An electronic accounting file grouping method, system and program product
By using AI big data models for semantic parsing of electronic accounting records and human-machine collaborative file assembly, the problems of low file assembly efficiency, inconsistent quality, and high maintenance costs in existing technologies have been solved, thus realizing an efficient and intelligent record management system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- YGSOFT INC
- Filing Date
- 2026-03-18
- Publication Date
- 2026-06-19
AI Technical Summary
In existing technologies, electronic accounting records are inefficient to assemble, of inconsistent quality, have rigid rule engines and high maintenance costs, lack intelligence and self-evolution capabilities, and result in a poor user experience.
Using a large AI model for semantic analysis, the key elements and relationships of electronic accounting records are identified. A human-machine collaborative mechanism is used to make file assembly decisions, and a feedback closed-loop optimization model is established to achieve automated initial screening and recommendation.
It has improved the efficiency and quality of document compilation, handled complex scenarios, enhanced user experience, reduced long-term maintenance costs, achieved breakthroughs in system intelligence and applicability, and promoted the digital accumulation of archives management.
Smart Images
Figure CN122240840A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of archives management technology, specifically relating to a method, system, and program product for assembling electronic accounting archives. Background Technology
[0002] With the increasing level of enterprise informatization, the quantity and complexity of electronic accounting records have also increased dramatically. Traditional electronic accounting record assembly mainly relies on two technical approaches: one is entirely manual operation, where archivists manually classify, sort, and assemble records based on their experience and understanding of business relationships. This method is inefficient, error-prone, and difficult to standardize; the other is rule-based automated assembly systems, which mechanically match records using predefined "IF-THEN" rules (e.g., merging records with consecutive voucher numbers and the same business type within the same accounting period). However, these two existing methods have the following significant drawbacks: 1. The efficiency bottleneck of manual mode is prominent: Faced with massive, multi-source, and heterogeneous electronic accounting archives, manual processing is time-consuming and labor-intensive, becoming the main bottleneck of archive management. Moreover, the quality of file compilation is highly dependent on personal experience and has poor consistency.
[0003] 2. Rigid and costly rule engine: To cover complex business scenarios, a large and complex rule base needs to be configured, which is difficult to implement and maintain. Rules are prone to conflict and cannot handle long-tail scenarios with undefined rules or files with ambiguous content. System performance decreases as the number of rules increases.
[0004] 3. The system interaction is passive and lacks intelligence: Existing automated systems are essentially passive "filters". When files do not meet any rules, the system can only issue a simple error message, unable to proactively provide constructive document assembly suggestions or alternative solutions, failing to solve the user's actual problems, resulting in a poor user experience.
[0005] 4. The system lacks self-evolution capabilities: Whether it's manual experience or static rules, knowledge updates rely on offline, periodic manual maintenance. The system cannot learn automatically from daily use, nor can it continuously optimize itself in response to business changes and user feedback, causing its capabilities to gradually lag behind. Summary of the Invention
[0006] The purpose of this invention is to provide an electronic accounting file compilation method, system, and program product to solve the above-mentioned problems existing in the prior art.
[0007] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, a method for compiling electronic accounting records is provided, including: Collect electronic accounting records transmitted from various business systems to form an electronic accounting record collection; The pre-trained AI model is invoked to perform semantic parsing on each electronic accounting record in the electronic accounting record collection, thereby obtaining the set of key elements for each electronic accounting record; The relationships between electronic accounting records are determined based on the set of key elements of each electronic accounting record. Each electronic accounting file with a corresponding relationship is used as an accounting voucher, and other accounting vouchers with the same relationship are searched from each business system, as well as the original vouchers associated with each accounting voucher. Based on the AI big data model, the accounting vouchers and their associated original vouchers are used to make a decision on the compilation of the accounting archives. The accounting vouchers and their associated original vouchers that meet the compilation decision are arranged in a set logical order to obtain the recommended accounting archive compilation. Send the recommended accounting file compilation to the administrator and receive feedback information from the administrator; When the feedback information contains fine-tuning instructions, the corresponding accounting vouchers and / or original vouchers in the recommended accounting file set are screened, merged and / or split according to the fine-tuning instructions to obtain the final accounting file set, and the final accounting file set is archived.
[0008] In one possible design, the method further includes: An optimized feedback dataset is created by combining recommended accounting file compilation, feedback information, and the final accounting file compilation. The optimized feedback dataset is input into the large AI model, enabling the large AI model to be trained and optimized using the optimized feedback dataset.
[0009] In one possible design, the method further includes: When the feedback information includes a confirmation instruction, the recommended accounting file set will be used as the final accounting file set, and the final accounting file set will be archived.
[0010] In one possible design, the electronic accounting record set includes orders, invoices, contracts, reports, payment requests, payment vouchers, and / or bank receipts.
[0011] In one possible design, the large AI model employs a deep neural network model based on the Transformer architecture.
[0012] In one possible design, the set of key elements includes business type, core project, accounting period, voucher number, related entity, retention period, confidentiality level and / or amount range, and the related relationships include belonging to the same business type, the same core project, the same accounting period and / or the same voucher number.
[0013] In one possible design, the satisfaction of the document grouping decision includes semantic similarity greater than a set similarity threshold and compliance with set accounting constraint rules, which include belonging to the same business type, the same core project, the same accounting period, and / or the same voucher number.
[0014] Secondly, an electronic accounting record assembly system is provided, including a record acquisition unit, a record parsing unit, an association determination unit, a voucher tracing unit, a scheme generation unit, a verification and transmission unit, a dynamic assembly unit, a record storage unit, and a feedback optimization unit, wherein: The document acquisition unit is used to connect with various business systems and collect the electronic accounting documents transmitted by each business system to form an electronic accounting document collection. The document parsing unit is used to call a pre-trained AI model to perform semantic parsing on each electronic accounting document in the electronic accounting document collection, and obtain the set of key elements of each electronic accounting document; The association determination unit is used to determine the association relationship between electronic accounting files based on the set of key elements of each electronic accounting file. The voucher traceability unit is used to use electronic accounting files with corresponding relationships as accounting vouchers, and to find other accounting vouchers with the same relationship from various business systems, as well as the original vouchers associated with each accounting voucher. The scheme generation unit is used to make a decision on the compilation of each accounting voucher and its associated original vouchers based on the AI big model. The accounting vouchers and their associated original vouchers that meet the compilation decision are arranged in a set logical order to obtain the recommended accounting file compilation. The verification and transmission unit is used to send the recommended accounting file set to the administrator and receive feedback information from the administrator. The dynamic file assembly unit is used to filter, merge, and / or split the corresponding accounting vouchers and / or original vouchers in the recommended accounting file assembly according to the fine-tuning instructions when the feedback information contains fine-tuning instructions, so as to obtain the final accounting file assembly; or when the feedback information contains confirmation instructions, the recommended accounting file assembly is used as the final accounting file assembly. The case file storage unit is used to archive the final accounting records. The feedback optimization unit is used to create an optimized feedback dataset by combining recommended accounting file sets, feedback information, and the final accounting file sets. The optimized feedback dataset is then input into the AI large model, which uses the optimized feedback dataset for training and optimization.
[0015] Thirdly, an electronic accounting record compilation system is provided, including: Memory, used to store instructions; A processor is configured to read instructions stored in the memory and execute any one of the electronic accounting record assembly methods described in the first aspect above, according to the instructions.
[0016] Fourthly, a computer-readable storage medium is provided, on which instructions are stored, which, when executed on a computer, cause the computer to perform any one of the electronic accounting record assembly methods described in the first aspect. Simultaneously, a computer program product is also provided, which, when executed on a computer, performs any one of the electronic accounting record assembly methods described in the first aspect.
[0017] Beneficial effects: This invention can automate the initial screening and recommendation of document assembly through a large AI model, significantly improving the efficiency and quality of document assembly. Simultaneously, the human-machine collaboration mechanism ensures the accuracy of the final document assembly results, avoiding the risks of pure automation. Through semantic understanding, it can handle massive amounts of undefined complex scenarios, solving the problems of rule base expansion and maintenance. Furthermore, the proactive recommendation method greatly improves the user experience, achieving a fundamental breakthrough in system intelligence and applicability. The unique feedback loop design allows the system to utilize high-value human decision-making data generated in daily work to achieve automatic model iteration and optimization, with long-term maintenance costs far lower than traditional systems that require regular manual updates to the rule base. The document assembly experience and judgment of archivists are continuously learned and solidified by the system through the feedback mechanism, realizing the digital inheritance of best practices in internal archival management, improving the overall business level of the team, and promoting the digital accumulation of archival management knowledge. Attached Figure Description
[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0019] Figure 1 This is a flowchart illustrating the method in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the system configuration in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the system configuration in Embodiment 3 of the present invention. Detailed Implementation
[0020] It should be noted that the descriptions of these embodiments are intended to aid in understanding the invention and do not constitute a limitation thereof. The specific structural and functional details disclosed herein are merely for describing exemplary embodiments of the invention. However, the invention may be embodied in many alternative forms and should not be construed as being limited to the embodiments described herein.
[0021] It should be understood that, unless otherwise explicitly specified and limited, the corresponding terms should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments according to the specific circumstances.
[0022] Specific details are provided in the following description to provide a complete understanding of the exemplary embodiments. However, those skilled in the art will understand that the exemplary embodiments can be implemented without these specific details. For example, the system may be shown in block diagrams to avoid obscuring the example with unnecessary details. In other embodiments, well-known processes, structures, and techniques may be shown without non-essential details to avoid obscuring the embodiments.
[0023] Example 1: This embodiment provides a method for assembling electronic accounting records, which can be applied to corresponding electronic record management servers, such as... Figure 1 As shown, the method includes the following steps: S1. Collect electronic accounting records transmitted from various business systems and form an electronic accounting record collection.
[0024] In practice, the server first connects to and collects electronic accounting records from various business systems, forming an electronic accounting record set. This set may include orders, invoices, contracts, reports, payment applications, payment vouchers, and / or bank receipts. For example, the server can automatically collect a batch of electronic accounting records to be archived from ERP (Enterprise Resource Planning) systems, procurement systems, engineering systems, and contract systems via a data interface service. These records include purchase orders, invoices from different suppliers, payment applications, payment vouchers, electronic bank receipts, and scanned copies of relevant technical agreements. After these records are combined into an electronic accounting record set, they are stored in an archive repository, triggering a file assembly task.
[0025] S2. Call the pre-trained AI model to perform semantic parsing on each electronic accounting file in the electronic accounting file collection to obtain the set of key elements of each electronic accounting file.
[0026] In practice, after the document assembly task is initiated, the server can invoke a pre-trained AI (Artificial Intelligence) model to perform deep semantic analysis (including understanding of relationships) on each electronic accounting document in the electronic accounting document collection. This results in a set of key elements for each electronic accounting document, which may include business type, core items, accounting period, voucher number, related entities, retention period, security classification, and / or amount range. The key point of this solution is that the AI model does not perform simple keyword matching, but rather deeply understands the business background, entity relationships, and contextual semantics of the documents, and automatically normalizes and extracts the key elements for document assembly (such as business type, core items, accounting period, voucher number, related entities, etc.), achieving a fundamental shift from "rule symbol matching" to "content semantic understanding."
[0027] The large AI model can be a deep neural network model based on the Transformer architecture (large language model or multimodal model), and the model as a whole consists of the following parts: 1. Embedding Layer. The embedding layer is used to convert the raw text data in accounting records into vector representations that the model can process. Specifically, it includes: token embedding, position embedding, and segment embedding. Through the above embedding methods, text information such as accounting voucher summaries, account names, amount descriptions, and attachment descriptions are mapped into vector sequences of a unified dimension.
[0028] 2. Multi-layer Transformer Encoder Layers. The number of layers can be configured to 12, 24, or 32 layers, depending on available computing resources. Each Transformer Encoder Layer includes a multi-head self-attention mechanism, a feedforward network, and residual connections with layer normalization (Residual+LayerNorm). The multi-head self-attention mechanism is used to model the contextual relationships between different fields in accounting document text; for example, the semantic association between voucher summary and accounting subject, the logical relationship between voucher date and its accounting period, and the correspondence between attachment description and voucher type. Through the multi-head attention mechanism, the model can learn text features in parallel from multiple semantic subspaces. The feedforward network typically consists of two fully connected layers, used to perform non-linear transformations on the attention output, improving the model's ability to express complex accounting semantic structures.
[0029] 3. Semantic Representation Layer. After multiple layers of Transformer encoding, the model outputs a high-dimensional semantic vector for each text unit. This vector comprehensively reflects the semantic information, contextual relationships, and business implications of the accounting document text. This semantic vector can be used for various subsequent tasks, such as classification, similarity calculation, and relationship inference.
[0030] 4. Task-specific Head: This layer performs task processing based on the high-dimensional semantic vector output from the semantic representation layer, including classification, similarity calculation, and relationship inference.
[0031] S3. Determine the relationships between electronic accounting records based on the set of key elements of each electronic accounting record.
[0032] In practice, the server can use an AI big data model to infer the relationship between electronic accounting files based on the key elements of each electronic accounting file, such as belonging to the same business type, the same core project, the same accounting period, and / or the same voucher number.
[0033] S4. Use the electronic accounting records with corresponding relationships as accounting vouchers, and search for other accounting vouchers with the same relationship from each business system, as well as the original vouchers associated with each accounting voucher.
[0034] In practice, the server can use electronic accounting files with corresponding relationships as accounting vouchers, and search for other accounting vouchers with the same relationship (i.e., other electronic accounting files with the same relationship existing in each business system) and the original vouchers associated with each accounting voucher. If combined with a few pre-set core compliance rules (such as accounting vouchers should be associated with corresponding receipts / payments, and receipts / payments should include corresponding upstream business processing applications, invoices, etc.), based on the items or accounting subjects extracted from the summary of the accounting vouchers to be archived, the server continues to search for corresponding upstream documents and other original voucher information related to the related items in each business system. Based on the actual business situation, the server dynamically traces and restores the original business, and automatically combines the accounting vouchers with the original vouchers in the business tracing relationship into a single volume.
[0035] For example, the server can receive these key elements through a recommender and inject a rule through a lightweight rule engine: "All documents under the same purchase order should be kept as related as possible." Based on its understanding of the business logic from a large model, the recommender dynamically generates a recommendation scheme, such as suggesting a file title like "XX Data Center Construction Project - Server Procurement Related Accounting Files," and including purchase orders, invoices, payment applications, payment vouchers, bank receipts, technical agreements, etc., within the file. Simultaneously, it identifies other accounting documents related to the project and the original documents associated with these accounting documents.
[0036] S5. Based on the AI big data model, the accounting vouchers and their associated original vouchers are used to make a decision on the compilation of the accounting archives. The accounting vouchers and their associated original vouchers that meet the compilation decision are arranged in a set logical order to obtain the recommended accounting archive compilation.
[0037] In practice, the server can use a large AI model to make decisions on the grouping of accounting vouchers and their associated original vouchers. The grouping decision logic is implemented using model reasoning and rule constraints: the large model outputs the semantic elements and relationships of the vouchers, and then combines them with preset accounting constraint rules, such as belonging to the same business type, the same core project, the same accounting period, the same voucher number, the same retention period, and / or the same security level, to automatically merge the files that meet the grouping decision. Example of grouping decision: The set of files to be grouped is vectorized, and the semantic similarity between any two files is calculated. When the similarity is greater than a preset threshold and meets the accounting constraint rules, they are determined to be the same file.
[0038] Then, the documents / files can be arranged according to a set logical order (such as by document issuance time, accounting period, or a set logical order of "payment voucher - payment application form - bank receipt - invoice - purchase order - technical agreement and other attachments") to merge the documents and obtain a recommended accounting file volume. For example, the documents can be arranged according to the issuance time of the documents: Accounting Voucher 1 - Accounting Voucher 1 associated original document + Accounting Voucher 2 - Accounting Voucher 2 associated original document + Accounting Voucher N - Accounting Voucher N associated original document to form one volume.
[0039] S6. Send the recommended accounting file compilation to the administrator and receive feedback information from the administrator.
[0040] In practice, the server can send the generated recommended accounting file assembly plan to the administrator, allowing the administrator to review and make decisions. The administrator can fine-tune the recommended accounting file assembly through intuitive operations such as dragging, merging, and splitting, or directly click to confirm the recommended accounting file assembly. The administrator generates corresponding feedback information based on the administrator's adjustment or confirmation operation and sends the feedback information to the server. For example, if the administrator sees the recommended accounting file assembly plan on the review interface and believes that the technical agreement belongs to the document category rather than accounting records and should not be included in this volume, they can drag the technical agreement file out of the recommended plan. The administrator's system interface will respond in real time, generating the corresponding adjustment instruction. Alternatively, if the administrator confirms that the recommended accounting file assembly plan is reasonable and clicks "Confirm Assembly," the administrator's system will immediately generate a confirmation instruction according to the instructions.
[0041] S7. When the feedback information contains a fine-tuning instruction, the corresponding accounting vouchers and / or original vouchers in the recommended accounting file set are screened, merged and / or split according to the fine-tuning instruction to obtain the final accounting file set. When the feedback information contains a confirmation instruction, the recommended accounting file set is taken as the final accounting file set.
[0042] In practice, after receiving feedback information, the server, if it determines that the feedback information contains a fine-tuning instruction, will filter, merge, and / or split the corresponding accounting vouchers and / or original vouchers in the recommended accounting file set according to the fine-tuning instruction to obtain the final accounting file set. If it determines that the feedback information contains a confirmation instruction, the recommended accounting file set will be used as the final accounting file set.
[0043] S8. Archive the final accounting records.
[0044] In practice, the server can automatically generate a case file catalog, cover, etc. according to the final accounting file compilation scheme, and store the final accounting file compilation package in a designated electronic warehouse for archiving.
[0045] S9. Use the recommended accounting file compilation, feedback information, and the final accounting file compilation to form an optimized feedback dataset, and input the optimized feedback dataset into the AI big model so that the AI big model can be trained and optimized using the optimized feedback dataset.
[0046] In practice, the server can utilize recommended accounting file compilation, feedback information, and the final accounting file compilation to form an optimized feedback dataset. This optimized feedback dataset reveals the differences between AI recommendations and human expert decisions and can be fed into the model optimization module. Through incremental learning, reinforcement learning, and other techniques, it learns and optimizes the methods for compiling and organizing accounting files for different business operations, continuously refining the parameters of the large AI model. This allows the server system to continuously learn the tacit knowledge of business experts, thereby achieving self-evolution in the accuracy of file compilation recommendations.
[0047] For example, during the above interaction, the server can collect key feedback data: the original set of documents and the features parsed by the AI, the AI-recommended solution including the technical agreement, and the solution not including the technical agreement as decided by the human team, the difference being that the technical agreement was excluded by the human team. The server system automatically labels this case as a training sample: the correct output corresponding to the feature vector (including labels such as "technical agreement" and "procurement") should be "not included in the accounting file." The model training module starts an incremental learning task, which uses all feedback data of the day, including this case, to fine-tune the AI model. The training objective is to minimize the difference between the model's recommendation and the human team's final decision. After training, the model parameters are updated, and it learns more deeply the business rule that "the weight of technical agreement documents in accounting file composition should be reduced." The updated AI model is then deployed to the corresponding production environment.
[0048] This method utilizes a large AI model to automate the initial screening and recommendation of document compilation, significantly improving both efficiency and quality. Simultaneously, a human-machine collaboration mechanism ensures the accuracy of the final compilation results, avoiding the risks of pure automation. Through semantic understanding, it can handle massive amounts of undefined and complex scenarios, solving the problems of rule base expansion and maintenance. Furthermore, the proactive recommendation method greatly improves user experience, achieving a fundamental breakthrough in system intelligence and applicability. A unique feedback loop design allows the system to leverage high-value human decision-making data generated in daily work to achieve automatic model iteration and optimization, resulting in long-term maintenance costs far lower than traditional systems requiring regular manual rule base updates. The document compilation experience and judgment of archivists are continuously learned and solidified by the system through the feedback mechanism, enabling the digital inheritance of best practices in internal archival management, improving the overall business level of the team, and promoting the digital accumulation of archival management knowledge.
[0049] Example 2: This embodiment provides an electronic accounting record compilation system, such as... Figure 2 As shown, it includes a file acquisition unit, a file parsing unit, an association determination unit, a voucher tracing unit, a scheme generation unit, a verification and transmission unit, a dynamic file assembly unit, a file storage unit, and a feedback optimization unit, wherein: The document acquisition unit is used to connect with various business systems and collect the electronic accounting documents transmitted by each business system to form an electronic accounting document collection. The document parsing unit is used to call a pre-trained AI model to perform semantic parsing on each electronic accounting document in the electronic accounting document collection, and obtain the set of key elements of each electronic accounting document; The association determination unit is used to determine the association relationship between electronic accounting files based on the set of key elements of each electronic accounting file. The voucher traceability unit is used to use electronic accounting files with corresponding relationships as accounting vouchers, and to find other accounting vouchers with the same relationship from various business systems, as well as the original vouchers associated with each accounting voucher. The scheme generation unit is used to make a decision on the compilation of each accounting voucher and its associated original vouchers based on the AI big model. The accounting vouchers and their associated original vouchers that meet the compilation decision are arranged in a set logical order to obtain the recommended accounting file compilation. The verification and transmission unit is used to send the recommended accounting file set to the administrator and receive feedback information from the administrator. The dynamic file assembly unit is used to filter, merge, and / or split the corresponding accounting vouchers and / or original vouchers in the recommended accounting file assembly according to the fine-tuning instructions when the feedback information contains fine-tuning instructions, so as to obtain the final accounting file assembly; or when the feedback information contains confirmation instructions, the recommended accounting file assembly is used as the final accounting file assembly. The case file storage unit is used to archive the final accounting records. The feedback optimization unit is used to create an optimized feedback dataset by combining recommended accounting file sets, feedback information, and the final accounting file sets. The optimized feedback dataset is then input into the AI large model, which uses the optimized feedback dataset for training and optimization.
[0050] Example 3: This embodiment provides an electronic accounting record compilation system, such as... Figure 3 As shown, at the hardware level, it includes: Data interface, used to establish data connection between the processor and various business systems; Memory, used to store instructions; The processor is used to read instructions stored in the memory and execute the electronic accounting file assembly method in Embodiment 1 according to the instructions.
[0051] Optionally, the system also includes an internal bus, through which the processor, memory, and data interface can be interconnected. This internal bus can be a PCIe (Peripheral Component Interconnect Eexpress) bus, which can be divided into an address bus, a data bus, a control bus, etc. The memory can include, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Flash Memory, First Input First Output (FIFO), and / or First In Last Out (FILO). The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0052] Example 4: This embodiment provides a computer-readable storage medium storing instructions. When these instructions are executed on a computer, the computer performs the electronic accounting file assembly method described in Embodiment 1. The computer-readable storage medium refers to a data storage medium, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or Memory Sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable systems.
[0053] This embodiment also provides a computer program product that, when run on a computer, executes the electronic accounting file assembly method in Embodiment 1. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system.
[0054] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for compiling electronic accounting records, characterized in that, include: Collect electronic accounting records transmitted from various business systems to form an electronic accounting record collection; The pre-trained AI model is invoked to perform semantic parsing on each electronic accounting record in the electronic accounting record collection, thereby obtaining the set of key elements for each electronic accounting record; The relationships between electronic accounting records are determined based on the set of key elements of each electronic accounting record. Each electronic accounting file with a corresponding relationship is used as an accounting voucher, and other accounting vouchers with the same relationship are searched from each business system, as well as the original vouchers associated with each accounting voucher. Based on the AI big data model, the accounting vouchers and their associated original vouchers are used to make a decision on the compilation of the accounting archives. The accounting vouchers and their associated original vouchers that meet the compilation decision are arranged in a set logical order to obtain the recommended accounting archive compilation. Send the recommended accounting file compilation to the administrator and receive feedback information from the administrator; When the feedback information contains fine-tuning instructions, the corresponding accounting vouchers and / or original vouchers in the recommended accounting file set are screened, merged and / or split according to the fine-tuning instructions to obtain the final accounting file set, and the final accounting file set is archived.
2. The method for compiling electronic accounting records according to claim 1, characterized in that, The method further includes: An optimized feedback dataset is created by combining recommended accounting file compilation, feedback information, and the final accounting file compilation. The optimized feedback dataset is input into the large AI model, enabling the large AI model to be trained and optimized using the optimized feedback dataset.
3. The method for compiling electronic accounting records according to claim 1, characterized in that, The method further includes: When the feedback information includes a confirmation instruction, the recommended accounting file set will be used as the final accounting file set, and the final accounting file set will be archived.
4. The method for compiling electronic accounting records according to claim 1, characterized in that, The electronic accounting archive collection includes orders, invoices, contracts, reports, payment applications, payment vouchers, and / or bank receipts.
5. The method for compiling electronic accounting records according to claim 1, characterized in that, The AI model uses a deep neural network model based on the Transformer architecture.
6. The method for compiling electronic accounting records according to claim 1, characterized in that, The set of key elements includes business type, core project, accounting period, voucher number, related entity, retention period, confidentiality level and / or amount range, and the related relationships include belonging to the same business type, the same core project, the same accounting period and / or the same voucher number.
7. The method for compiling electronic accounting records according to claim 1, characterized in that, The criteria for meeting the document grouping decision include semantic similarity greater than a set similarity threshold and compliance with set accounting constraint rules, which include belonging to the same business type, the same core project, the same accounting period, and / or the same voucher number.
8. An electronic accounting record compilation system, characterized in that, It includes a file acquisition unit, a file parsing unit, an association determination unit, a voucher tracing unit, a scheme generation unit, a verification and transmission unit, a dynamic file assembly unit, a file storage unit, and a feedback optimization unit, among which: The document acquisition unit is used to connect with various business systems and collect the electronic accounting documents transmitted by each business system to form an electronic accounting document collection. The document parsing unit is used to call a pre-trained AI model to perform semantic parsing on each electronic accounting document in the electronic accounting document collection, and obtain the set of key elements of each electronic accounting document; The association determination unit is used to determine the association relationship between electronic accounting files based on the set of key elements of each electronic accounting file. The voucher traceability unit is used to use electronic accounting files with corresponding relationships as accounting vouchers, and to find other accounting vouchers with the same relationship from various business systems, as well as the original vouchers associated with each accounting voucher. The scheme generation unit is used to make a decision on the compilation of each accounting voucher and its associated original vouchers based on the AI big model. The accounting vouchers and their associated original vouchers that meet the compilation decision are arranged in a set logical order to obtain the recommended accounting file compilation. The verification and transmission unit is used to send the recommended accounting file set to the administrator and receive feedback information from the administrator. The dynamic file assembly unit is used to filter, merge, and / or split the corresponding accounting vouchers and / or original vouchers in the recommended accounting file assembly according to the fine-tuning instructions when the feedback information contains fine-tuning instructions, so as to obtain the final accounting file assembly; or when the feedback information contains confirmation instructions, the recommended accounting file assembly is used as the final accounting file assembly. The case file storage unit is used to archive the final accounting records. The feedback optimization unit is used to create an optimized feedback dataset by combining recommended accounting file sets, feedback information, and the final accounting file sets. The optimized feedback dataset is then input into the AI large model, which uses the optimized feedback dataset for training and optimization.
9. An electronic accounting record compilation system, characterized in that, include: Memory, used to store instructions; A processor is configured to read instructions stored in the memory and execute the electronic accounting record assembly method according to any one of claims 1-7.
10. A computer program product, characterized in that, When the computer program product is run on a computer, it executes the electronic accounting file compilation method according to any one of claims 1-7.