Salary individual tax automatic accounting method and system based on RPA

Abstract

The invention discloses a salary individual tax automatic accounting method and system based on RPA, and relates to the technical field of data processing, and the method comprises the steps: reading the text data of a tax policy information source, executing condition features, calculating actions and limiting factor extraction, and building a semantic triple set; generating a tax rule template according to the semantic triple set, and loading the tax rule template to an RPA accounting unit in a parameterized calculation logic form; constructing a multi-dimensional income feature vector after collecting employee salary composition data; executing mode clustering of the multi-dimensional income feature vectors to generate a salary mode cluster; performing template matching on the salary mode cluster and a tax rule template, and establishing a matching result; and after the calculation logic of the RPA accounting unit is configured according to the matching result, automatic accounting processing is executed. The technical problems that in the prior art, salary individual tax accounting is tedious, many manual operations exist, and errors are prone to occurring are solved, and the technical effect of improving the efficiency and accuracy of salary individual tax accounting is achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to an automated payroll and individual income tax calculation method and system based on RPA. Background Technology

[0002] The process of calculating individual income tax on payroll typically involves several complex steps, including collecting employee payroll data, calculating tax, generating tax returns, and submitting the data to the tax system. This process often relies on manual operation, and the synchronization of employee payroll data with tax policies frequently requires manual input and verification, which is prone to errors. Furthermore, the data transfer and verification work between departments during tax calculation and filing further increases the complexity and risk of errors. Manual operation not only increases workload but also leads to low efficiency, especially in large-scale employee payroll management, where the probability of operational errors increases significantly, affecting the accuracy of individual income tax calculation. Summary of the Invention

[0003] This application provides an RPA-based automated payroll and individual income tax calculation method and system to address the technical problems of cumbersome payroll and individual income tax calculation, excessive manual operation, and easy error in existing technologies.

[0004] In view of the above problems, this application provides an RPA-based method and system for automated calculation of payroll and individual income tax.

[0005] The first aspect of this application provides an RPA-based automated payroll and individual income tax calculation method, the method comprising: The system reads text data from tax policy information sources, extracts conditional features, calculation actions, and limiting factors from the text data using a semantic parsing engine, and establishes a semantic triplet set. A tax rule template is generated based on the semantic triplet set, and the tax rule template is loaded into the RPA accounting unit in the form of parameterized calculation logic. After collecting employee salary composition data, a multidimensional income feature vector is constructed. An adaptive clustering model is used to perform pattern clustering of the multidimensional income feature vector, generating salary pattern clusters. The salary pattern clusters are then matched with the tax rule template to establish a matching result. Based on the matching result, the calculation logic of the RPA accounting unit is configured, and automated accounting processing is performed.

[0006] A second aspect of this application provides an RPA-based automated payroll and individual income tax calculation system, the system comprising: The module consists of four parts: a set creation module, a semantic triplet module, and a template generation module. The former reads text data from tax policy information sources and uses a semantic parsing engine to extract conditional features, computational actions, and limiting factors from the text data to create semantic triplet sets. The latter generates tax rule templates based on the semantic triplet sets and loads these templates into the RPA accounting unit using parameterized computational logic. The former constructs a vector model to build multidimensional income feature vectors after collecting employee salary composition data. The latter uses an adaptive clustering model to perform pattern clustering of the multidimensional income feature vectors, generating salary pattern clusters. The former matches the salary pattern clusters with the tax rule templates to establish matching results. The latter performs automated accounting processing after configuring the computational logic of the RPA accounting unit based on the matching results.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application reads text data from tax policy information sources, uses a semantic parsing engine to extract conditional features, calculation actions, and limiting factors from the text data, and establishes a semantic triplet set; it generates a tax rule template based on the semantic triplet set, and loads the tax rule template into the RPA accounting unit in the form of parameterized calculation logic; after collecting employee salary composition data, it constructs a multi-dimensional income feature vector; it uses an adaptive clustering model to perform pattern clustering of the multi-dimensional income feature vector to generate salary pattern clusters; it performs template matching between the salary pattern clusters and the tax rule template to establish matching results; after configuring the calculation logic of the RPA accounting unit according to the matching results, it performs automated accounting processing. This invention solves the technical problems of cumbersome salary and individual income tax calculation, excessive manual operation, and easy error in the prior art. By using an RPA robot to automatically perform salary and individual income tax calculation, declaration, and data processing, it achieves the technical effect of improving the efficiency and accuracy of salary and individual income tax calculation. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0009] Figure 1 A schematic diagram of the RPA-based automated payroll and individual income tax calculation method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of an RPA-based automated payroll and individual income tax calculation system provided in an embodiment of this application.

[0010] Figure labeling: Module 11 for set creation, Module 12 for template generation, Module 13 for vector construction, Module 14 for clustering, Module 15 for matching, and Module 16 for kernel calculation. Detailed Implementation

[0011] This application provides an RPA-based automated payroll and individual income tax calculation method and system, which addresses the technical problems of cumbersome payroll and individual income tax calculation, excessive manual operation, and easy errors in the existing technology. By utilizing RPA robots to automatically perform payroll and individual income tax calculation, declaration, and data processing, it achieves the technical effect of improving the efficiency and accuracy of payroll and individual income tax calculation.

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0013] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0014] Example 1, as Figure 1 As shown, this application provides an RPA-based automated payroll and individual income tax calculation method, the method comprising: Step S100: Read the text data of the tax policy information source, use the semantic parsing engine to extract the conditional features, calculation actions and limiting factors of the text data, and establish a semantic triple set.

[0015] In this embodiment of the application, the text data of the tax policy information source is first read. The text data includes various tax rules, calculation methods and related restrictions for individual income tax.

[0016] Next, the semantic parsing engine is used to extract conditional features, computed actions, and limiting factors from the text data. In this process, the format structure recognition layer in the semantic parsing engine first performs structured recognition on the text data, dividing it into clause-level, sentence-level, and phrase-level semantic layers. Then, the cross-sentence dependency parsing layer is used to identify cross-sentence dependencies between conditional trigger statements and action instruction statements, and to configure limiting semantic paths. Finally, semantic triples are generated based on the limiting semantic paths, achieving the extraction and structured representation of key features in the text data.

[0017] Furthermore, the method provided in the application embodiments, which utilizes a semantic parsing engine to extract conditional features, computational actions, and limiting factors from text data to establish a semantic triplet set, also includes: The semantic parsing engine's format structure recognition layer is used to perform format structure recognition of text data, dividing it into semantic levels including clause level, sentence level, and phrase level; the cross-sentence dependency parsing layer is used to identify cross-sentence dependencies of conditional trigger statements and action instruction statements based on the semantic levels, and configure restricted semantic paths; and a set of semantic triples is generated based on the restricted semantic paths.

[0018] In this embodiment, when the format structure recognition layer of the semantic parsing engine performs format structure recognition of text data, it uses Natural Language Processing (NLP) technology, employing syntactic analysis and dependency analysis to perform format structure recognition on the text data, dividing the text data according to a hierarchical structure and identifying different semantic levels, including clause level, sentence level, and phrase level. At the clause level, macro-level regulations or rules are identified; at the sentence level, specific operational steps or calculation rules are identified; and at the phrase level, specific conditions, values, and restriction information are extracted.

[0019] Next, a cross-sentence dependency parsing layer is used to identify cross-sentence dependencies between conditional trigger statements and action instruction statements based on semantic hierarchy. In this process, the cross-sentence dependency parsing layer employs pronoun resolution and semantic role labeling methods for relationship identification. Pronoun resolution identifies and associates pronouns or noun phrases pointing to the same entity in different sentences; semantic role labeling determines which parts of a sentence are conditions, which are actions, and their respective triggers and receivers. Through this process, conditional trigger statements and action instruction statements distributed at the sentence level and even the clause level that logically possess causal or conditional relationships are identified, and their cross-sentence dependencies are clarified, thus forming a constrained semantic path.

[0020] Finally, based on the configured restricted semantic paths, a set of semantic triples is generated using entity recognition and relation extraction methods. Entity recognition extracts key entities from the paths, such as income and tax amount, which serve as the subject and object of the triples. Next, relation extraction defines the logical or computational relationships between entities, such as "income > 5000" representing a condition and "tax amount = income × tax rate" representing a calculation formula; these relationships constitute the predicate of the triples. This process deconstructs and transforms the text data into a standard triple form, i.e., (subject, predicate, object), providing a clear rule structure for subsequent automated calculations. Finally, all triples are aggregated into semantic triples.

[0021] Step S200: Generate a tax rule template based on the semantic triple set, and load the tax rule template into the RPA accounting unit in the form of parameterized calculation logic.

[0022] In this embodiment of the application, the subject, predicate and object content of the semantic triple set are first sorted out, and the condition information, calculation information and restriction information contained in the semantic triple set are classified respectively to ensure that the key logic in the policy rules is accurately decomposed.

[0023] Subsequently, the categorized semantic triples are combined according to the logical sequence of tax policies to construct a tax rule template that includes conditional judgment paths, calculation formula structures, and constraint verification content. This allows the tax rule template to fully express the execution process of the original policy. Next, data items in the tax rule template, such as tax rates, tax bases, and deduction standards, are parameterized, transforming these variable contents into automatically populated parameters. This enables the tax rule template to be directly used in different accounting objects and different accounting periods.

[0024] Finally, the parameterized tax rule template is loaded into the RPA accounting unit, enabling the RPA accounting unit to perform automated accounting based on the condition judgments, calculation logic, and restrictions set in the tax rule template, thereby realizing the direct application of tax rules in the accounting process.

[0025] Step S300: After collecting employee salary composition data, construct a multidimensional income feature vector.

[0026] In this embodiment, employee salary composition data is first collected from a preset database. Then, the collected data is structured and organized, classifying basic salary, performance-based salary, allowances and subsidies, overtime pay, one-time income, social insurance contributions, housing provident fund contributions, and deductions according to predefined fields. Subsequently, data cleaning is performed on each field, including format validation, missing value handling, and outlier identification, to ensure the accuracy and consistency of the salary composition data.

[0027] After data cleaning, the categorized fields are standardized to convert data of different dimensions into comparable numerical forms, and categorical fields are uniformly coded. Then, according to the feature design requirements, multiple feature dimensions reflecting income structure are extracted from the standardized fields, such as the proportion of fixed income, fluctuations in variable income, special deductions, and one-time income indicators.

[0028] Finally, after numericalizing and normalizing the above feature dimensions, they are combined into a set of vector data in a predetermined order to form a multi-dimensional income feature vector representing the employee income structure.

[0029] Step S400: Perform pattern clustering of the multidimensional income feature vector using an adaptive clustering model to generate salary pattern clusters.

[0030] In this embodiment of the application, when performing pattern clustering of multidimensional income feature vectors using an adaptive clustering model, the similarity between multidimensional income feature vectors is measured by calculating the Euclidean distance between them using the K-means clustering algorithm in the adaptive clustering model. Vectors with similar features are automatically grouped into one class. Through this process, salary pattern clusters are obtained.

[0031] Furthermore, in the method provided in the application embodiments, after generating the salary pattern cluster, it further includes: The salary datasets from two consecutive accounting periods are statistically compared to calculate the distribution change index of the main salary features. The feature distribution difference is calculated using the distribution change index to generate the clustering drift. When the clustering drift exceeds a preset trigger threshold, a retraining instruction is triggered. The adaptive clustering model is updated and managed according to the retraining instruction.

[0032] In this embodiment, the method of calculating the mean difference and variance ratio is first used to statistically compare the salary datasets over two consecutive accounting periods. By calculating the magnitude of the change in the average value and the proportion of the change in the dispersion of each major salary feature over the two periods, a numerical set of quantitative feature distribution changes is obtained, i.e., the distribution change index.

[0033] Next, the Euclidean distance metric is used to process these distribution variation indicators. By calculating the straight-line distance from the origin to the current point in multidimensional space, all variations are combined into a specific numerical value, which is the feature distribution dissimilarity. Then, the feature distribution dissimilarity is processed using the min-max normalization method, linearly transforming the value to the range of zero to one, generating a clustering drift index with uniform dimensions. This index can intuitively reflect the degree of change in data distribution.

[0034] When the clustering drift metric exceeds a preset trigger threshold, a retraining instruction is generated. This instruction contains key information such as the current drift value and the time of occurrence. Finally, based on the received retraining instruction, an adaptive clustering model update management method is used. All parameters of the clustering model are retrained using the latest complete salary data to ensure that the model can adapt to the latest data distribution characteristics.

[0035] Step S500: Perform template matching between the salary pattern cluster and the tax rule template to establish a matching result.

[0036] In this embodiment of the application, a template matching is performed between the salary pattern cluster and the tax rule template. That is, the cosine similarity between the feature vector of the salary pattern cluster and the parameter vector of the tax rule template is calculated. Then, the calculated cosine similarity is compared with a preset threshold. When the preset threshold is exceeded, a matching result between the salary pattern cluster and the corresponding tax rule template is established, that is, a matching result is generated.

[0037] Step S600: After configuring the calculation logic of the RPA accounting unit according to the matching results, execute the automated accounting process.

[0038] In this embodiment, the calculation logic of the RPA accounting unit is configured based on the matching results through rule mapping and parameter binding. In this process, the tax rule template corresponding to the salary pattern cluster in the matching results is first parsed into executable calculation steps. For example, calculation actions in the template, such as tax payable = taxable income × tax rate - quick calculation deduction, are converted into calculation instructions recognizable by the RPA accounting unit. Simultaneously, parameter variables in the template, such as basic deduction and special additional deductions, are bound to the employee's actual salary data fields. After configuration, the RPA accounting unit performs automated accounting processing.

[0039] In the automated accounting process, computational behavior logs and link identifiers are first generated at the task nodes of the RPA accounting unit, constructing an accounting link graph covering the pre-tax, post-tax, and declaration stages. Then, node computation authentication is performed at the task nodes, establishing an explainable anomaly report. This report includes the policy template, computational parameters, and rule confidence score corresponding to the anomaly node. Finally, the explainable anomaly report is mapped and bound to the accounting link graph, and anomaly reporting is executed to achieve traceability and anomaly management in the accounting process.

[0040] Furthermore, the method provided in the application embodiments, in performing automated accounting processing, further includes: During the automated accounting process executed by the RPA accounting unit, calculation behavior logs and link identifiers are generated based on task nodes to construct an accounting link graph that includes pre-tax, post-tax, and declaration stages. Node calculation authentication is performed at the task nodes, and an interpretable anomaly report is established. The interpretable anomaly report includes the policy template, calculation parameters, and rule confidence score corresponding to the anomaly node. After the interpretable anomaly report is mapped and bound to the accounting link graph, anomaly reporting is executed.

[0041] In this embodiment, during the automated accounting process executed by the RPA accounting unit, a directed graph modeling method is first used to construct an accounting link graph. In this process, by defining the dependencies between task nodes, key steps such as income aggregation and tax-exempt item identification in the pre-tax calculation stage, tax calculation and post-tax salary calculation in the post-tax accounting stage, and report generation and data submission in the declaration stage are linked into a complete workflow. Each task node generates a calculation behavior log containing timestamps and input / output parameters during execution and is assigned a globally unique link identifier, ultimately forming the accounting link graph.

[0042] Next, in the node computation authentication stage, a rule verification engine is used to perform computation accuracy checks. During this process, a similarity calculation algorithm is used to generate a rule confidence score by comparing the actual computation results with the expected results from the tax rule template. Specifically, a similarity analysis is performed between the parameter set defined in the tax rule template and the parameter set actually used in the computation. By measuring the consistency of the two parameter sets in numerical distribution and logical relationships, a rule confidence score within the range of 0-100 is obtained. When the score is lower than the preset passing standard, an interpretable anomaly report is generated. This report fully records the tax rule template number corresponding to the anomaly node, the actual computational parameter values ​​used, and the calculated rule confidence score.

[0043] Finally, graph mapping technology is used to associate anomaly reports with the accounting link graph. This process uses a node identifier matching method to establish a correspondence between the anomaly description information in the explainable anomaly report and the specific task node in the accounting link graph, thus forming a precise location of the anomaly in the overall process. After the association is completed, an anomaly reporting operation is performed, that is, the result of binding the explainable anomaly report with the accounting link graph is reported.

[0044] Furthermore, in the method provided in the application embodiments, after binding the explainable anomaly report with the accounting link graph mapping and then performing anomaly reporting, it further includes: The causal backtracking identification of the explainable anomaly reports is performed to construct an accounting correction instruction set; the accounting correction instruction set is used to perform recalculation and re-verification repair management of the RPA accounting unit.

[0045] In this embodiment of the application, when identifying causal backtracking for explainable anomaly reports, a causal graph analysis method is used. The specific tax rule is located based on the policy template number recorded in the anomaly report. By analyzing the correlation between the abnormal fluctuations of the calculated parameter values ​​and the rule confidence score, a complete abnormal causal chain from the abnormal result to the root cause is constructed.

[0046] Then, based on the identified causes of the anomalies, conditional matching rules are used to match the anomaly types with predefined correction strategies, generating a set of accounting correction instructions that includes specific operations such as parameter calibration, rule updates, and data verification.

[0047] Subsequently, when using the accounting correction instruction set to execute the recalculation and re-verification repair management of the RPA accounting unit, the instruction set is converted into machine-executable code through an instruction compilation mechanism. This drives the RPA accounting unit to execute operations such as parameter re-input, calculation logic adjustment, and data re-collection in the instruction sequence, completing targeted recalculation processing. After the recalculation is completed, a result verification process is initiated, employing a dual verification mechanism to check the integrity of the recalculation results. This ensures that the new calculation results comply with tax rules and that the rule confidence score meets the passing standard, ultimately completing the closed-loop management of anomaly handling.

[0048] Furthermore, in the method provided in the application embodiments, before performing automated accounting processing, it further includes: Construct a dynamic task topology graph based on the dependencies between accounting tasks; optimize the graph structure by optimizing the execution order, parallel dependencies, and resource allocation of accounting tasks using the dynamic task topology graph; when node resource conflicts or dependency delays are detected, automatically reconstruct the concurrent scheduling scheme of the accounting link, and perform automated accounting processing based on the reconstruction results.

[0049] In this embodiment of the application, when constructing a dynamic task topology graph based on the dependency relationship of accounting tasks, a directed acyclic graph modeling method is adopted to abstract accounting tasks such as salary data collection, individual income tax calculation, and result verification into graph nodes. Edge connection relationships are established according to the data flow between tasks to form a dynamic task topology graph structure.

[0050] Next, a dynamic task topology graph is used to optimize the graph structure, including the execution order, parallel dependencies, and resource allocation of the accounting tasks. In this process, a topology sorting method is first applied to determine the executable order of tasks, ensuring that payroll data collection precedes individual income tax calculation, and individual income tax calculation precedes result verification. Then, dependency analysis identifies subsets of independent tasks that can be executed in parallel, such as individual income tax calculation tasks from different departments that can be processed concurrently. Finally, a resource scheduling method is used to allocate available computing resources to parallel tasks, allocating more CPU resources to computationally intensive tasks and more memory resources to data-intensive tasks, ensuring that each task can obtain the necessary resources without conflict during execution.

[0051] When node resource conflicts or task dependency delays are detected, an automatic refactoring mechanism is triggered. For resource conflicts, a priority scheduling method is used to reassess task importance, suspend non-urgent tasks, and ensure that critical computing tasks are executed first. For dependency delays, a path optimization method is used to adjust the execution sequence of subsequent tasks, and cached data is used to maintain process operation. During the refactoring process, the node dependencies in the topology graph are re-analyzed, the task execution order and resource allocation are adjusted, and a new concurrent scheduling scheme is generated, while maintaining the directed acyclic property of the dynamic task topology graph.

[0052] Finally, automated accounting processing is executed according to the reconstructed concurrent scheduling scheme. The RPA accounting unit executes tasks according to the optimized topology and resource allocation. The task scheduler parses the new scheduling scheme into specific instructions, controlling the RPA accounting unit to perform payroll data collection according to the updated task sequence, process individual income tax calculation according to the optimized parallel strategy, and verify the execution results according to the reallocated resources, ensuring the correct order of the accounting chain in the pre-tax, post-tax, and declaration stages, and ultimately completing the fully automated accounting processing.

[0053] Furthermore, the method provided in the application embodiments, in performing automated accounting processing, further includes: During the automated accounting process, full-process tracking is performed, and a full-process tracking log is established; the full-process tracking log is synchronized to the enterprise database, and record management is performed.

[0054] In this embodiment of the application, during the automated accounting process, log tracking technology is used to perform full-process tracking. By embedding log recording points at each task node such as salary data collection, individual income tax calculation, and result verification, key information such as task start time, execution status, input parameters, and output results are collected in real time. These scattered log records are integrated and linked together according to the execution time sequence to establish a full-process tracking log containing a complete operation trajectory.

[0055] Subsequently, the completed full-process tracking log is transmitted to the enterprise database through a data synchronization mechanism. New log entries are created in the designated data table of the enterprise database to fully save the full-process tracking log of this accounting task and complete the record management operation of the enterprise database.

[0056] In summary, the embodiments of this application have at least the following technical effects: This application reads text data from tax policy information sources, uses a semantic parsing engine to extract conditional features, calculation actions, and limiting factors from the text data, and establishes a semantic triplet set; it generates a tax rule template based on the semantic triplet set, and loads the tax rule template into the RPA accounting unit in the form of parameterized calculation logic; after collecting employee salary composition data, it constructs a multi-dimensional income feature vector; it uses an adaptive clustering model to perform pattern clustering of the multi-dimensional income feature vector to generate salary pattern clusters; it performs template matching between the salary pattern clusters and the tax rule template to establish matching results; after configuring the calculation logic of the RPA accounting unit according to the matching results, it performs automated accounting processing. This invention solves the technical problems of cumbersome salary and individual income tax calculation, excessive manual operation, and easy error in the prior art. By using an RPA robot to automatically perform salary and individual income tax calculation, declaration, and data processing, it achieves the technical effect of improving the efficiency and accuracy of salary and individual income tax calculation.

[0057] Example 2, based on the same inventive concept as the RPA-based automated payroll and individual income tax calculation method in the previous examples, such as... Figure 2 As shown, this application provides an RPA-based automated payroll and individual income tax calculation system. The system and method embodiments in this application are based on the same inventive concept. The system includes: The three-tuple set creation module 11 is used to read text data from tax policy information sources, extract conditional features, calculation actions, and limiting factors from the text data using a semantic parsing engine, and establish a semantic triplet set; the template generation module 12 is used to generate a tax rule template based on the semantic triplet set, and load the tax rule template into the RPA accounting unit in the form of parameterized calculation logic; the vector construction module 13 is used to construct a multi-dimensional income feature vector after collecting employee salary composition data; the clustering module 14 is used to perform pattern clustering of the multi-dimensional income feature vector using an adaptive clustering model to generate salary pattern clusters; the matching module 15 is used to perform template matching between the salary pattern clusters and the tax rule template to establish a matching result; and the accounting module 16 is used to configure the calculation logic of the RPA accounting unit according to the matching result and then perform automated accounting processing.

[0058] Furthermore, the system is also used to implement the following functions: During the automated accounting process executed by the RPA accounting unit, calculation behavior logs and link identifiers are generated based on task nodes to construct an accounting link graph that includes pre-tax, post-tax, and declaration stages. Node calculation authentication is performed at the task nodes, and an interpretable anomaly report is established. The interpretable anomaly report includes the policy template, calculation parameters, and rule confidence score corresponding to the anomaly node. After the interpretable anomaly report is mapped and bound to the accounting link graph, anomaly reporting is executed.

[0059] Furthermore, the system is also used to implement the following functions: The causal backtracking identification of the explainable anomaly reports is performed to construct an accounting correction instruction set; the accounting correction instruction set is used to perform recalculation and re-verification repair management of the RPA accounting unit.

[0060] Furthermore, the system is also used to implement the following functions: The semantic parsing engine's format structure recognition layer is used to perform format structure recognition of text data, dividing it into semantic levels including clause level, sentence level, and phrase level; the cross-sentence dependency parsing layer is used to identify cross-sentence dependencies of conditional trigger statements and action instruction statements based on the semantic levels, and configure restricted semantic paths; and a set of semantic triples is generated based on the restricted semantic paths.

[0061] Furthermore, the system is also used to implement the following functions: The salary datasets from two consecutive accounting periods are statistically compared to calculate the distribution change index of the main salary features. The feature distribution difference is calculated using the distribution change index to generate the clustering drift. When the clustering drift exceeds a preset trigger threshold, a retraining instruction is triggered. The adaptive clustering model is updated and managed according to the retraining instruction.

[0062] Furthermore, the system is also used to implement the following functions: Construct a dynamic task topology graph based on the dependencies between accounting tasks; optimize the graph structure by optimizing the execution order, parallel dependencies, and resource allocation of accounting tasks using the dynamic task topology graph; when node resource conflicts or dependency delays are detected, automatically reconstruct the concurrent scheduling scheme of the accounting link, and perform automated accounting processing based on the reconstruction results.

[0063] Furthermore, the system is also used to implement the following functions: During the automated accounting process, full-process tracking is performed, and a full-process tracking log is established; the full-process tracking log is synchronized to the enterprise database, and record management is performed.

[0064] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An RPA-based automated payroll and individual income tax calculation method, characterized in that, The method includes: Read text data from tax policy information sources, use a semantic parsing engine to extract conditional features, computational actions, and limiting factors from the text data, and establish a semantic triple set; A tax rule template is generated based on the semantic triple set, and the tax rule template is loaded into the RPA accounting unit in the form of parameterized calculation logic. After collecting employee salary composition data, a multidimensional income feature vector is constructed; The multidimensional income feature vector is subjected to pattern clustering using an adaptive clustering model to generate salary pattern clusters; The salary pattern cluster is matched with the tax rule template to establish a matching result; After configuring the calculation logic of the RPA accounting unit based on the matching results, the automated accounting process is executed.

2. The RPA-based automated payroll and individual income tax calculation method as described in claim 1, characterized in that, Perform automated accounting processing, including: During the automated accounting process performed by the RPA accounting unit, calculation behavior logs and link identifiers are generated based on task nodes to construct an accounting link graph that includes the pre-tax, post-tax, and declaration stages. The task node performs node computation authentication and establishes an interpretable anomaly report, which includes the policy template, computation parameters and rule confidence score corresponding to the anomaly node. After binding the explainable anomaly report with the accounting link graph mapping, the anomaly reporting is executed.

3. The RPA-based automated payroll and individual income tax calculation method as described in claim 2, characterized in that, After binding the explainable anomaly report with the accounting link graph mapping, the anomaly reporting is executed, including: The causal backtracking identification is performed on the explained anomaly reports to construct an accounting correction instruction set; The RPA accounting unit performs recalculation and re-verification repair management using the aforementioned accounting correction instruction set.

4. The RPA-based automated payroll and individual income tax calculation method as described in claim 1, characterized in that, Using a semantic parsing engine, conditional features, computational actions, and limiting factors are extracted from text data to construct semantic triple sets, including: The format structure recognition layer of the semantic parsing engine is used to perform format structure recognition of text data and divide it into semantic levels including clause level, sentence level and phrase level; The cross-sentence dependency parsing layer is used to identify cross-sentence dependencies between conditional trigger statements and action instruction statements based on the semantic hierarchy, and restricts semantic paths accordingly. Generate a set of semantic triples based on the restricted semantic path.

5. The RPA-based automated payroll and individual income tax calculation method as described in claim 1, characterized in that, After generating the salary pattern cluster, it includes: Statistical distribution comparison of salary datasets from two consecutive accounting periods is performed to calculate the distribution change index of key salary characteristics. The characteristic distribution difference degree is calculated using the aforementioned distribution change index to generate cluster drift degree; When the clustering drift is determined to exceed a preset trigger threshold, a retraining instruction is triggered; The adaptive clustering model is updated and managed according to the retraining instructions.

6. The RPA-based automated payroll and individual income tax calculation method as described in claim 1, characterized in that, Before performing automated accounting processing, the following steps are included: Construct a dynamic task topology graph based on the dependencies between accounting tasks; The dynamic task topology graph is used to optimize the graph structure by calculating the execution order, parallel dependencies, and resource allocation of the accounting tasks. When node resource conflicts or dependency delays are detected, the concurrent scheduling scheme of the accounting link is automatically reconstructed, and automated accounting processing is performed based on the reconstruction results.

7. The RPA-based automated payroll and individual income tax calculation method as described in claim 1, characterized in that, Perform automated accounting processing, including: During automated accounting processing, full-process tracking is performed, and a full-process tracking log is established. The entire process tracking logs are synchronized to the enterprise database for record management.

8. An RPA-based automated payroll and individual income tax calculation system, characterized in that, The system is used to execute the RPA-based automated payroll and individual income tax calculation method as described in any one of claims 1-7, and the system includes: The set building module is used to read text data from tax policy information sources, and use the semantic parsing engine to extract conditional features, calculation actions and limiting factors from the text data to build semantic triple sets. The template generation module is used to generate a tax rule template based on the semantic triple set, and load the tax rule template into the RPA accounting unit in the form of parameterized calculation logic. The vector construction module is used to construct multidimensional income feature vectors after collecting employee salary composition data; The clustering module is used to perform pattern clustering of the multidimensional income feature vector using an adaptive clustering model to generate salary pattern clusters; The matching module is used to perform template matching between the salary pattern cluster and the tax rule template to establish a matching result; The accounting module is used to configure the calculation logic of the RPA accounting unit based on the matching results and then execute automated accounting processing.

Images