that indicates accuracy for each account item. After inputting only the abstract and learning a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer, input the abstract and the attribute data to the input layer and apply the weight to the abstract. A learning device that fixes the weight matrix and learns a weight matrix to be applied to the attribute data and a weight matrix to be applied to the intermediate layer.
A program to function as computer,
a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. Model generation unit that generates the learning model to be output
Equipped with
The model generation unit inputs a summary of the transaction included in the learning data and the input layer of the learning model, which is a neural network configured of an input layer, an intermediate layer, and an output layer indicating accuracy for each account item. The attribute data is input as a one-hot expression, the weight of the summary is set to 1, the weight of the attribute data is set to a value smaller than 1 or larger than 1, and the weight matrix applied to the input layer and the weight matrix applied to the intermediate layer are input. A learning device that learns weight matrices
A program to function as computer,
a learning data storage unit that stores learning data having a summary of a transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. Model generation unit that generates the learning model to be output
Equipped with
The model generation unit includes a summary of the transaction included in the learning data and a summary of the transaction included in the learning data in the input layer of the learning model, which is a neural network composed of an input layer, a middle layer, and an output layer indicating accuracy for each account item. , input first attribute data indicating ON or OFF, second attribute data indicating ON or OFF, and correlation data indicating XOR of the first attribute data and the second attribute data. , a learning device that learns a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer.
A program to function as A learning method performed by a computer, the method comprising:
a storing step of storing learning data including a summary of the transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction in a learning data storage unit;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. a model generation stage for generating a learning model to be output ;
In the model generation stage, the transaction included in the learning data is added to the input layer of the learning model, which is a neural network composed of an input layer, an intermediate layer, and an output layer that indicates accuracy for each account item. After inputting only the abstract and learning a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer, input the abstract and the attribute data to the input layer and apply the weight to the abstract. A learning method that fixes the weight matrix and learns a weight matrix to be applied to the attribute data and a weight matrix to be applied to the intermediate layer . A learning method performed by a computer, the method comprising:
a storing step of storing learning data including a summary of the transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction in a learning data storage unit;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. a model generation stage that generates a learning model to be output;
Equipped with
In the model generation stage, the input layer of the learning model, which is a neural network composed of an input layer, a middle layer, and an output layer indicating accuracy for each account item, is provided with a summary of the transaction included in the learning data and The attribute data is input as a one-hot expression, the weight of the summary is set to 1, the weight of the attribute data is set to a value smaller than 1 or larger than 1, and the weight matrix applied to the input layer and the weight matrix applied to the intermediate layer are input. A learning method for learning weight matrices. A learning method performed by a computer, the method comprising:
a storing step of storing learning data including a summary of the transaction, an account item corresponding to the transaction, and attribute data including at least one of a subject attribute and a transaction attribute of the transaction in a learning data storage unit;
Based on the learning data stored in the learning data storage unit, a summary of the transaction and attribute data including at least one of the subject attribute and the transaction attribute of the transaction are input, and an account item corresponding to the transaction is determined. a model generation stage that generates a learning model to be output;
Equipped with
In the model generation stage, the input layer of the learning model, which is a neural network composed of an input layer, a middle layer, and an output layer indicating accuracy for each account item, is provided with a summary of the transaction included in the learning data. , input first attribute data indicating ON or OFF, second attribute data indicating ON or OFF, and correlation data indicating XOR of the first attribute data and the second attribute data. , a learning method for learning a weight matrix to be applied to the input layer and a weight matrix to be applied to the intermediate layer.

Images