A method and system for automatically generating enterprise financial statements

Abstract

The present invention provides a method and system for automatically generating corporate financial statements, which relates to the technical field of data processing. The method includes: obtaining corporate financial data and financial reporting items in corporate financial statements; preliminarily screening corporate financial data to remove non-financial data; extracting associated sentences based on belief propagation for the preliminarily screened corporate financial data to retain multiple initial financial sentences with financial quadruplets; mapping bill predicates to financial reporting items, and merging initial financial sentences according to the mapping results to obtain primary financial data with the same bill predicate; verifying the consistency of primary financial data and corporate financial data; merging primary financial data according to bill predicates when the verification is passed, and outputting secondary financial data; filling the secondary financial data into the corresponding financial reporting items in the corporate financial statements, and outputting the obtained target corporate financial statements. The efficiency and accuracy of generating corporate financial statements are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for automatically generating enterprise financial statements. Background Art

[0002] Corporate financial statements are formal records that detail a company's financial status and operating results over a specific period of time. These statements, including balance sheets, income statements, cash flow statements, etc., provide key economic information for corporate managers, investors, creditors and other stakeholders to assess the company's financial health, profitability, capital structure and cash liquidity. Financial statements are not only a reflection of a company's financial transparency to the public, but also the basis for internal decision-making and strategic planning.

[0003] As the size of enterprises expands and transaction volume increases, manual preparation of financial statements becomes increasingly time-consuming, labor-intensive, and error-prone. Automated generation of financial statements can significantly improve data processing speed and accuracy, reduce labor costs and error rates. Automated tools can update financial information in real time, support quick decision-making, and enhance financial transparency by integrating electronic data collection, processing, and report generation. This automation is of great significance for ensuring compliance with financial reporting rules and standards, improving efficiency, and optimizing financial resource management.

[0004] However, existing methods directly perform global processing in the process of generating corporate financial statements without considering the redundancy of corporate financial data, resulting in an extremely complex analysis process, high generation difficulty and affected accuracy. Moreover, existing corporate financial statements often use pre-trained models to perform semantic processing of financial data, which is heavily dependent on the quality of training set data. The accuracy of the generated corporate financial statements is highly random and the accuracy is insufficient. They cannot meet the high accuracy requirements of the corporate financial statement generation process, are insufficiently practical, and their accuracy needs to be improved. Summary of the invention

[0005] In order to solve the technical problems that in the process of generating corporate financial statements, global processing is directly performed without considering the redundancy of corporate financial data, resulting in extremely complex analysis process, high generation difficulty and affected accuracy, and the existing corporate financial statements often use pre-trained models to perform semantic processing of financial data, which is heavily dependent on the quality of training set data. The generated corporate financial statements have high randomness in accuracy and insufficient accuracy, cannot meet the high accuracy requirements of the corporate financial statement generation process, are insufficient in practicality, and the accuracy needs to be improved, the present invention provides a method and system for automatically generating corporate financial statements.

[0006] The technical solution provided by the embodiment of the present invention is as follows:

[0007] First aspect

[0008] An embodiment of the present invention provides a method for automatically generating corporate financial statements, comprising:

[0009] S1: Obtain the financial data of the enterprise and the financial reporting items in the enterprise financial statements;

[0010] S2: Preliminary screening of the company’s financial data to remove non-financial data;

[0011] S3: extracting related sentences based on belief propagation from the initially screened enterprise financial data, and retaining multiple initial financial sentences with financial quadruple, where the financial quadruple includes billing time adverbial, billing subject, billing predicate and billing object;

[0012] S4: Map the bill predicate to the financial report item, and merge the initial financial sentences according to the mapping result to obtain the first-level financial data with the same bill predicate;

[0013] S5: Verify the consistency between primary financial data and enterprise financial data;

[0014] S6: If the verification is passed, the primary financial data is merged according to the bill predicate, and the secondary financial data having the bill predicate, the bill object and the bill time adverbial is output; otherwise, return to step S2;

[0015] S7: Fill the secondary financial data into the corresponding financial reporting items in the enterprise financial statements, and output the target enterprise financial statements.

[0016] Second aspect

[0017] An embodiment of the present invention provides a system for automatically generating corporate financial statements, including:

[0018] processor;

[0019] A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for automatically generating enterprise financial statements as described in the first aspect is implemented.

[0020] The third aspect

[0021] An embodiment of the present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for automatically generating corporate financial statements as described in the first aspect is implemented.

[0022] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0023] In the present invention, the acquired corporate financial data is first screened to remove irrelevant data, effectively reducing the complexity of subsequent sentence dependency parsing. Then, belief propagation-based associated sentence extraction is performed based on the financial reporting items required by the corporate financial data, and data elements related to specific financial events are accurately identified and associated in an efficient probabilistic reasoning manner, retaining important financial quadruplets. Then, synonym mapping of predicate verbs is performed, and the initial financial sentences are merged to further reduce the processing complexity of scattered data. Before the final merger of the financial data, consistency verification with the initial corporate financial data is performed to ensure that the merged data is supported by the original data and the data is traceable, ensuring the accuracy and interpretability of the generated corporate financial statements, avoiding the problem of insufficient interpretability caused by the black box effect of traditional models in the process of generating corporate financial statements, meeting the requirements of high accuracy and traceability of financial data, and being more practical. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 A schematic diagram of a process for automatically generating enterprise financial statements provided by an embodiment of the present invention;

[0026] Figure 2 A schematic diagram of the structure of a system for automatically generating enterprise financial statements provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0028] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0029] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0030] In the embodiments of the present invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are consistent.

[0031] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0032] Reference Manual Attached Figure 1 , showing a flow chart of a method for automatically generating enterprise financial statements provided by an embodiment of the present invention.

[0033] The embodiment of the present invention provides a method for automatically generating enterprise financial statements, which can be implemented by an automatic generation device for enterprise financial statements, and the automatic generation device for enterprise financial statements can be a terminal or a server. The processing flow of the method for automatically generating enterprise financial statements can include the following steps:

[0034] S1: Obtain the financial data of the enterprise and the financial reporting items in the enterprise's financial statements.

[0035] Among them, corporate financial data refers to the digital records of the economic activities of the enterprise generated during its daily operations. These data usually include but are not limited to income and expenditure records, assets and liabilities, investment activities, tax information, etc. Specifically, in terms of data types, it can involve cash flow details, various sales and purchase invoices, bank transaction records, salary payment records, etc. These data are the basis for understanding the economic situation of the enterprise and the original input for preparing financial statements. Corporate financial statements are a formal record that summarizes the financial activities and results of the enterprise over a certain period of time. Financial reporting items refer to the specific content elements that constitute the financial statements. They are the various items in the financial statements, such as assets, liabilities, income, expenses, etc. Each reporting item corresponds to a certain value, which is obtained by summarizing, classifying and calculating the financial data of the enterprise. For example, in the balance sheet, financial reporting items include cash, accounts receivable, inventory, fixed assets, short-term debt, long-term debt, etc. In the income statement, reporting items include sales revenue, costs, taxes, net profit, etc. These data are the data basis for the automatic generation process of corporate financial statements.

[0036] S2: Conduct preliminary screening of corporate financial data and remove non-financial data.

[0037] It should be noted that the initial screening of corporate financial data to remove non-financial data ensures the quality and relevance of the data. By excluding irrelevant information, the efficiency and accuracy of subsequent processing steps are improved. Secondly, this screening process can significantly reduce the complexity and computational burden of data processing, making the entire financial statement generation process more efficient. In addition, the initial screening helps reduce potential information overload, allowing the financial team to focus more on analyzing and reporting truly important financial indicators and trends, which not only speeds up the preparation of financial statements, but also improves the availability of reports and the quality of decision support.

[0038] In one possible implementation, the enterprise financial data is preliminarily screened using a Kuramoto model based on a coupled phase oscillator to remove non-financial data.

[0039] Among them, the Kuramoto model is a mathematical model that describes the synchronization phenomenon in a network of oscillators (or periodic systems). In this model, each oscillator has a phase, and these oscillators influence each other through a coupling mechanism and tend to synchronize their phases. This model is particularly suitable for studying complex dynamic interactions between a large number of elements. Using the Kuramoto model based on coupled phase oscillators for the initial screening of corporate financial data can effectively identify and separate data elements related to core financial activities. This method can automatically filter out non-financial data by analyzing the synchronization between data elements, thereby retaining the most relevant financial information. It not only improves the accuracy of data processing, but also greatly enhances the degree of automation of data processing, reduces the need for manual screening, and improves efficiency. In addition, this model can handle large-scale data sets and adapt to data of different types and sources. It is an effective tool for processing complex and large financial data sets.

[0040] In a possible implementation, S2 specifically includes:

[0041] S201: Convert each entire sentence in the enterprise financial data into a numerical feature vector.

[0042] Specifically, corporate financial data can be converted into numerical feature vectors through bag-of-words models, TF-IDF, word embedding, or pre-trained language models such as BERT.

[0043] S202: each data feature vector is used as an oscillator for self-classification until the synchronization value of each data feature vector is greater than a preset synchronization value, thereby obtaining a plurality of data feature vector sets:

[0044]

[0045] ωi = PCA(X i )

[0046]

[0047] r i ≥r0

[0048] Among them, PCA(X i ) represents the extraction of the i-th data feature vector X i The principal component of i represents the natural frequency of the ith data eigenvector, that is, the ith data eigenvector X i The principal component after dimensionality reduction, θ i and θ j denote the phase of the i-th oscillator and the phase of the j-th oscillator respectively, N denotes the total number of oscillators, K denotes the coupling strength, sin denotes the sine function, r i represents the local synchronization value of the i-th oscillator, N i represents the number of oscillators coupled to the i-th oscillator, cos represents the cosine function, and r0 represents the preset synchronization value.

[0049] It should be noted that those skilled in the art can set the size of the preset synchronization value according to actual needs, and the present invention is not limited here.

[0050] S203: Input each data feature vector set obtained by self-classification into a pre-trained classification model, and output target financial data, wherein the pre-trained classification model includes a pre-trained support vector machine and a pre-trained random forest.

[0051] S204: Based on the target financial data, remove non-financial data from the enterprise financial data.

[0052] Specifically, first, the text data is converted into numerical feature vectors, so that unstructured text data can be quantified and analyzed in the computational model, which enhances the operability of the data and the depth of the analysis. Then, the oscillator model is used to self-classify the feature vectors, and the relevance of the data is judged by the synchronization value, which effectively identifies the data closely related to financial activities and improves the accuracy of data screening. In addition, the use of pre-trained classification models such as support vector machines and random forests for further precise classification ensures the accuracy and efficiency of the screening process. Ultimately, this method not only optimizes the data processing process and reduces the burden of manual screening, but also improves the automation level of data processing, providing a high-quality data foundation for subsequent financial analysis and report production.

[0053] S3: Extract related sentences based on belief propagation for the initially screened corporate financial data, and retain multiple initial financial sentences with financial quadruplets.

[0054] Among them, the financial quadruple includes bill time adverbial, bill subject, bill predicate and bill object.

[0055] Among them, belief propagation is an algorithm for probabilistic reasoning in a graphical model. It is mainly used to calculate the marginal probability distribution of each node in the graph and transmit "messages" in the network structure to estimate these probabilities. This process involves the nodes (representing variables) and edges (representing the probabilistic dependencies between variables) of the graph. Applying the belief propagation algorithm allows the system to effectively extract associated sentences containing financial quadruplets by accurately modeling and calculating the complex probabilistic dependencies between elements in the financial data. This method improves the accuracy and efficiency of data extraction, especially when dealing with large and complex corporate financial data. It can ensure that the extracted information is highly relevant and accurate, providing a solid data foundation for subsequent report generation.

[0056] Specifically, in the financial quadruple, the bill time adverbial indicates the data time, the bill subject indicates the data subject, the bill predicate indicates the data behavior, such as "paid", "received", and the bill object indicates the data object, such as a specific amount or related accounts.

[0057] It should be noted that the extraction of related sentences containing financial quadruplets based on the belief propagation method greatly enhances the structuring of data and the extractability of information. The financial quadruplets include billing time adverbials, billing subjects, billing predicates, and billing objects, which together form a complete description of financial events. This method uses the powerful reasoning ability of the belief propagation algorithm to effectively extract key financial information from a large amount of unstructured or semi-structured data to ensure the integrity and accuracy of this information. Such processing not only improves the accuracy of data analysis, but also provides a reliable foundation for subsequent automated processing and report generation, thereby supporting complex decision-making processes and ensuring the compliance and transparency of financial reports.

[0058] In a possible implementation, S3 specifically includes:

[0059] S301: Select an original financial sentence from enterprise financial data.

[0060] S302: Take each element in the financial quadruple in the original financial sentence as a node, and the relationship between the elements as an edge, to construct an association graph.

[0061] S303: Establish a factor function describing the degree of association between nodes based on the node part of speech, node relative position and node dependency direction:

[0062] F(x i ,x j )=exp(θ1·δ pos-pair (xi ,x j )+θ2·|ij|+θ3·sign(ji))

[0063] Among them, F(x i ,x j ) represents node x i With node x j The factor function value between exp represents the exponential function, θ1, θ2 and θ3 represent the weight parameters whose sum is 1, δ pos-pair (x i ,x j ) means that if the i-th node x i is a noun, the jth node x j The indicator function returns 1 if it is a verb, otherwise it returns 0. |ij| represents the absolute value of the position index difference between the i-th node and the j-th node. sign(ji) indicates that if the i-th node x i At the jth node x j The sign function returns 1 if the value is in front of the , otherwise it returns -1.

[0064] S304: Determine the joint probability of each factor in the association graph under a specific arrangement state:

[0065]

[0066] Among them, p(A) represents the joint probability of the factor set m when the arrangement state is A, Represents the node x in the factor set m i With node x j The factor value when the permutation state is A, Z represents the normalization function that ensures that the sum of the joint probabilities is 1, Ω represents the permutation state set, and Π represents the cumulative multiplication operation.

[0067] S305: Message transmission is performed on each node in the association graph. The transmission process is specifically as follows:

[0068]

[0069] in, The message sent by node V to factor function F in the k+1 iteration represents the belief of node V on state v based on the influence of other adjacent factors. κ represents the normalization constant. The message sent from the factor function G to the node V in the kth iteration represents the support degree of the factor function G for the state v of the node V based on the information of other nodes related to V. represents the message sent from the factor function F to the node V in the k+1 iterations, which represents the support degree of the factor function F for the state v of the node V according to the information of other nodes related to V. N(V)\F represents the set of neighboring nodes of the node V without considering the factor function F. N(F)\V represents the set of neighboring nodes associated with the factor function F without including the node V. Ω:Ω[V]=v represents the permutation state of the node V in the set of all possible permutation states Ω. ∑ represents the summation. F A represents the factor function value in the arrangement state A, represents the message A[U] sent from node U to the factor function F in the kth iteration when node U is in the permutation state A.

[0070] Among them, belief is also called probability value. Through these message transmissions, the algorithm can gradually approach the marginal probability distribution of each variable, thereby achieving effective inference.

[0071] S306: Update the information of each node and factor function after the message is transmitted:

[0072]

[0073] in, It means that after the k+1th iteration, node V takes the marginal probability of node v, that is, belief, represents the message sent by the factor function G to the node V in state v during the kth iteration for all associated factor functions of the node V, represents the belief of the factor function in the permutation state A after the k+1th iteration, Represents the message sent by A[U] to the factor function F for node U which is one of all the adjacent nodes belonging to the factor function F in the arrangement state A.

[0074] It should be noted that the purpose of belief updating is to use local message passing to approximate the global optimal probability distribution. Each node and factor function achieves this by collecting information from its neighboring nodes and updating its own beliefs. In this way, the entire association graph can efficiently infer the marginal probability distribution of each variable to support decision-making. This method is particularly useful when dealing with large-scale complex networks because it avoids directly calculating the global joint probability distribution, thereby greatly reducing the computational complexity.

[0075] S307: According to the updated belief, extract the node with the highest belief value from the association graph as the financial quadruple of the original financial sentence, wherein the financial quadruple is the initial financial sentence corresponding to the original financial sentence.

[0076] It should be noted that the termination condition of belief update can be terminated by setting the maximum number of iterations, or based on whether the belief change is stable within the expected error, that is: Among them, max means taking the maximum value, It represents the belief of node i in state v at the tth iteration, and it represents the expected error, which is used to judge whether the change is small enough.

[0077] S308: Update the original financial sentence and return to step S302 until the financial quadruple in the enterprise financial data is extracted.

[0078] It should be noted that by constructing an association graph and applying factor functions to describe the complex relationships between nodes, an efficient and accurate way is provided to process and analyze corporate financial data. This method allows the system to understand and infer the dependencies between nodes at a global level, thereby accurately identifying and extracting sentences containing key financial information in financial texts. Through iterative message passing and belief updating, the algorithm can adaptively adjust the state estimate of each node, gradually approaching the real data model, and significantly improving the accuracy and reliability of data processing. This method is particularly suitable for processing large-scale and complex financial data sets, and can effectively extract the most valuable information from massive data, supporting companies to make more accurate financial decisions and strategy adjustments. Through this technology, companies can significantly improve the efficiency and quality of financial data analysis, and enhance financial transparency, compliance and accuracy.

[0079] S4: Map the bill predicate to the financial reporting item, and merge the initial financial sentences according to the mapping result to obtain the first-level financial data with the same bill predicate.

[0080] Among them, the primary financial data has related bill predicates, such as how much is spent, how much is consumed, how much is earned, etc. The method of mapping bill predicates to financial reporting items and merging data optimizes the organization and presentation of financial information. Through this mapping, the same or similar financial activities (such as payments, income, etc.) can be systematically classified, thereby improving the consistency and readability of data processing. In addition, merging financial sentences with the same predicate not only simplifies the data structure, but also enhances the accuracy and comprehensibility of the report, allowing enterprises to analyze financial status faster and make information-driven decisions accordingly. This method improves the efficiency of report generation while ensuring the integrity and transparency of the data.

[0081] In a possible implementation manner, S4 is specifically:

[0082] The bill predicate is mapped to the financial reporting item through the mapping function. The specific mapping function is:

[0083]

[0084] C n→o ={0,1}

[0085]

[0086] Among them, max means taking the maximum value, C n and C o They represent the semantics of the nth bill predicate and the semantics of the oth financial report item, respectively. n→o represents the decision variable of whether to map the nth bill predicate to the oth financial reporting item, δ represents the regularization parameter, ||C||1 represents the L1 norm of the decision variable matrix C, and m represents the mth bill predicate.

[0087] It should be noted that C n→o ={0,1} indicates that C n→o It is a binary variable that can only take 0 or 1, indicating whether C n Mapping to C o , if C n =C o , then the decision variable takes 1, otherwise it takes 0. The constraint ensures that each billing predicate n can be mapped to at most one financial reporting predicate o, preventing multiple mappings. It is used to maintain the consistency and transitivity of the mapping. If n and m are both mapped to o, then m and cannot be mapped to n, avoiding mapping and circular references between bill predicates and ensuring the consistency of the mapping logic. The introduction of the δ||C||1 regularization term promotes the sparsity of the mapping scheme, ensures the effectiveness of the mapping, and completes the process of mapping the bill predicate to the financial reporting item with the least possible mapping times under the condition that the conditions are met, thereby improving the data processing efficiency and the accuracy of the generated financial statements.

[0088] Specifically, accurate mapping ensures data consistency and standardization, reduces the complexity of data processing and potential error rates. The mapping process is highly optimized through decision variables and regularization techniques, making the mapping results not only accurate but also sparse, avoiding unnecessary duplication and redundant mapping, and improving processing efficiency. In addition, the implementation of this method can ensure the accurate correspondence and integrity of data, prevent data confusion in classification and subsequent processing, and thus improve the quality and credibility of the final financial report. This mapping strategy makes the generation of financial statements more systematic and automated, greatly improving the financial management capabilities of enterprises and the real-time update speed of reports.

[0089] S5: Verify the consistency between primary financial data and corporate financial data.

[0090] It should be noted that in the process of consistency verification, since the financial object of the secondary financial data has changed, the consistency of the target enterprise financial statements and the enterprise financial data is verified in step S5 by verifying the consistency of the primary financial data and the enterprise financial data. The consistency verification requires that the primary financial data needs and must be supported by the original text of the enterprise financial data to ensure the accuracy and traceability of the generated enterprise financial statements. By comparing and verifying data consistency, this step helps confirm the accuracy of data processing and conversion, avoiding data errors from being transmitted to the final financial statements. This verification process enhances the credibility of the statements and ensures the authenticity and integrity of the enterprise financial statements.

[0091] In a possible implementation, S5 specifically includes:

[0092] S501: vectorizing the primary financial data and the corresponding original financial sentences to obtain vectorized primary financial data and vectorized original financial sentences.

[0093] S502: Calculate the similarity between the vectorized primary financial data and the vectorized original financial sentence through cosine similarity to verify the consistency between the primary financial data and the enterprise financial data:

[0094]

[0095] Among them, S(B1, B0) represents the cosine similarity between the vectorized first-level financial data B1 and the vectorized original financial sentence B0, B1·B0 represents the dot product of the vectorized first-level financial data and the vectorized original financial sentence, and ||B1|| and ||B0|| represent the modulus of the vectorized first-level financial data and the vectorized original financial sentence, respectively.

[0096] S503: When the cosine similarity is greater than a preset cosine similarity, it is determined that the primary financial data and the enterprise financial data have passed the consistency verification; otherwise, it is determined that the primary financial data and the enterprise financial data have not passed the consistency verification.

[0097] In practical applications, by quantizing the primary financial data and the original financial sentences and calculating their cosine similarity, an accurate and scientific method is provided to verify the consistency of financial data. This method intuitively displays the degree of matching between data through numerical similarity scores, thereby ensuring that only verified consistent data is used for further financial report production. The use of cosine similarity reduces the interference of human judgment and improves the objectivity and accuracy of the verification process. In addition, this method can effectively screen out records that are inconsistent with the original data, ensuring the accuracy and reliability of the report, and is a key step in financial data management and control.

[0098] It should be noted that those skilled in the art can set the size of the preset cosine similarity according to actual needs, and the present invention is not limited here.

[0099] S6: If the verification is passed, the primary financial data is merged according to the bill predicate, and the secondary financial data with the bill predicate, the bill object and the bill time adverbial is output; otherwise, return to step S2.

[0100] Among them, the secondary financial data includes expenditure and consumption data within a specific period, such as how much was spent and consumed in July, etc. After the consistency verification is passed, the primary financial data is merged according to the bill predicate, which not only reduces data duplication, but also enhances the efficiency and accuracy of data analysis, ensuring that only verified data is further processed, thereby ensuring the accuracy and reliability of financial statements. In this way, enterprises can more effectively monitor and manage their financial activities, provide more accurate financial insights, and support decision making.

[0101] In a possible implementation, S6 specifically includes:

[0102] S601: Remove the bill subject in the primary financial data.

[0103] S602: Calculate the bill object value under the bill time constraint under each bill predicate:

[0104]

[0105] Among them, S c (T) represents the bill object value of the bill predicate c within the preset time length T, It represents the amount of the i-th bill with bill predicate c within the preset time period T, and I represents the total number of bills.

[0106] It should be noted that those skilled in the art can set the preset duration according to actual needs, and the present invention is not limited thereto.

[0107] S603: Concatenate the bill predicate, the bill object value and the bill time adverbial to obtain secondary financial data.

[0108] It should be noted that by removing the bill subject, this step focuses on the core numerical information of financial activities, namely the bill predicate and object, thereby simplifying the data structure and reducing redundant information. Calculating the total value of the object related to the bill predicate within a specific time period allows for aggregate analysis of financial flows within a time period, providing companies with a clear overview of financial flows. Finally, these data are spliced ​​together to form structured secondary financial data, which effectively supports accurate and dynamic financial decision-making and improves the practicality and readability of financial statements. This approach ensures the systematic and consistent processing of data, helping companies to better monitor and manage their financial status.

[0109] S7: Fill the secondary financial data into the corresponding financial reporting items in the enterprise financial statements, and output the target enterprise financial statements.

[0110] In a possible implementation manner, after S7, the method further includes:

[0111] The accuracy of collecting the target enterprise's financial statements is determined. If the accuracy is lower than a preset accuracy, the process returns to step S2.

[0112] It should be noted that this approach significantly improves the quality and reliability of financial statements because it ensures that financial statements undergo multiple rounds of verification and optimization before release, reducing financial risks caused by data errors or inconsistencies. In addition, real-time accuracy assessment allows for timely identification and resolution of potential data processing problems, ensuring that the company's financial information is transparent and accurate, supporting efficient financial decision-making and compliance management. This self-correcting process greatly enhances the integrity and reliability of the financial reporting system.

[0113] Specifically, this solution first ensures accurate data processing and capture of key information by efficiently screening and extracting financial data, including identifying and retaining financial quadruplets, thereby improving the accuracy of the report and reducing the complexity of subsequent dependency analysis. The use of belief propagation to extract related sentences enhances the logical correlation between data, making financial reports more detailed and credible. In addition, consistency verification ensures the traceability of data and the authenticity of reports, which is extremely important for meeting financial requirements and supporting decision making. The automated process reduces manual errors and increases processing speed, allowing companies to quickly respond to market and internal management needs, and enhancing the company's financial transparency and internal control efficiency. Overall, this approach provides companies with an efficient, reliable and compliant financial reporting tool, greatly improving the professionalism and systematicness of corporate financial management.

[0114] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0115] In the present invention, the acquired corporate financial data is first screened to remove irrelevant data, effectively reducing the complexity of subsequent sentence dependency parsing. Then, belief propagation-based associated sentence extraction is performed based on the financial reporting items required by the corporate financial data, and data elements related to specific financial events are accurately identified and associated in an efficient probabilistic reasoning manner, retaining important financial quadruplets. Then, synonym mapping of predicate verbs is performed, and the initial financial sentences are merged to further reduce the processing complexity of scattered data. Before the final merger of the financial data, consistency verification with the initial corporate financial data is performed to ensure that the merged data is supported by the original data and the data is traceable, ensuring the accuracy and interpretability of the generated corporate financial statements, avoiding the problem of insufficient interpretability caused by the black box effect of traditional models in the process of generating corporate financial statements, meeting the requirements of high accuracy and traceability of financial data, and being more practical.

[0116] Reference Manual Attached Figure 2 , showing a schematic diagram of the structure of a system for automatically generating enterprise financial statements provided by the present invention.

[0117] The present invention further provides a system 20 for automatically generating enterprise financial statements, which is applied to the above-mentioned method for automatically generating enterprise financial statements, and comprises:

[0118] Processor 201.

[0119] The memory 202 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 201, the method for automatically generating enterprise financial statements as in the method embodiment is implemented.

[0120] The enterprise financial statement automatic generation system 20 provided by the present invention can execute the above-mentioned enterprise financial statement automatic generation method and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate on it.

[0121] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0122] In the present invention, the acquired corporate financial data is first screened to remove irrelevant data, effectively reducing the complexity of subsequent sentence dependency parsing. Then, belief propagation-based associated sentence extraction is performed based on the financial reporting items required by the corporate financial data, and data elements related to specific financial events are accurately identified and associated in an efficient probabilistic reasoning manner, retaining important financial quadruplets. Then, synonym mapping of predicate verbs is performed, and the initial financial sentences are merged to further reduce the processing complexity of scattered data. Before the final merger of the financial data, consistency verification with the initial corporate financial data is performed to ensure that the merged data is supported by the original data and the data is traceable, ensuring the accuracy and interpretability of the generated corporate financial statements, avoiding the problem of insufficient interpretability caused by the black box effect of traditional models in the process of generating corporate financial statements, meeting the requirements of high accuracy and traceability of financial data, and being more practical.

[0123] It should be understood that the processor in the embodiment of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0124] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0125] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0126] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0127] In the present invention, "at least one" means one or more, and "more" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be represented by: a, b, c, ab, ac, bc, or abc, where a, b, c can be single or plural.

[0128] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0129] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0130] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0131] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0132] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0133] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0134] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0135] An embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for automatically generating enterprise financial statements as described in the method embodiment is implemented.

[0136] A computer-readable storage medium provided by the present invention can implement the steps and effects of the method for automatically generating corporate financial statements in the above method embodiment. To avoid repetition, the present invention will not go into details.

[0137] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0138] In the present invention, the acquired corporate financial data is first screened to remove irrelevant data, effectively reducing the complexity of subsequent sentence dependency parsing. Then, belief propagation-based associated sentence extraction is performed based on the financial reporting items required by the corporate financial data, and data elements related to specific financial events are accurately identified and associated in an efficient probabilistic reasoning manner, retaining important financial quadruplets. Then, synonym mapping of predicate verbs is performed, and the initial financial sentences are merged to further reduce the processing complexity of scattered data. Before the final merger of the financial data, consistency verification with the initial corporate financial data is performed to ensure that the merged data is supported by the original data and the data is traceable, ensuring the accuracy and interpretability of the generated corporate financial statements, avoiding the problem of insufficient interpretability caused by the black box effect of traditional models in the process of generating corporate financial statements, meeting the requirements of high accuracy and traceability of financial data, and being more practical.

[0139] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0140] There are a few points to note:

[0141] (1) The drawings of the embodiments of the present invention only relate to the structures related to the embodiments of the present invention, and other structures may refer to the general design.

[0142] (2) For the sake of clarity, in the drawings used to describe the embodiments of the present invention, the thickness of the layers or regions is exaggerated or reduced, that is, these drawings are not drawn according to the actual scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element may be "directly" "on" or "under" the other element or there may be intermediate elements.

[0143] (3) In the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other to obtain new embodiments.

[0144] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for automatically generating enterprise financial statements, characterized in that: include: S1: Obtaining financial data of an enterprise and financial reporting items in the financial statements of the enterprise; S2: Preliminary screening of the enterprise's financial data to remove non-financial data; S3: extracting related sentences based on belief propagation from the initially screened enterprise financial data, and retaining a plurality of initial financial sentences with financial quadruple groups, wherein the financial quadruple group includes a billing time adverbial, a billing subject, a billing predicate, and a billing object; S4: Mapping the bill predicate to the financial report item, and merging the initial financial sentences according to the mapping result to obtain primary financial data with the same bill predicate; S5: Verify the consistency between the primary financial data and the enterprise financial data; S6: If the verification is passed, the primary financial data is merged according to the bill predicate, and the secondary financial data having the bill predicate, the bill object and the bill time adverbial is output; otherwise, return to step S2; S7: Fill the secondary financial data into the corresponding financial reporting items in the enterprise financial statements, and output the obtained target enterprise financial statements.

2. The method for automatically generating enterprise financial statements according to claim 1, characterized in that: The enterprise financial data is preliminarily screened by a Kuramoto model based on a coupled phase oscillator to remove the non-financial data.

3. The method for automatically generating enterprise financial statements according to claim 2, characterized in that: The S2 specifically includes: S201: Convert each entire sentence in the enterprise financial data into a numerical feature vector; S202: each data feature vector is used as an oscillator for self-classification until the synchronization value of each data feature vector is greater than a preset synchronization value, thereby obtaining a plurality of data feature vector sets: oh i =PCA(X i ) r i ≥r0 Among them, PCA(X i ) represents the extraction of the i-th data feature vector X i The principal component of i represents the natural frequency of the ith data eigenvector, that is, the ith data eigenvector X i The principal component after dimensionality reduction, θ i and θ j denote the phase of the i-th oscillator and the phase of the j-th oscillator respectively, N denotes the total number of oscillators, K denotes the coupling strength, sin denotes the sine function, r i represents the local synchronization value of the i-th oscillator, N i represents the number of oscillators coupled to the i-th oscillator, cos represents a cosine function, and r0 represents the preset synchronization value; S203: Inputting each data feature vector set obtained by self-classification into a pre-trained classification model, and outputting target financial data, wherein the pre-trained classification model includes a pre-trained support vector machine and a pre-trained random forest; S204: Based on the target financial data, remove the non-financial data from the enterprise financial data.

4. The method for automatically generating enterprise financial statements according to claim 1, characterized in that: The S3 specifically includes: S301: Select an original financial sentence from the enterprise financial data; S302: Taking each element in the financial quadruple in the original financial sentence as a node and the relationship between the elements as an edge, a relationship graph is constructed; S303: Establish a factor function describing the degree of association between nodes based on the node part of speech, node relative position and node dependency direction: F(x i ,x j )=exp(θ1·δ pos-pair (x i ,x j )+θ2·|i-j|+θ3·sign(j-i)) Among them, F(x i ,x j ) represents node x i With node x j The factor function value between exp represents the exponential function, θ1, θ2 and θ3 represent the weight parameters whose sum is 1, δ pos-pair (x i ,x j ) means that if the i-th node x i is a noun, the jth node x j The indicator function returns 1 if it is a verb, otherwise it returns 0. |ij| represents the absolute value of the position index difference between the i-th node and the j-th node. sign(ji) indicates that if the i-th node x i At the jth node x j The sign function returns 1 if it is in front of , otherwise it returns -1; S304: Determine the joint probability of each factor in the association graph in a specific arrangement state: Among them, p(A) represents the joint probability of the factor set m when the arrangement state is A, Represents the node x in the factor set m i With node x j The factor value when the permutation state is A, Z represents a normalization function that ensures that the sum of the joint probabilities is 1, Ω represents a permutation state set, and Π represents a cumulative multiplication operation; S305: Message transmission is performed on each node in the association graph, and the transmission process is specifically as follows: in, The message sent by node V to factor function F in the k+1 iteration represents the belief of node V on state v based on the influence of other adjacent factors. κ represents the normalization constant. The message sent from the factor function G to the node V in the kth iteration represents the support degree of the factor function G for the state v of the node V based on the information of other nodes related to V. represents the message sent from the factor function F to the node V in the k+1 iterations, which represents the support degree of the factor function F for the state v of the node V according to the information of other nodes related to V. N(V)\F represents the set of neighboring nodes of the node V without considering the factor function F. N(F)\V represents the set of neighboring nodes associated with the factor function F without including the node V. Ω:Ω[V]=v represents the permutation state of the node V in the set of all possible permutation states Ω. Σ represents the summation. F A represents the factor function value in the arrangement state A, represents the message A[U] sent from node U to factor function F in the kth iteration when node U is in the permutation state A; S306: Update the information of each node and factor function after the message is transmitted: in, It means that after the k+1th iteration, node V takes the marginal probability of node v, that is, belief, represents the message sent by the factor function G to the node V in state v during the kth iteration for all associated factor functions of the node V, represents the belief of the factor function in the permutation state A after the k+1th iteration, represents the message sent by A[U] to the factor function F for the node U that is one of all the adjacent nodes of the factor function F in the arrangement state A; S307: extracting a node with the highest belief value from the association graph according to the updated belief as a financial quadruple of the original financial sentence, wherein the financial quadruple is an initial financial sentence corresponding to the original financial sentence; S308: Update the original financial sentence and return to step S302 until the financial quadruple in the enterprise financial data is completely extracted.

5. The method for automatically generating enterprise financial statements according to claim 1, characterized in that: The S4 is specifically: The bill predicate is mapped to the financial report item through a mapping function, and the mapping function is specifically: C n→o ＝{0,1} Among them, max means taking the maximum value, C n and C o They represent the semantics of the nth bill predicate and the semantics of the oth financial report item, respectively. n→o represents the decision variable of whether to map the nth bill predicate to the oth financial reporting item, δ represents the regularization parameter, ||C||1 represents the L1 norm of the decision variable matrix C, and m represents the mth bill predicate.

6. The method for automatically generating enterprise financial statements according to claim 4, characterized in that: The S5 specifically includes: S501: vectorizing the primary financial data and the corresponding original financial sentences to obtain vectorized primary financial data and vectorized original financial sentences; S502: Calculate the similarity between the vectorized primary financial data and the vectorized original financial sentence by using cosine similarity to verify the consistency between the primary financial data and the enterprise financial data: Wherein, S(B1, B0) represents the cosine similarity between the vectorized primary financial data B1 and the vectorized original financial sentence B0, B1·B0 represents the dot product of the vectorized primary financial data and the vectorized original financial sentence, ||B1|| and ||B0|| represent the modulus of the vectorized primary financial data and the vectorized original financial sentence, respectively; S503: When the cosine similarity is greater than a preset cosine similarity, it is determined that the primary financial data and the enterprise financial data have passed the consistency verification; otherwise, it is determined that the primary financial data and the enterprise financial data have not passed the consistency verification.

7. The method for automatically generating enterprise financial statements according to claim 1, characterized in that: The S6 specifically includes: S601: Remove the bill subject in the primary financial data; S602: Calculate the bill object value under the bill time constraint under each bill predicate: Among them, S c (T) represents the bill object value of the bill predicate c within the preset time length T, represents the amount of the i-th bill with bill predicate c within the preset time T, and I represents the total number of bills; S603: Concatenate the bill predicate, the bill object value and the bill time adverbial to obtain the secondary financial data.

8. The method for automatically generating enterprise financial statements according to claim 1, characterized in that: After S7, the method further includes: The accuracy of the target enterprise's financial statements is collected, and when the accuracy is lower than a preset accuracy, the process returns to step S2.

9. An automatic generation system for enterprise financial statements, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the method for automatically generating enterprise financial statements as claimed in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for automatically generating corporate financial statements as described in any one of claims 1 to 8 is implemented.

Images