JP7515037B1 - Risk Management Support System - Google Patents

Abstract

[Problem] To provide a risk management support system that can automatically detect if a document under review contains any accounting or tax issues that require consultation with an expert. [Solution] A risk management support system 1 that detects if a document under review contains any accounting or tax issues that require consultation with an expert. The system includes a database section 22 that stores information about accounting or tax issues, a document acquisition section 31 that imports the document under review, a related information extraction section 33 that extracts related information related to the document under review from the database section by comparing the wording contained in the document under review with the information in the database section, and an issue determination section 34 that determines whether or not the document under review contains any issues based on the wording contained in the document under review and the extracted related information. [Selected Figure] Figure 1

Description

The present invention relates to a risk management support system that detects if a document under review contains issues that require consultation with an accounting or tax expert.

As disclosed in Patent Documents 1 and 2, a system has been developed that enables even inexperienced or inexperienced personnel to create or review contracts efficiently and accurately.

For example, the document review support device in Patent Document 1 displays the document to be reviewed and a checklist including check items related to risks at the time of contracting on the same screen, and creates a work record including the contents entered as check results for the check items included in the checklist.

On the other hand, in the field of accounting and taxation, there are increasing opportunities for companies to make advanced judgments in their accounting and tax procedures due to complex transactions and the introduction of new accounting standards and tax laws. When complex accounting and tax procedures are required, companies may seek advice from accounting and taxation experts.

JP 2023-64113 A JP 2023-88560 A

However, if company personnel fail to detect accounting or tax issues that require consultation with an expert, there is a risk that they will carry out incorrect accounting and tax procedures without obtaining advice from an expert. There are limitations to the reduction of such risks using systems such as those disclosed in Patent Documents 1 and 2, which are targeted at contracts with clear check items.

The present invention aims to provide a risk management support system that can automatically detect if there are any issues in the documents under review that require consultation with an accounting or tax expert.

In order to achieve the above-mentioned objective, the risk management support system of the present invention is a risk management support system that detects if a document under review contains an accounting or tax issue that requires consultation with an expert, and is characterized by comprising a database unit that stores information related to accounting or tax issues, a document acquisition unit that imports the document under review, a related information extraction unit that extracts related information related to the document under review from the database unit by comparing the wording contained in the document under review with the information in the database unit, and an issue determination unit that determines whether the document under review contains the issue based on the wording contained in the document under review and the extracted related information.

Here, the database unit may be configured to record cases in which transactions have accounting or tax considerations, in association with the issues and accounting or tax standards or laws.

In the risk management support system of the present invention thus configured, the related information extraction unit compares the information in the database unit, which contains information related to accounting or tax issues, with documents to be considered, such as minutes and approval documents, imported by the document acquisition unit, to extract related information. Then, based on the extracted related information and the documents to be considered, the issue determination unit determines whether there are any issues regarding accounting or tax that require consultation with an expert.

Therefore, if there are any accounting or tax issues in the documents under review that require consultation with an expert, it can be automatically detected. Furthermore, if the database section contains cases where transactions have accounting or tax considerations and is linked to the issues and standards, etc., it will be possible to accurately detect issues that are relevant to practical work.

1 is a block diagram illustrating a configuration of a risk management support system according to an embodiment of the present invention. FIG. 2 is an explanatory diagram of case data stored in a database unit. FIG. 4 is an explanatory diagram of reference data stored in a database unit. FIG. 11 is an explanatory diagram of reference change point data stored in the database unit. FIG. 11 is an explanatory diagram of case data extracted as related information. FIG. 11 is an explanatory diagram of reference data extracted as related information. FIG. 1 is an explanatory diagram illustrating an example of a dataset used for machine learning.

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram illustrating the configuration of a risk management support system 1 according to the present embodiment.

The risk management support system 1 of this embodiment is a device that enables company personnel to seek advice from experts when complex accounting or tax procedures are required in the fields of accounting or taxation, even if they are unable to detect that there are accounting or tax issues that require consultation. The following explanation will be given mainly using the field of accounting as an example.

In the risk management support system 1 of this embodiment, as shown in FIG. 1, calculation processing or acquisition of calculation processing results is performed by the calculation processing unit 3 or the like based on data input from the input device 2, memory unit (4, 41), various databases (21, 22), etc., and the results are output to the display device 5 or notification device 51.

The input device 2 for inputting various data is a data input means connected to or equipped on a personal computer (PC), a notebook computer, a tablet terminal, a wearable terminal, a smartphone, etc. Examples of the input device 2 include a keyboard, a mouse, a touch panel, a touch pad, a scanner, a microphone for voice input, a camera, etc.

On the other hand, the display device 5 can be a liquid crystal display, an organic EL (Electro-Luminescence) display, a printer, etc., and the notification device 51 can be an electronic method such as a mailer that sends e-mail, various chat tools, and various message tools.

The hardware of the calculation processing unit 3 is composed of a CPU (Central Processing Unit), an MPU (Micro-processing unit), a GPU (Graphics Processing Unit), etc., and is equipped with memories such as RAM (Random Access Memory) and ROM (Read Only Memory).

The arithmetic processing unit 3 can execute various functions by applications installed on the computer. In addition, part or all of the arithmetic processing unit 3 can be executed by a server connected via a network such as the Internet. The network is composed of part or all of the Internet, a Wide Area Network (WAN), a wired Local Area Network (LAN), a wireless LAN (Wi-Fi), a provider device, a wireless base station, a dedicated line, etc.

The storage unit 4 is a storage medium that records data used in processing in the arithmetic processing unit 3 and data generated by the arithmetic processing, and includes a hard disk, a solid state drive (SSD), a flash memory (such as an SD memory card), a magnetic disk, and an optical disk. In addition, an external online storage (cloud storage) such as a server connected via a network can also be used as the storage unit 4. Note that the area of the storage medium that stores the machine learning model is described as being distinguished by the reference numeral 41, but as hardware it may be integrated with the storage unit 4 or may be separate.

The calculation processing unit 3 of the risk management support system 1 is mainly composed of a document acquisition unit 31 that imports documents to be considered, a database update unit 32, a related information extraction unit 33 that extracts related information related to the documents to be considered, an issue determination unit 34 that determines whether or not there are issues in the documents to be considered, and a notification sending unit 35.

The document to be reviewed in this embodiment is any document that may contain accounting issues that require consultation with an expert. For example, documents such as minutes and approval documents prepared in relation to corporate transactions are examples of documents to be reviewed.

Documents to be reviewed, such as minutes, can be stored in a database that serves as the document storage unit 21 as needed during daily operations using a management system or the like. The document acquisition unit 31 can automatically import documents stored in the document storage unit 21 and determine whether or not there are any points of contention, or it can import appropriately selected documents from the document storage unit 21 or the like and determine whether or not there are any points of contention.

The database unit 22 is a database that stores information related to accounting issues. For example, cases that summarize corporate transactions and the accounting issues involved, accounting standards and laws, etc. are recorded in the database unit 22. Accounting standards include accounting standards, auditing standards, and information on changes to accounting standards, auditing standards, etc.

Figure 2 is an explanatory diagram of case data stored in the database unit 22. For example, when a maintenance service is provided with the sale of an item, even if the customer pays in a lump sum, the sales from the maintenance service may be required to be recorded over the maintenance period rather than in a lump sum. In such a case, if the company's personnel are not aware that such accounting procedures are required and the same procedures are followed as for previous sales of items, this will result in incorrect accounting procedures, and the case is recorded in the database unit 22 as a controversial case.

In the case example table shown in Figure 2, the details of the transaction, etc. are recorded, and the associated issues (accounting standards related to revenue recognition) and the titles of related accounting standards, etc. that should be referenced for those issues are recorded. It is also possible to record keywords associated with the case examples.

As shown in FIG. 2, there are many accounting issues, and it is possible to select relevant accounting standards to refer to for each issue. In addition, since the transaction contents may differ even for the same issue, it is desirable to accumulate many cases with different transaction contents in the database unit 22. Keywords to be associated with the cases can be selected from, for example, the transaction contents, the issues, and terms described in the relevant accounting standards.

Figure 3 is an explanatory diagram of the standards data stored in the database unit 22. In the standards table, the type of accounting standard, auditing standard, etc., the title indicating the content of the standard, and the content of the standard are recorded in association with each other.

For example, as shown in FIG. 3, there are many accounting standards and guidelines, such as accounting standards for revenue recognition, application guidelines for accounting standards for impairment of fixed assets, and accounting standards for financial instruments. In addition, there are many audit standards and guidelines, such as audits of accounting estimates and special considerations in group audits. By storing this standards data in the database unit 22, it becomes possible to easily extract the contents of accounting standards and other information associated with cases as related information.

When new cases that involve issues such as those described above arise, or when accounting standards, auditing standards, etc. change, the database update unit 32 adds that information to the database unit 22. The database update unit 32 can also delete cases, standards, etc. that are no longer appropriate to store in the database unit 22.

Figure 4 is an explanatory diagram of the criteria change data stored in the database unit 22. In the criteria change table, the content before the change, the content after the change, and the points of the change are recorded along with information on the change date in association with the type and title of the criteria. The change date information includes the revision date and the application date, and it is also possible to select related information to be extracted based on the relationship between the document under consideration and the application date.

The related information extraction unit 33 compares the wording contained in the document to be examined with the information in the database unit 22 to extract related information related to the document to be examined from the database unit 22. In other words, the related information extraction unit 33 obtains information on cases, accounting standards, etc. related to the contents written in the minutes of meetings, etc. from the database unit 22.

When the related information extraction unit 33 obtains information, it can use keyword searches or vector searches to obtain related cases, related accounting standards, etc. It can also use topic models, etc.

As shown in FIG. 2, for example, if keywords are linked and recorded in the case table stored in the database unit 22, when the minutes or the like contain the keywords, the keywords can be extracted as related cases. For example, when the keyword "maintenance service" is included, the "Accounting Standard for Revenue Recognition" is linked to the case, so that the case and related accounting standards can be extracted as related information along with the contents of the "Accounting Standard for Revenue Recognition" in the standards table shown in FIG. 3 (see FIGS. 5 and 6). In addition, by keyword search, it is also possible to directly extract information on related accounting standards and changes to accounting and auditing standards from the standards data (FIG. 3) and standard change data (FIG. 4) as related information.

When the related information extraction unit 33 is made to perform a vector search, the text is vectorized and the vector similarity is used to match the text. For example, information such as cases and accounting standards stored in advance in the database unit 22 is vectorized using a machine learning model recorded in the learning model storage unit 41 or the like. Then, data such as minutes of meetings acquired by the document acquisition unit 31 from the document storage unit 21 is converted into vectors, and by comparing these vectorized data, related cases and related accounting standards stored in the database unit 22 are identified and extracted as related information.

A machine learning model can also be used to extract related information by the related information extraction unit 33. For example, as shown in FIG. 7, machine learning is performed on information such as past approval documents and information such as corresponding cases as a dataset for training data, and the machine learning model generated thereby is recorded in the learning model storage unit 41. In this way, cases related to newly acquired documents under consideration can be extracted from the database unit 22. A machine learning model provided by an external organization such as OpenAI can also be extracted from the database unit 22 via an API (Application Programming Interface). In this way, at least one of information on related cases, issues, related accounting standards, and changes in standards can be extracted as related information.

In addition, a model for the related information extraction unit 33 that can determine related cases, etc. from minutes, etc. may be constructed by fine-tuning existing large-scale language models (LLMs).

The reason why the related information extraction unit 33 performs the process of extracting related information from the database unit 22 in this way is that in the field of accounting, where accounting standards and the like are frequently updated, it is a heavy burden to always machine-learn the latest accounting standards. Therefore, after narrowing down the information to be used as reference for related information, the issue determination unit 34 determines whether there is an issue.

The issue determination unit 34 determines whether or not there are any issues in the document under review based on the wording contained in the document under review and the related information extracted by the related information extraction unit 33. For example, by setting essential keywords or using processing with a large-scale language model (LLM), it determines whether the contents of the minutes or the like include accounting issues that should be discussed with an accounting firm or auditing corporation.

For example, when using keywords to have the issue determination unit 34 determine whether an issue exists, multiple keywords can be set and it can be determined that there is an issue only for documents under review that contain all of those keywords. For example, by specifying two keywords, "sales of new products" and "maintenance services," if both keywords are included in a document, it can be determined that there is an issue in the minutes, etc. that should be discussed.

In addition, the issue determination unit 34 not only compares the document under review with related information, but also sets a threshold amount based on the size of the company being audited, making it possible to narrow down the items whose impact amount exceeds the threshold amount when determining whether they contain accounting issues. Here, when calculating the impact amount, exchange rate information can be used for conversion.

On the other hand, information such as minutes and related information can be input into a natural language processing model such as a large-scale language model to determine whether the contents of the minutes include points of discussion, etc.

For example, the following prompt can be entered into a large-scale language model (LLM):
"Refer to the following related accounting standards and other documents to determine whether the request includes accounting issues. If it does, reply "Yes" and state what the issues are. If it does not, reply "No." If you know the monetary impact, also reply with the amount. If you do not know, reply "Unknown."
#### Example Answer
## If there is any impact,
“Impact”: “Yes”, “Issues”: “This is a request regarding the decline in profitability of fixed assets, and it is necessary to consider whether or not to write down the fixed assets”, “Impact amount”: “100,000,000 yen”}
## If there is no impact
{"Impact":"No"}
##### Accounting standards, etc. (accounting standards entered as related information)
Accounting Standards for Revenue Recognition Definitions of Terms...
##### Request document (document to be reviewed)
Regarding the launch of new product XX
New product A will be released on XXXX/X/XXXX.
The unit price is 100,000 yen, and the annual sales target is 100 million yen.
This product comes with a three-year maintenance service.

Examples of answers that large-scale language models can give include the following:
-{“Impact”:”Yes”, “Issue”:”This is a request for approval regarding the release of a new product that includes maintenance services, and it is necessary to consider revenue recognition for the maintenance services, “Impact amount”:”100,000,000 yen”}
-{“Impact”:”No”}

The notification sending unit 35 displays the judgment result (answer) of the issue judgment unit 34 on the display device 5, and depending on the settings, if the judgment result includes accounting issues, etc., the notification device 51 can automatically send a notification based on the judgment result to a pre-set notification destination such as an accounting firm or auditing corporation.

If automatic notification is not provided to accounting firms, etc., a configuration is possible in which, for example, a company employee determines whether or not notification is necessary based on the judgment result (answer) of the issue judgment unit 34, and only when it is determined that notification is necessary and an instruction to send is given to the notification sending unit 35, notification is sent to the accounting firm, etc. by the notification device 51.

A method of using the risk management support system 1 according to this embodiment will be described below.
First, in the course of daily business, documents such as meeting minutes and approval documents are accumulated in the document accumulation unit 21 (see FIG. 1). Among these documents are documents that may contain issues that require the consultation of an accounting expert.

On the other hand, the data in the database unit 22 is updated by the database update unit 32 when a new case arises involving a corporate transaction or the like and an associated accounting matter (see FIG. 2), or when there is a change in accounting or auditing standards (see FIG. 4).

The risk management support system 1 of this embodiment is a device that allows a company's accounting staff to seek advice from an expert even if they are unable to detect that there are accounting issues that require consultation. For this reason, one way of using it is to have the document acquisition unit 31 automatically acquire newly recorded documents in the document storage unit 21 one after another and make a judgment. Another way of using it is to have the document acquisition unit 31 acquire documents from related departments all at once, such as before the end of the fiscal year, through the operation of the accounting staff.

For documents (documents to be considered) imported into the risk management support system 1 by the document acquisition unit 31, the related information extraction unit 33 extracts related information from the database unit 22. Related information is extracted from information with high similarity, for example, using an index called similarity. The number of related cases, related accounting standards, etc. to be extracted as related information (order of similarity), the similarity threshold, etc. can be set arbitrarily.

On the other hand, if the related information extraction unit 33 extracts related information such as related cases and related accounting standards, the issue determination unit 34 determines whether the document under review contains any accounting issues that should be discussed with an accounting firm, auditing corporation, etc.

If the issue determination unit 34 determines that there is an accounting issue that should be discussed with an accounting firm or the like, the notification sending unit 35 and notification device 51 send the document to be considered to the accounting firm or the like. Note that before sending to the accounting firm or the like, the accounting staff may judge the result of the issue determination unit 34 displayed on the display device 5 and ultimately decide whether or not to send the document.

Next, the operation of the risk management support system 1 of this embodiment will be described.
The risk management support system 1 of this embodiment configured as described above extracts related information by comparing the information in the database unit 22, which stores information on accounting issues, with documents to be considered, such as minutes and approval documents, captured by the document acquisition unit 31, in the related information extraction unit 33. Then, based on the extracted related information and the documents to be considered, the issue determination unit 34 determines whether there are any accounting issues that should be discussed with an expert.

Therefore, if there are any accounting issues in the documents under review that require consultation with an expert, it can be automatically detected. Furthermore, if the database unit 22 records cases in which accounting issues for transactions need to be considered, linked to the issues and standards, etc., it becomes possible to accurately detect issues that are relevant to practical matters.

In addition, by providing a notification sending unit 35 that sends a notification based on the determination result to a pre-set notification destination such as an accounting firm when the issue determination unit 34 determines that there is an issue, the relevant institutions can quickly and efficiently begin considering ways to reduce risks.

In addition, if the issue determination unit 34 determines whether there are any issues using information on the base amount set by the company that manages the document under review, it will be possible to review appropriate issues according to the scale of the company's profits, assets, etc.

Furthermore, if the relevant information extraction unit 33 extracts relevant information from the database unit 22 and then the issue determination unit 34 determines whether there is an issue, it becomes easy to continuously import the latest accounting standards, etc., and it becomes possible to reduce risks in line with practical applications.

The embodiment of the present invention has been described in detail above with reference to the drawings, but the specific configuration is not limited to this embodiment, and design changes that do not deviate from the gist of the present invention are included in the present invention.

For example, in the above embodiment, the explanation is mainly given using the field of accounting as an example, but the present invention is not limited to this. When determining whether or not there are tax issues, cases having tax considerations can be stored in the database unit 22 in association with the issues, tax laws, changes to tax laws, etc.

1: Risk management support system 21: Document storage unit 22: Database unit 31: Document acquisition unit 33: Related information extraction unit 34: Issue determination unit 35: Notification transmission unit

Claims (6)

A risk management support system that detects if a document to be examined contains any issues that require consultation with an accounting or tax expert,
a database section containing information relating to accounting or tax issues;
a document acquisition unit for acquiring the document to be examined;
a related information extraction unit that extracts related information related to the document to be examined from the database unit by comparing a wording included in the document to be examined with information in the database unit;
A risk management support system characterized by having an issue determination unit that determines whether or not the document under review contains the issue based on the wording contained in the document under review and the extracted related information. The risk management support system according to claim 1, characterized in that the database section stores cases in which transactions have accounting or tax considerations, linked to the issues and accounting or tax standards or laws. The risk management support system according to claim 1 or 2, further comprising a notification sending unit that sends a notification based on the judgment result to a preset notification destination when the issue judgment unit judges that the issue exists. The risk management support system according to claim 1 or 2, characterized in that the document to be considered is a meeting minutes or a request for approval. The risk management support system according to claim 2, characterized in that the related information extraction unit extracts at least one of the cases, the issues, and the accounting or taxation standards or laws as the related information by keyword search or vector search. The risk management support system according to claim 1 or 2, characterized in that the issue determination unit also uses information on the standard amount set for the company that manages the document under review to make the determination.