Intelligent financial management system and method based on multi-modal large model

Abstract

The invention provides an intelligent financial management system and method based on a multi-modal large model, and the method comprises the following steps: obtaining a paper original voucher picture through photographing or scanning; extracting specified information in the picture through a multi-modal large model, wherein the specified information comprises a voucher type, an amount and a cause; inputting the specified information into a trained multi-mode large model, judging a financial accounting subject to which the voucher belongs based on a cue word, and generating an accounting voucher; the generated accounting voucher is submitted to financial staff for auditing, and the accounting voucher is confirmed or modified; and storing the confirmed or modified accounting voucher into a financial management background to generate a financial statement. According to the invention, the working efficiency and accuracy of financial accounting bookkeeping are greatly improved, and the level of financial management is improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent financial accounting, and in particular to an intelligent financial management system and method based on a multimodal large model. Background Art

[0002] The existing financial accounting is based on manual accounting. When financial personnel conduct accounting and operations, they need to compile the original vouchers into accounting vouchers that meet financial requirements according to accounting standards and personal work experience, and manually enter them into the financial system for review and accounting. This accounting method has the following problems:

[0003] (1) Due to factors such as different professional levels of financial personnel, differences in personal accounting habits, personnel turnover, and differences in understanding of accounting standards, accounting results are inconsistent and non-standard.

[0004] (2) Inaccurate voucher data may be entered due to factors such as the financial staff's computer proficiency, data entry errors, and staff fatigue when entering large amounts of data.

[0005] (3) Financial statements cannot be automatically generated, resulting in poor financial management effectiveness and inconvenience in reviewing and evaluating financial data.

[0006] (4) Manual accounting methods involve a lot of mechanical and repetitive work, high labor and time costs, and low work efficiency. Summary of the invention

[0007] The purpose of the present invention is to provide an intelligent financial management system and method based on a multimodal large model to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: an intelligent financial management system and method based on a multimodal large model, comprising the following steps:

[0009] S1. Obtain the image of the original paper voucher by taking a photo or scanning;

[0010] S2. Extracting designated information from the image through a multimodal large model, wherein the designated information includes voucher type, amount, and reason;

[0011] S3, inputting the specified information into the trained multimodal large model, judging the financial accounting subject to which the voucher belongs based on the prompt word, and generating an accounting voucher;

[0012] S4. Submit the generated accounting vouchers to the financial staff for review and confirmation or modification of the accounting vouchers;

[0013] S5. Save the confirmed or modified accounting vouchers into the financial management background and generate financial statements.

[0014] Furthermore, the types of paper original documents in step S1 include receipts, invoices, approval forms, bank receipts, contract agreements, meeting minutes and work records.

[0015] Furthermore, the step of constructing a method for generating accounting vouchers using a multimodal big model also includes: integrating the multimodal big model with natural language processing and image processing technology, performing semantic analysis on text data, extracting key information including amount, transaction type, and reason, and identifying key entities in the text including company name and account information.

[0016] Furthermore, the step of constructing a method for generating accounting vouchers using a multimodal large model also includes: performing multimodal data fusion and correlating text data, image data, and numerical data to ensure data consistency and accuracy of the accounting vouchers.

[0017] Furthermore, the step of constructing a multimodal large model to generate accounting vouchers also includes: verifying the uniqueness and authenticity of the invoice.

[0018] Furthermore, the construction of the multimodal large model trained in step S3 specifically includes the following steps:

[0019] S31. Data collection: Collect financial data including different types of original vouchers, including receipts, invoices, approval forms, bank receipts, contract agreements, meeting minutes and work records;

[0020] S32. Data cleaning and processing: Clean and process the collected financial data to ensure that the data input into the multimodal large model is accurate and consistent, including deleting duplicate values, processing missing values, processing outliers, and consistent processing;

[0021] S33, model selection: select a suitable training model as the basis according to the number of parameters and inference speed;

[0022] S34, Model training: Design training tasks, including joint representation learning of images and texts, so that the model can understand and process data of multiple modalities. Based on the training tasks, fine-tune the model for the specific task of automatic financial accounting, design fine-tuning tasks, train the model using labeled financial data, select appropriate loss functions and optimization algorithms, adjust model parameters, and improve the accuracy of the model.

[0023] S35. Model evaluation, optimization and deployment: Use evaluation indicators to evaluate the performance of the model on the validation set. Adjust and optimize the model based on the evaluation results. For the result data generated by the model that does not meet the expectations, use it as historical training data to participate in the model training. Deploy the trained model to actual applications to ensure that it can receive actual financial voucher input and automatically generate accounting vouchers.

[0024] Furthermore, the step of constructing and using a multimodal large model to generate accounting vouchers also includes: construction of vector data, including feature extraction and vectorization, and feature extraction includes data conversion, feature selection, and feature calculation.

[0025] Furthermore, the step of constructing a multimodal large model to generate accounting vouchers also includes: processing of RAG (retrieval enhanced generation) data, including vector storage and management and performance optimization, and performance optimization includes dimensionality reduction, parameter tuning, verification of vector quality, and testing performance.

[0026] Furthermore, the step S4 submits the generated accounting voucher to the financial personnel for review, including: the financial personnel checks the consistency of the generated accounting voucher with the original voucher to ensure the preparation and compliance of the accounting voucher; if there is an error, the financial personnel modifies and corrects the accounting voucher according to the corresponding original voucher; if there is no error, the confirmed financial accounting voucher is issued.

[0027] Furthermore, the step S5 generates financial statements from the financial accounting vouchers in the financial management background, including: registering detailed ledgers and general ledgers according to the accounting vouchers; preparing financial statements according to the general ledger and detailed ledgers, and the financial statements include: balance sheet, income and income distribution statement.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] (1) The intelligent financial management system and method based on the multimodal large model automatically identifies the contents of various original vouchers, accurately extracts key economic behavior elements, automatically verifies whether the business process is compliant, and completes intelligent calculation and professional processing of economic behavior to generate standardized accounting vouchers that meet the specifications, and finally classifies and stores the original vouchers. Implement standardized processes to ensure the standardization of financial activities.

[0030] (2) The intelligent financial management system and method based on the multimodal large model, through natural language processing capabilities and image processing technology, enables the model to accurately identify and parse various financial vouchers, significantly reduce human errors, and ensure the consistency of financial data and the accuracy of accounting vouchers.

[0031] (3) The intelligent financial management system and method based on the multimodal large model can seamlessly integrate various financial data, generate various financial statements in real time, realize centralized management and intelligent analysis of data, and thus improve the efficiency of digital application and management.

[0032] (4) The intelligent financial management system and method based on the multimodal large model deeply understands and analyzes national accounting standards, systems and complex economic behaviors, relies on samples and real-account accounting entries for deep learning training iterations, solves the accounting subject methods of different industry types, reduces the learning cost of financial personnel, and ensures the accuracy of accounting vouchers in different industries.

[0033] (5) The intelligent financial management system and method based on the multimodal large model can significantly reduce the workload of financial personnel in manual data entry by automatically generating accounting vouchers and processing financial data, thus reducing time and labor costs and improving financial work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart of the intelligent financial management system and method based on multimodal large model of the present invention;

[0035] Figure 2 A flowchart of model training construction for the intelligent financial management system and method based on a multimodal large model of the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] See also Figure 1 and Figure 2 The present invention provides a technical solution: an intelligent financial management system and method based on a multimodal large model, comprising the following steps:

[0038] S1. Obtain the image of the original paper voucher by taking a photo or scanning.

[0039] The types of paper original documents include receipts, invoices, approval forms, bank receipts, contract agreements, meeting minutes and work records; the types of invoices include but are not limited to special value-added tax invoices, general value-added tax invoices, train tickets, flight itineraries, tickets, etc.

[0040] Use external devices to take photos or scan paper original vouchers. External devices include mobile phones, high-definition scanners, scanners, etc. The equipment is also connected to the financial system data to facilitate the transmission of electronic image information to the system for storage and backup, and generate corresponding codes.

[0041] S2. Extracting designated information from the image through a multimodal large model, wherein the designated information includes voucher type, amount, and reason;

[0042] The multimodal large model uses bill verification technology, combined with predefined verification prompts and function calling technology, to verify key information in the document, such as amount, invoice authenticity, invoice duplication, and bank receipt duplication. After the verification is successfully passed, the model will further use the specific prompts extracted from the cause information to extract key information such as the cause description, specific amount, and bank account related information from the document. The key information will be organized and responded to the user interface, and the user will double-confirm the generated cause information to ensure the accuracy of the cause and amount.

[0043] S3, inputting the specified information into the trained multimodal large model, judging the financial accounting subject to which the voucher belongs based on the prompt word, and generating an accounting voucher;

[0044] S4. Submit the generated accounting vouchers to the financial staff for review and confirmation or modification of the accounting vouchers;

[0045] The financial staff will check the consistency between the generated accounting vouchers and the original vouchers to ensure the preparation and compliance of the accounting vouchers; if there are any errors, the financial staff will modify and correct the accounting vouchers based on the corresponding original vouchers; if there are no errors, the confirmed financial accounting vouchers will be issued.

[0046] S5. Save the confirmed or modified accounting vouchers into the financial management background and generate financial statements.

[0047] Register detailed ledgers and general ledgers according to the accounting vouchers; prepare financial statements based on the general ledger and detailed ledgers, and the financial statements include: balance sheet, income and income distribution statement.

[0048] In this embodiment,

[0049] The step of constructing a method for generating accounting vouchers using a multimodal big model also includes: integrating the multimodal big model with natural language processing and image processing technology, performing semantic analysis on text data, extracting key information including amount, transaction type, and reason, and identifying key entities in the text including company name and account information.

[0050] The step of constructing a multimodal large model to generate accounting vouchers also includes: performing multimodal data fusion and correlating text data, image data and numerical data to ensure data consistency and accuracy of accounting vouchers.

[0051] In the present invention, the construction of the multimodal large model trained in step S3 specifically includes the following steps:

[0052] S31. Data collection: Collect financial data including different types of original vouchers, including receipts, invoices, approval forms, bank receipts, contract agreements, meeting minutes and work records;

[0053] S32. Data cleaning and processing: Clean and process the collected financial data to ensure that the data input into the multimodal large model is accurate and consistent, including deleting duplicate values, processing missing values, processing outliers, and consistent processing;

[0054] S33, model selection: select a suitable training model as the basis according to the number of parameters and inference speed;

[0055] S34, Model training: Design training tasks, including joint representation learning of images and texts, so that the model can understand and process data of multiple modalities. Based on the training tasks, fine-tune the model for the specific task of automatic financial accounting, design fine-tuning tasks, train the model using labeled financial data, select appropriate loss functions and optimization algorithms, adjust model parameters, and improve the accuracy of the model.

[0056] S35. Model evaluation, optimization and deployment: Use evaluation indicators to evaluate the performance of the model on the validation set. Adjust and optimize the model based on the evaluation results. For the result data generated by the model that does not meet the expectations, use it as historical training data to participate in the model training. Deploy the trained model to actual applications to ensure that it can receive actual financial voucher input and automatically generate accounting vouchers.

[0057] The step of constructing and using a multimodal large model to generate accounting vouchers also includes: construction of vector data, including feature extraction and vectorization, and feature extraction includes data conversion, feature selection, and feature calculation.

[0058] The step of constructing a multimodal large model to generate accounting vouchers also includes: processing of RAG (retrieval enhanced generation) data, including vector storage and management and performance optimization, and performance optimization includes dimensionality reduction, parameter tuning, verification of vector quality, and testing performance.

[0059] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in the field. Although the embodiments of the present invention have been shown and described, it is understood by ordinary technicians in the field that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the attached claims and their equivalents.

Claims

1. An intelligent financial management system and method based on a multimodal large model, characterized in that: The following steps are involved: S1. Obtain the image of the original paper voucher by taking a photo or scanning; S2. Extracting designated information from the image through a multimodal large model, wherein the designated information includes voucher type, amount, and reason; S3, inputting the specified information into the trained multimodal large model, judging the financial accounting subject to which the voucher belongs based on the prompt word, and generating an accounting voucher; S4. Submit the generated accounting vouchers to the financial staff for review and confirmation or modification of the accounting vouchers; S5. Save the confirmed or modified accounting vouchers into the financial management background and generate financial statements.

2. According to claim 1, an intelligent financial management system based on a multimodal large model and a method thereof, characterized in that: The types of paper original vouchers in step S1 include receipts, invoices, approval forms, bank receipts, contract agreements, meeting minutes and work records.

3. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The step of constructing a method for generating accounting vouchers using a multimodal big model also includes: integrating the multimodal big model with natural language processing and image processing technology, performing semantic analysis on text data, extracting key information including amount, transaction type, and reason, and identifying key entities in the text including company name and account information.

4. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The step of constructing a multimodal large model to generate accounting vouchers also includes: performing multimodal data fusion and correlating text data, image data and numerical data to ensure data consistency and accuracy of accounting vouchers.

5. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The step of constructing a multimodal large model to generate accounting vouchers also includes: verifying the uniqueness and authenticity of the invoice.

6. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The construction of the multimodal large model trained in step S3 specifically includes the following steps: S31. Data collection: Collect financial data including different types of original vouchers, including receipts, invoices, approval forms, bank receipts, contract agreements, meeting minutes and work records; S32. Data cleaning and processing: Clean and process the collected financial data to ensure that the data input into the multimodal large model is accurate and consistent, including deleting duplicate values, processing missing values, processing outliers, and consistent processing; S33, model selection: select a suitable training model as the basis according to the number of parameters and inference speed; S34, Model training: Design training tasks, including joint representation learning of images and texts, so that the model can understand and process data of multiple modalities. Based on the training tasks, fine-tune the model for the specific task of automatic financial accounting, design fine-tuning tasks, train the model using labeled financial data, select appropriate loss functions and optimization algorithms, adjust model parameters, and improve the accuracy of the model. S35. Model evaluation, optimization and deployment: Use evaluation indicators to evaluate the performance of the model on the validation set. Adjust and optimize the model based on the evaluation results. For the result data generated by the model that does not meet the expectations, use it as historical training data to participate in the model training. Deploy the trained model to actual applications to ensure that it can receive actual financial voucher input and automatically generate accounting vouchers.

7. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The step of constructing and using a multimodal large model to generate accounting vouchers also includes: construction of vector data, including feature extraction and vectorization, and feature extraction includes data conversion, feature selection, and feature calculation.

8. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The step of constructing a multimodal large model to generate accounting vouchers also includes: processing of RAG (retrieval enhanced generation) data, including vector storage and management and performance optimization, and performance optimization includes dimensionality reduction, parameter tuning, verification of vector quality, and testing performance.

9. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The step S4 submits the generated accounting voucher to the financial personnel for review, including: the financial personnel checks the consistency of the generated accounting voucher with the original voucher to ensure the preparation and compliance of the accounting voucher; if there is an error, the financial personnel modifies and corrects the accounting voucher according to the corresponding original voucher; if there is no error, the confirmed financial accounting voucher is issued.

10. The intelligent financial management system and method based on multimodal large model according to claim 1 is characterized in that: The step S5 generates financial statements from the financial accounting vouchers in the financial management background, including: registering detailed ledgers and general ledgers according to the accounting vouchers; preparing financial statements according to the general ledger and detailed ledgers, and the financial statements include: balance sheet, income and income distribution statement.

Images