Search

CN-122023049-A - Cross-platform financial data automatic accounting method and system

CN122023049ACN 122023049 ACN122023049 ACN 122023049ACN-122023049-A

Abstract

The invention relates to the technical field of enterprise financial automation accounting, in particular to a cross-platform financial data automatic accounting method and system, comprising the following steps: establishing connection with a plurality of heterogeneous financial systems, periodically acquiring an original flowing water containing transaction remark texts, and performing field extraction and format cleaning to generate standardized flowing water entries. The standardized entries are fed into a pre-trained text semantic classifier, and economic substantive categories of transaction behaviors are parsed based on remark text. According to the category, matching corresponding account entry rules from an accounting rule knowledge base, performing accounting decomposition on the flow entry, and generating a standardized accounting entry draft. And performing hook auditing relation verification on the entry draft, and checking the internal lending balance and the logic conflict with the historical posted transaction in subject continuity. The invention realizes intelligent understanding and automation and precise logic accounting of multi-source heterogeneous financial data.

Inventors

  • QIN YAZHOU

Assignees

  • 深圳赛盒科技有限公司

Dates

Publication Date
20260512
Application Date
20260331

Claims (9)

  1. 1. A method for automatically accounting cross-platform financial data, the method comprising: establishing communication connection with a plurality of external heterogeneous financial systems, and periodically acquiring original transaction running water from the heterogeneous financial systems, wherein the original transaction running water comprises a transaction time stamp, transaction party information, transaction amount and transaction remark text; Performing field extraction and format cleaning on the obtained original transaction flow, eliminating the difference of different heterogeneous financial systems in data formats, character codes and measurement units, and generating standardized flow entries; sending the standardized running water item to a pre-trained text semantic classifier, wherein the text semantic classifier analyzes economic essence categories of transaction behaviors based on the content of the transaction remark text; Matching corresponding account entering rules from a preset accounting rule knowledge base according to the economic essence category of the transaction behavior; performing accounting decomposition on the standardized running water item by using the matched accounting rules to generate a standardized accounting entry draft; executing hook audit relation verification on the standardized accounting entry draft, including: Extracting all debit accounting subjects and corresponding accounting amounts from the standardized accounting entry draft, and calculating the sum of the debit amounts; extracting all credit accounting subjects and corresponding accounting amounts from the standardized accounting entry draft, and calculating the sum of the credit amounts; Comparing the sum of the borrowing amount with the sum of the lending amount, and if the difference exceeds an allowable numerical tolerance range, marking that the standardized accounting entry has a lending imbalance error; Querying an end-of-term balance of an accounting subject during a previous accounting period from a historical accounting database for the accounting subject related to assets, liabilities, equity; Checking whether the variation direction and the numerical value of the accounting amount of the standardized accounting entry draft containing the accounting subject cause the situation that the balance of the subject does not accord with business logic, and marking the accounting entry draft as continuity logic conflict if the accounting entry draft exists; Generating a hook auditing relation verification report containing all error marks and conflict marks, and marking the standardized accounting entry draft passing verification as an entry to be checked; Checking a lending balance relationship among a plurality of standardized accounting documents resolved by the same original transaction flow, and checking whether a logic conflict exists between the lending balance relationship and a historical posted transaction in subject continuity.
  2. 2. The method for automatically accounting cross-platform financial data according to claim 1, wherein said performing field extraction and format cleansing on said raw transaction stream obtained comprises: Identifying a source system identifier of each original transaction flow, and loading a corresponding data analysis template from a preset format mapping template library according to the source system identifier; Applying the data analysis template to extract key value pairs of unstructured text or semi-structured data of the original transaction flow, and extracting the transaction time stamp, the transaction party information, the transaction amount and the original character string of the transaction remark text; uniformly converting the extracted transaction time stamp into a world coordination time format, uniformly converting the transaction amount into a specified home currency, and uniformly converting the transaction party information and the character code of the transaction remark text into a universal character set code; Carrying out named entity recognition processing on the transaction party information and the transaction remark text after the completion of format and code conversion, and recognizing and correcting misspellings, inconsistent abbreviations and multi-word meaning expressions existing in the transaction party information and the transaction remark text; and repackaging the data after finishing all the cleaning conversion steps according to a preset internal data model, and outputting the data as the standardized pipeline item.
  3. 3. The method of claim 2, wherein the standardized flow entry is fed into a pre-trained text semantic classifier that resolves economic parenchymal categories of transaction behavior based on the content of the transaction remark text, comprising: Constructing a semantic understanding model based on a deep neural network architecture, and training the semantic understanding model by using massive marked historical transaction remark texts and corresponding accounting subjects to form the pre-trained text semantic classifier; Extracting the transaction remark text from the standardized journal entry, and inputting the transaction remark text into the pre-trained text semantic classifier; the pre-trained text semantic classifier performs word segmentation, vectorization and context semantic coding on the transaction remark text and outputs a high-dimensional semantic feature vector; Mapping the high-dimensional semantic feature vector to a predefined economic parenchyma category probability space, and calculating a probability value of the economic parenchyma category belonging to each transaction behavior; And selecting the economic essence category of the transaction behavior with the highest probability value as the final classification result of the standardized running water item.
  4. 4. A method of automatically accounting cross-platform financial data according to claim 3, wherein matching corresponding accounting rules from a preset accounting rule knowledge base according to economic parenchymal categories of the transaction behavior comprises: the accounting rule knowledge base contains structured posting rule entries converted from accounting criteria terms; the accounting rule knowledge base is organized in a directed graph structure, nodes of the graph represent specific accounting subjects, edges of the graph represent accounting processing logic derived from accounting criterion clauses, and each edge is attached with an economic essence category label of the transaction behavior to which the edge is applicable; taking the economic essence category of the transaction behavior as a query index, and searching all edges with economic essence category labels in a directed graph structure of the accounting rule knowledge base; Further screening out entry rules matched with other key attributes in the standardized pipeline items from the retrieved edges, wherein the key attributes comprise transaction main body types, transaction occurrence countries or regions; Selecting an entry rule with highest priority from the screened multiple entry rules according to a preset rule priority policy as a matching result; binding the core data field of the standardized running water entry with the matched accounting rule to form an accounting decomposition task to be executed.
  5. 5. The method for automatically accounting cross-platform financial data according to claim 4, wherein said accounting of said standardized flow entry using said matched entry rules comprises: the standardized accounting entry draft comprises a debit accounting subject, a credit accounting subject, an accounting amount and an affiliated accounting period; Analyzing the logic structure of the posting rule, and identifying an amount calculation field, a subject judgment condition and a lending direction defined in the rule; Extracting the transaction amount and other relevant numerical value fields from the standardized journal entry, and calculating the accounting amount which should be counted into an accounting entry according to the amount calculation logic defined in the entry rule; determining at least two accounting subjects related to the transaction according to the standardized attribute of the flow entry and subject judgment conditions in the account entry rule; designating each determined accounting subject as a borrower accounting subject or a lender accounting subject according to the lending direction defined in the accounting rule; determining the belonging accounting period of the standardized accounting entry draft according to the transaction time stamp in the standardized flow entry and in combination with a preset accounting period division strategy; summarizing the debit accounting subjects, the credit accounting subjects, the accounting amount and the period of the accounting, and generating a complete standardized accounting entry draft.
  6. 6. The method for automatically accounting cross-platform financial data according to claim 5, further comprising the step of multi-dimensionally rechecking a standardized accounting entry draft marked as an entry to be checked: Performing risk dimension screening on the to-be-checked entry based on a preset checking strategy rule set, wherein the checking strategy rule set comprises a large transaction checking threshold value, a sensitive transaction opponent list and an abnormal transaction time mode; Triggering a business authenticity checking flow of the to-be-checked entry, and calling an electronic invoice and a contract scanning piece supporting file related to the to-be-checked entry from an original evidence image database; automatically comparing the key information in the to-be-checked entry with the called supporting file, wherein the key information comprises an amount, a trade opponent, a date and a trade summary; for the to-be-checked entry which fails to pass the automatic comparison, or the entry triggering the high risk rule in the risk dimension screening, pushing the to-be-checked entry to a manual rechecking queue, and attaching the related original transaction running water to be linked with a supporting file; And approving the formal account entry of the to-be-checked entry which passes through all automatic rechecks, writing the formal account entry into a general account database, and simultaneously updating the real-time balance of the related account.
  7. 7. The method for automatically accounting cross-platform financial data according to claim 6, wherein the triggering the business authenticity checking process for the entry to be checked comprises: Reversely tracing the source of the original transaction running water from the to-be-checked address book, and acquiring a global unique identifier of the original transaction running water for generating the to-be-checked address book; taking the global unique identifier as an index, performing full-text retrieval in the original voucher image database, and searching storage paths of all associated electronic invoices and contract scanning piece supporting files; Carrying out content identification on the searched supporting file through a file analysis service, and extracting structured data in the file, wherein the structured data comprises invoice codes, invoice numbers, invoicing dates, tax free amount, tax amount, purchaser and seller information; Establishing a mapping relation between structured data recorded from the to-be-checked file to the supporting file, and carrying out consistency check item by item according to preset auditing logic; And recording the result of each consistency check, and generating a structured business authenticity check report as a part of the check conclusion of the to-be-checked entry.
  8. 8. The method for automatically accounting cross-platform financial data according to claim 7, further comprising the step of periodically pre-generating trial balance and report: at the expiration time of each accounting period, extracting formal accounting records of all accounts entered in the current period from the general account database; classifying and summarizing all the formal accounting records according to the accounting subjects, and calculating the occurrence amount of the borrower, the occurrence amount of the lender and the end balance of each accounting subject in the period; generating a pre-adjustment trial balance table based on the end-of-period balances of all accounting purposes, checking whether the asset balance of the trial balance table is equal to the liability and owner equity balance; If the trial balance table is unbalanced, starting a balance tracing program, and performing tracing inspection on all accounting entries and balance calculation processes in the current period until a difference reason is found and an adjustment entry is generated; and after the trial balance is passed, extracting data from the general ledger database according to a preset report template, and automatically filling the main parts of the asset liability list, the profit list and the cash flow list to generate a financial report draft.
  9. 9. A cross-platform financial data automatic accounting system comprising a memory, a processor and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, performs the steps of a cross-platform financial data automatic accounting method as claimed in any one of claims 1 to 8.

Description

Cross-platform financial data automatic accounting method and system Technical Field The invention relates to the technical field of enterprise financial automation accounting, in particular to a cross-platform financial data automatic accounting method and system. Background When performing financial accounting, current enterprises often need to interface with multiple external heterogeneous financial or business systems, such as banks, paymate platforms, ERP systems, and the like. The raw transaction stream produced by these systems varies significantly in data format, field definition, character encoding, and monetary units. Handling these differences relies primarily on financial personnel to manually perform data cleansing, format conversion and standardization, or to write a large number of script rules at once. This approach is inefficient in handling multi-source, high frequency data, and is highly prone to errors due to rule omission or human negligence, and the data preprocessing stage becomes the bottleneck in the automated process. In the transaction essence judging link, the prior art scheme generally adopts a rule matching method based on fixed keywords or regular expressions to classify the transaction remark text. However, natural language expressions are of diversity and complexity, and transactions of the same economic nature may correspond to a vast number of different textual descriptions, and rule bases are difficult to exhaust and costly to maintain. This results in a large number of transactions that cannot be accurately classified, still require manual review of the decisions, and have limited automation. In terms of accounting logic verification, existing automated accounting systems typically perform only the most basic debit balance verification, lacking deep checking of subject use logic continuity and cross-term transaction relevance. How to accurately and automatically identify the economic essence of the transaction from unstructured texts and how to implement verification which exceeds that of single transaction and ensures the logic rigor of the whole accounting after automatic entry generation are key barriers for improving the intelligent level and reliability of the whole process of financial accounting. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a cross-platform financial data automatic accounting method and system. In order to achieve the purpose, the invention adopts the following technical scheme that the cross-platform financial data automatic accounting method comprises the following steps: establishing communication connection with a plurality of external heterogeneous financial systems, and periodically acquiring original transaction running water from the heterogeneous financial systems, wherein the original transaction running water comprises a transaction time stamp, transaction party information, transaction amount and transaction remark text; Performing field extraction and format cleaning on the obtained original transaction flow, eliminating the difference of different heterogeneous financial systems in data formats, character codes and measurement units, and generating standardized flow entries; sending the standardized running water item to a pre-trained text semantic classifier, wherein the text semantic classifier analyzes economic essence categories of transaction behaviors based on the content of the transaction remark text; Matching corresponding account entering rules from a preset accounting rule knowledge base according to the economic essence category of the transaction behavior; performing accounting decomposition on the standardized running water item by using the matched accounting rules to generate a standardized accounting entry draft; Executing hook auditing relation verification on the standardized accounting entry draft, checking lending balance relations among a plurality of standardized accounting entry drafts decomposed by the same original transaction flow, and checking whether logic conflict exists between the lending balance relations and historical credited transactions in subject continuity. As a further aspect of the present invention, the performing field extraction and format cleaning on the obtained raw transaction flowing water includes: Identifying a source system identifier of each original transaction flow, and loading a corresponding data analysis template from a preset format mapping template library according to the source system identifier; Applying the data analysis template to extract key value pairs of unstructured text or semi-structured data of the original transaction flow, and extracting the transaction time stamp, the transaction party information, the transaction amount and the original character string of the transaction remark text; uniformly converting the extracted transaction time stamp into a world coordination time format, uniformly converting the tran