Declaration Logic Validation Method and System Based on Tax Policy Knowledge Base

Abstract

This invention provides a method and system for verifying declaration logic based on a tax policy knowledge base. It calls upon a pre-set tax policy knowledge base to visualize policy items, generating a set of visualized policy elements. It then acquires the declaration data and scenario descriptions submitted by taxpayers, and through structured decomposition and analysis, forms a scenario-based declaration element set. The scenario-based declaration element set and the policy element visualization set are imported into an element association mechanism. Through scenario matching, suitable policies are identified, and attribute comparisons are performed to generate element mapping results. Deviation links are filtered to locate the root causes of differences and track their chain effects, obtaining declaration difference transmission information. Finally, the declaration difference transmission information is integrated with the scenario-based declaration element set, combined with policy element generation field adaptation suggestions and policy basis, to generate a declaration logic verification report according to the declaration order and push it to the declaration terminal. This invention effectively improves the accuracy and efficiency of tax declarations, reduces tax risks, and enhances the intelligence level of declaration verification.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and more specifically, to a method and system for verifying declaration logic based on a tax policy knowledge base. Background Technology

[0002] In the field of tax filing, the complexity of tax filing is increasing with the continuous updating and improvement of tax policies. Taxpayers need to accurately understand and comply with the requirements of various tax policies when filing. However, due to the large number of tax policy items, diverse forms of expression, and differences in applicable policies under different filing scenarios, taxpayers are prone to making logical errors in their filings, leading to inaccurate data and potentially triggering tax risks. Traditional tax filing verification methods mainly rely on manual review. Reviewers need to check each item of the filing data against the policy requirements based on their familiarity with the tax policies. This method is not only inefficient but also prone to omissions and false positives due to reviewers' misunderstandings or oversights. Furthermore, with the development of tax informatization, although some tax systems provide certain automatic verification functions, these functions are often limited to simple data format checks or basic logical judgments, failing to deeply understand the connotation of tax policies and making it difficult to accurately verify the complex logical relationship between filing data and policy requirements. Therefore, there is an urgent need for a tax filing logic verification method that can effectively solve the above problems. Summary of the Invention

[0003] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a declaration logic verification method based on a tax policy knowledge base, the method comprising: The system calls upon a pre-defined tax policy knowledge base, performs visualization processing on the policy entries within the base, extracts the elements related to the declaration in the policy description and the relationships between these elements, and generates a visualization set of policy elements. This visualization set of policy elements contains the corresponding information of policy entries and elements. Obtain the declaration data submitted by taxpayers and the associated declaration scenario descriptions, decompose the declaration data into basic field elements, analyze the correlation attributes between field elements to form a declaration element set, and bind the declaration element set with the declaration scenario descriptions to obtain a scenario-based declaration element set; The scenario-based application element set and the policy element representation set are imported into the element association mechanism. The appropriate policy is locked through scenario matching, the attribute comparison between field elements and policy elements is performed, and a mapping link is established to generate element mapping results. By filtering deviation links from the element mapping results, extracting the corresponding declaration field elements and policy elements, locating the root cause of the difference and tracking the chain effect of the difference on related fields, and recording the transmission path to obtain the declaration difference transmission information; Integrate the information on differences in the application process with the set of application elements for specific scenarios, combine policy elements to generate field adaptation suggestions and policy basis, organize the content according to the application order to form an application logic verification report, and push the report to the application terminal.

[0004] In another aspect, embodiments of the present invention also provide a declaration logic verification system based on a tax policy knowledge base, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0005] Based on the above, this embodiment of the invention constructs a tax policy knowledge base and visualizes the policy entries within the base to generate a visualized set of policy elements. This allows for the accurate extraction of elements related to the declaration and the relationships between these elements from policy statements. After obtaining the declaration data submitted by taxpayers and the associated declaration scenario descriptions, the declaration data is structurally decomposed and the correlation attributes between field elements are analyzed to form a scenario-based declaration element set. This ensures that the declaration data is closely integrated with specific scenarios and more accurately reflects the actual declaration situation. By importing the scenario-based declaration element set and the visualized set of policy elements into the element association mechanism, suitable policies can be quickly identified, and attribute comparisons between field elements and policy elements can be performed to establish mapping links and generate element mapping results, achieving accurate matching between declaration data and policy requirements. By filtering deviation links from the element mapping results, locating the root causes of differences, and tracking the chain effects of differences on related fields, problems existing in the declaration data can be discovered in depth. Finally, by integrating the information on differences in tax declarations with the set of scenario-based tax declaration elements, and combining policy elements to generate field adaptation suggestions and policy basis, a declaration logic verification report is generated and pushed to the declaration terminal. This provides taxpayers with clear and detailed declaration guidance, effectively improves the accuracy and efficiency of tax declarations, reduces tax risks, and enhances the level of intelligence in tax declaration verification. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of the execution flow of the declaration logic verification method based on the tax policy knowledge base provided in the embodiments of the present invention.

[0007] Figure 2 This is a schematic diagram of exemplary hardware and software components of the declaration logic verification system based on a tax policy knowledge base provided in an embodiment of the present invention. Detailed Implementation

[0008] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1This is a flowchart illustrating a declaration logic verification method based on a tax policy knowledge base, as provided in one embodiment of the present invention. The following is a detailed description of this declaration logic verification method based on a tax policy knowledge base.

[0009] Step S110: Call the preset tax policy knowledge base, perform visualization processing on the policy items in the base, extract the elements related to the declaration in the policy description and the relationship between the elements, and generate a visualization set of policy elements. The visualization set of policy elements contains the corresponding information of policy items and elements.

[0010] In this embodiment, the system first needs to access a pre-built tax policy knowledge base. This knowledge base is pre-constructed and contains various tax policy documents, such as tax laws and regulations, departmental rules, and normative documents issued by the State Taxation Administration. During the access process, the system establishes a connection with the tax policy knowledge base through an interface to obtain all policy entry data stored in the database. This policy entry data exists in text form and covers tax regulations for different tax types and business types, such as corporate income tax revenue recognition policies and value-added tax input tax deduction policies.

[0011] Step S111: Traverse all policy entries in the tax policy knowledge base, split each policy entry into natural language statements to obtain multiple independent policy statements, and add a corresponding policy entry number to each split policy statement.

[0012] Next, the system iterates through the acquired policy entries. The iteration order can be based on policy release date, tax type, or business area to ensure no policy entry is missed. For each policy entry, the system uses a sentence splitting tool in natural language processing technology to break it down into multiple independent policy statements based on punctuation marks (such as periods, semicolons, exclamation marks, etc.). For example, a policy entry regarding the additional deduction of R&D expenses for corporate income tax, stating that "For R&D expenses actually incurred by enterprises in R&D activities that do not form intangible assets and are included in current profit and loss, in addition to the deduction based on actual expenses as stipulated, an additional 75% of the actual amount incurred can be deducted before tax from January 1, 2018 to December 31, 2023; if intangible assets are formed, 175% of the intangible asset cost can be amortized before tax during the aforementioned period," results in two independent policy statements after splitting. After splitting, the system adds a corresponding policy entry number to each policy statement, which is associated with the unique identifier of the original policy entry. For example, if the original policy item is numbered "Finance and Taxation

[2018] No. 99", then the two policy statements after splitting can be numbered as "Finance and Taxation

[2018] No. 99-1" and "Finance and Taxation

[2018] No. 99-2" respectively.

[0013] Step S112: Perform semantic analysis on the policy statement to identify tax business keywords, declaration constraint statements and applicable prerequisites contained in the statement, and use the tax business keywords, declaration constraint statements and applicable prerequisites as initial elements.

[0014] Then, the system performs semantic parsing on the decomposed policy statements. During the semantic parsing process, the system loads a pre-trained tax domain semantic model, which can identify tax business keywords, declaration constraints, and applicable prerequisites in the policy statements. Tax business keywords refer to the core tax concepts involved in the policy statements, such as "R&D expenses," "additional deduction," "intangible assets," and "pre-tax amortization." Declaration constraints refer to the restrictions on declaration behavior or declaration data, such as "not forming intangible assets and included in current profit and loss" or "during the period from January 1, 2018 to December 31, 2023." Applicable prerequisites refer to the specific circumstances or conditions under which the policy applies, such as "actual R&D expenses incurred by the enterprise in its R&D activities." The system performs word-by-word analysis and contextual understanding of the policy statements through the semantic model, extracting the identified tax business keywords, declaration constraints, and applicable prerequisites as initial elements. For example, regarding the first policy statement after the above split, "For R&D expenses actually incurred by enterprises in their R&D activities that have not formed intangible assets and are included in the current period's profit and loss, in addition to the deduction based on actual expenses as stipulated, an additional 75% of the actual expenses can be deducted before tax during the period from January 1, 2018 to December 31, 2023," the identified initial elements include tax business keywords such as "R&D expenses," "not formed intangible assets," "current period's profit and loss," "deduction based on actual expenses," and "additional deduction before tax." The declaration constraints are "during the period from January 1, 2018 to December 31, 2023" and "75% of the actual expenses." The applicable premise is "R&D expenses actually incurred by enterprises in their R&D activities."

[0015] Step S113: Standardize the initial elements by merging initial elements with different descriptions but pointing to the same application requirements into the same standard element, and assign a unique element code to each standard element formed.

[0016] After obtaining the initial elements, the system needs to standardize them. Since different policy statements may use different expressions to convey the same reporting requirements, these initial elements with different expressions but pointing to the same goal need to be merged. For example, different policy statements may use the expressions "R&D expenditure" and "R&D expenses," which actually point to the same reporting requirement. Therefore, they need to be merged into the same standard element, "R&D expenses." The system determines whether the initial elements point to the same reporting requirement by comparing their semantic similarity. The semantic similarity calculation is based on a pre-trained word vector model. After converting the initial elements into vectors, the cosine similarity is calculated. When the similarity exceeds a preset threshold (e.g., 0.85), it is determined that they are different expressions but point to the same reporting requirement. After merging, the system assigns a unique element code to each standard element. The element code adopts specific coding rules, such as consisting of tax type code, business type code, and sequence number, ensuring that each standard element has a unique identifier. For example, the code for the standard element "R&D expenses" can be "QYDS-R&D-001," where "QYDS" represents corporate income tax, "R&D" represents R&D-related business, and "001" is the sequence number.

[0017] Step S114: Analyze the association patterns of standard elements in the same policy statement, identify and determine the preconditions, constraints, and execution order between each standard element in combination with tax policy semantic rules, and generate element association links.

[0018] Next, the system analyzes the association patterns of standard elements within the same policy statement. The system first extracts all standard elements from the same policy statement, and then uses tax policy semantic rules to identify the relationships between them. These tax policy semantic rules are summarized based on tax domain knowledge and policy expression habits. For example, in a policy statement, in the structure "If A, then B," A is usually a precondition of B; in the structure "Within the scope of C, execute D," C is the constraint scope of D; and in the structure "Execute E first, then execute F," E and F have an execution order relationship. Through syntactic analysis and semantic understanding, the system identifies association patterns such as preconditions, constraint scope, and execution order between standard elements. For example, in the policy statement regarding the additional deduction of R&D expenses mentioned above, "actual R&D expenses incurred by enterprises in carrying out R&D activities" is a prerequisite for "not forming intangible assets and being included in current profit and loss," and "the period from January 1, 2018 to December 31, 2023" is the scope of the constraint of "additional deduction of 75% of the actual amount incurred before tax." There is an execution order relationship between "deduction based on actual expenses as stipulated" and "additional deduction of 75% of the actual amount incurred before tax," meaning that actual deduction is performed first, followed by additional deduction. After identifying these association patterns, the system generates element association links, which are represented in the form of a directed graph. Nodes are standard elements, and directed edges represent the association relationships between standard elements, with the association type (such as prerequisites, constraint scope, and execution order) labeled on the edges.

[0019] Step S1141: Mark the position of the standard elements in a single policy statement, and record the grammatical position and order of each standard element in the policy statement.

[0020] When analyzing the association patterns of standard elements within the same policy statement, the first step is to mark the positions of the standard elements in each individual policy statement. The system performs syntactic analysis on the policy statement to determine the part of speech, syntactic components (such as subject, predicate, object, attributive, adverbial, etc.), and the position of the beginning and end characters of each standard element, thus recording its syntactic position. Simultaneously, the system numbers the standard elements according to the order in which they appear in the policy statement, recording their chronological order. For example, in the policy statement "For R&D expenses actually incurred by enterprises in their R&D activities that have not formed intangible assets and are included in the current period's profit and loss, in addition to the deduction based on actual expenses as stipulated, an additional 75% of the actual expenses can be deducted before tax during the period from January 1, 2018 to December 31, 2023," the standard elements include "actual R&D expenses actually incurred by enterprises in their R&D activities," "not formed intangible assets and included in the current period's profit and loss," "deducted based on actual expenses as stipulated," "during the period from January 1, 2018 to December 31, 2023," and "an additional deduction of 75% of the actual expenses before tax." Their order in the statement is 1, 2, 3, 4, and 5, respectively, and their grammatical positions are subject, conditional adverbial, predicate, time adverbial, and predicate.

[0021] Step S1142: Identify the grammatical relationships between standard elements by combining grammatical association rules, and distinguish between subject-verb association type, verb-object association type, and modifier association type.

[0022] Combining grammatical association rules, the system identifies grammatical associations between standard elements. These rules are based on summaries of Chinese grammatical structures; for example, there is a subject-verb association between the subject and predicate, a verb-object association between the predicate and object, and a modifying association between attributives / adverbs and the headword. The system determines the type of grammatical association between standard elements based on their grammatical position and part of speech. For example, there is a subject-verb association between "actual R&D expenses incurred by the enterprise in its R&D activities" (subject) and "deducted according to regulations based on actual expenses" (predicate); in "deducted at 75% of the actual amount incurred before tax" (predicate), "deducted at 75% of the actual amount incurred" is the core verb, and its associated object is likely "R&D expenses," thus exhibiting a verb-object association; "from January 1, 2018 to December 31, 2023" (time adverb) modifies "deducted at 75% of the actual amount incurred before tax" (predicate), thus exhibiting a modifying association.

[0023] Step S1143: In accordance with the expression habits of tax policies, strengthen the identification of special relationships between standard elements, specifically covering the relationship between preconditions and implementation content, the relationship between the scope of constraints and core requirements, and the relationship between process steps and their sequence.

[0024] In addition to grammatical connections, the system also incorporates the conventions of tax policy expression to strengthen the identification of special relationships between standard elements. Tax policy expressions typically exhibit certain norms and conventions; for example, they often explicitly stipulate the preconditions for policy application, the scope of constraints on the content of implementation, and the sequence of related procedural steps. Through studying a large number of tax policy texts, the system has summarized the expressive characteristics of these special relationships. For example, "not forming intangible assets and included in current profit and loss" is usually a precondition for "deducting an additional 75% of the actual amount incurred before tax," forming a connection between the precondition and the content of implementation; "from January 1, 2018 to December 31, 2023" is the time constraint for "deducting an additional 75% of the actual amount incurred before tax," forming a connection between the scope of constraints and the core requirements; "deducting according to regulations based on actual expenses" and "deducting an additional 75% of the actual amount incurred before tax" are two procedural steps for deducting R&D expenses, and they have a sequential order, forming a connection between procedural steps and their sequence.

[0025] Step S1144: Present the relationships between standard elements in the form of directed links. The link nodes are standard elements and element codes, and the connection between nodes is labeled with the association type and association description.

[0026] After identifying the various relationships between standard elements, the system presents these relationships in the form of directed links. Each node in the link contains the name and element code of the standard element; for example, a node can be represented as "R&D Expenses (QYDS-R&D-001)". The connection between nodes uses directed arrows to indicate the direction of the connection, with the arrow pointing from the source node to the target node. The connection type and description are marked on the connection lines. For example, on the connection line from the node "Not forming intangible assets and included in current profit and loss (QYDS-R&D-002)" to the node "75% of the actual amount incurred can be deducted before tax (QYDS-R&D-005)", the following is noted: "Precondition: This implementation content applies when not forming intangible assets and included in current profit and loss".

[0027] Step S1145: Perform logical verification on the element association links, correct grammatical conflicts or semantic contradictions in the association relationships, and ensure that the element association links directly reflect the inherent logic of the policy statements.

[0028] To ensure the accuracy of the element association links, the system needs to perform logical verification. Logical verification primarily checks for syntactic conflicts or semantic contradictions in the links. For example, it checks for circular associations (e.g., A is a precondition of B, and B is a precondition of A), incorrect association types (e.g., marking an association that should be a constraint as a precondition). When syntactic conflicts or semantic contradictions are found, the system adjusts according to preset correction rules. These correction rules are based on the inherent logic and semantic understanding of tax policies. For example, when circular associations occur, the system analyzes the context of the policy statement to determine the correct association direction and type; when the association type is incorrect, the system re-determines the association type based on the semantics of both parties and the policy expression habits. Through logical verification and correction, the element association links accurately and directly reflect the inherent logic of the policy statements.

[0029] Step S115: Link the standard elements, element codes, element association links, and policy item numbers, and classify and organize them according to tax business types to form a structured set of concrete policy elements.

[0030] Finally, the system integrates and links standard elements, element codes, element association links, and corresponding policy entry numbers. Specifically, each standard element is mapped to its own element code, the element association link of its corresponding policy statement, and the policy entry number. Then, these linked data are categorized and organized according to tax business types, which may include corporate income tax, value-added tax, and individual income tax. Under each business type, the relevant standard elements, element codes, element association links, and policy entry numbers are organized into a structured data format, such as JSON or XML. For example, under the corporate income tax R&D business type, it includes standard elements such as "R&D expenses" and "additional deduction ratio," along with their corresponding element codes, element association links, and policy entry numbers. This processing ultimately forms a structured set of visualized policy elements, which clearly displays the elements in the policy entries and their relationships.

[0031] Step S120: Obtain the declaration data submitted by the taxpayer and the associated declaration scenario description, decompose the declaration data into basic field elements, analyze the correlation attributes between field elements to form a declaration element set, and bind the declaration element set with the declaration scenario description to obtain a scenario-based declaration element set.

[0032] After generating the concrete set of policy elements, the system enters the stage of acquiring taxpayer declaration data and generating a set of scenario-based declaration elements. The system receives declaration data and associated declaration scenario descriptions submitted by taxpayers through an interface with the tax declaration system. Declaration data consists of various forms filled out by taxpayers when filing tax returns, such as the annual corporate income tax return and the value-added tax return. This data is transmitted to the system in the form of electronic forms. The declaration scenario description is a description of the relevant circumstances of the current declaration business by the taxpayer. It is selected or filled in by the taxpayer during the declaration process based on the actual business situation, covering the industry to which the declared business belongs (such as manufacturing, service industry, finance, etc.), the time period of the business (such as 2023, the first quarter of 2023, etc.), and the characteristics of the economic behavior (such as R&D activities, sales of goods, provision of services, etc.).

[0033] Step S121: Receive the declaration data and declaration scenario description submitted by the taxpayer through the declaration terminal. The declaration scenario description covers the industry to which the declaration business belongs, the time period of the business, and the characteristics of the economic behavior.

[0034] When receiving data, the system first performs a preliminary check on its completeness and format. It checks whether the declared data includes all required forms and fields, and whether the data format conforms to the system's standards (e.g., date format YYYY-MM-DD, numeric data retained to two decimal places). Simultaneously, it checks whether the declaration scenario description fully includes information such as the industry, time period of the business, and characteristics of the economic behavior. If incomplete data or non-compliant formatting is found, the system will return a prompt to the declaration terminal, requesting the taxpayer to supplement or correct the data. For example, if the taxpayer has not filled in the industry of the declared business, the system will prompt "Please select the industry of the declared business."

[0035] Step S122: Decompose the declaration data into a structured form, and extract the field name, data content, reporting scope and data source of each field as basic field elements.

[0036] Next, the system performs structured decomposition on the declared data that has passed the initial check. This decomposition process involves splitting the electronic forms in the declared data according to their structure and field definitions, resulting in multiple independent fields. For example, the "Detailed Statement of Additional Deduction for R&D Expenses" in the annual corporate income tax return contains multiple fields, such as "R&D Project Name," "R&D Personnel Labor Costs," and "Direct Input Costs." The system uses a form parsing tool to identify and decompose these fields based on a predefined form template. For each decomposed field, the system extracts its field name (e.g., "R&D Personnel Labor Costs"), data content (i.e., the specific numerical value or text entered by the taxpayer), reporting scope (e.g., according to which accounting system or tax regulations this field should be reported), and data source (e.g., from the company's accounting books, original vouchers, etc.).

[0037] Step S123: Analyze the association attributes of different basic field elements, identify field combinations with numerical derivation, conditional dependence, and hierarchical subordination based on the field data association characteristics, and record the association methods within the field combinations as associated field elements.

[0038] After obtaining the basic field elements, the system needs to analyze the association attributes between different basic field elements. Association attributes reflect the inherent connections between fields and are crucial for understanding the logical structure of the declaration data and performing declaration logic verification. By analyzing the association characteristics of field data, the system identifies field combinations with numerical derivation, conditional dependence, and hierarchical subordinate relationships, and records the association methods within these combinations, forming associated field elements.

[0039] Step S1231: Select a basic field element from the declaration data as the target element, and retrieve other basic field elements in the declaration element set that have data association with the target element to form a candidate set of associations.

[0040] The system first selects one target element from the basic field elements. The selection order can follow the order of the fields in the form or be random. For example, "Labor Costs for R&D Personnel" can be selected as the target element. Then, the system retrieves all other basic field elements in the application element set that have data relationships with the target element. Data relationships can be determined by analyzing the relevance of field names, the logical relationships of data content, and the cross-referencing relationships in the form. For example, "Labor Costs for R&D Personnel" may have data relationships with fields such as "Salaries and Wages of Personnel Directly Engaged in R&D Activities" and "Social Insurance and Housing Fund for Personnel Directly Engaged in R&D Activities," because "Labor Costs for R&D Personnel" is usually obtained by adding the values ​​of these fields. These field elements that have data relationships with the target element are collected to form a candidate set of relationships.

[0041] Step S1232: Analyze the data content of the target element and each element in the candidate set of related elements, compare the numerical correlation characteristics of the target element and related elements to determine the source of the target element's numerical value, and record the relevant correlation characteristics as numerical derived correlation.

[0042] For each element in the candidate set of associations, the system analyzes its data content with the target element and compares their numerical association characteristics. Numerical association characteristics include numerical calculation relationships (such as addition, subtraction, multiplication, and division), proportional relationships, etc. By analyzing these characteristics, the system determines whether the value of the target element is derived from the value of an element in the candidate set of associations. For example, if the value of the target element "labor costs for R&D personnel" is equal to the sum of the values ​​of the two elements "salaries and wages of personnel directly engaged in R&D activities" and "social insurance and housing fund for personnel directly engaged in R&D activities" in the candidate set of associations, then it means that "labor costs for R&D personnel" is derived from the values ​​of these two elements. The system records this numerical derivation relationship, including the field combinations involved in the derivation (such as "salaries and wages of personnel directly engaged in R&D activities" and "social insurance and housing fund for personnel directly engaged in R&D activities") and the specific operational relationships (such as summation), as a numerical derivation association.

[0043] Step S1233: Check whether the attributes of the target element are restricted by the attributes of other elements. When a certain element meets a specific attribute condition, the attribute of the target element needs to meet the corresponding requirements. Record the restriction relationship between the attribute condition of a certain element and the attribute of the target element as a conditional dependency association.

[0044] In addition to numerical derivation associations, the system also needs to check whether the attributes of the target element are restricted by the attributes of other elements, i.e., conditional dependency associations. Attributes of the target element include data type, value range, and mandatory content. The system examines each element in the candidate association set to determine whether its attributes restrict the attributes of the target element. For example, if the candidate association set contains a "R&D activity type" field, when the attribute value of this field is "independent R&D," the value range of the target element "R&D personnel labor costs" may have specific requirements (such as not exceeding a certain limit); when the "R&D activity type" is "commissioned R&D," the value range of the target element may be different. The system records this kind of constraint relationship where, when a certain element meets a specific attribute condition, the target element's attribute must meet the corresponding requirement, as a conditional dependency association.

[0045] Step S1234: Identify whether there is a feature attribute that is a detailed description of the target feature attribute. When a feature attribute is a detailed description of the target feature attribute, the feature attribute belongs to the target feature attribute. Record the hierarchical relationship between the feature attribute and the target feature attribute as a hierarchical subordinate association.

[0046] In addition, the system also needs to identify hierarchical subordinate relationships. Hierarchical subordinate relationships refer to a situation where a certain element attribute is a more detailed description of a target element attribute; that is, the element attribute is subordinate to the target element attribute. For example, if the target element is "total R&D expenses," the candidate set of related elements might include elements such as "personnel costs," "direct input costs," and "depreciation expenses." These elements are all more detailed classifications of "total R&D expenses," and their attributes are subordinate to the attributes of "total R&D expenses." The system identifies this hierarchical subordinate relationship by analyzing the semantics of the fields and the structural relationships of the forms, recording the hierarchical relationship where a certain element attribute is subordinate to the target element attribute as a hierarchical subordinate relationship.

[0047] Step S1235: Integrate the numerical derivation association, conditional dependency association, and hierarchical subordinate association, and label the field combinations and operation relationships corresponding to the numerical derivation association, the field combinations and restriction relationships corresponding to the conditional dependency association, and the field combinations and hierarchical relationships corresponding to the hierarchical subordinate association to form a set of associated field elements.

[0048] After identifying numerical derivation associations, conditional dependency associations, and hierarchical subordinate associations, the system integrates these associations. During the integration process, the system labels the corresponding field combinations and specific association relationships for each association type. For numerical derivation associations, it labels the field combinations involved in the derivation (e.g., field A and field B) and the operational relationship (e.g., A+B); for conditional dependency associations, it labels the field combinations that trigger the condition (e.g., field C equals a specific value) and the restriction relationship on the target field (e.g., the value range of field D); for hierarchical subordinate associations, it labels the subordinate field combinations (e.g., field E and field F) and the hierarchical relationship (e.g., E and F are children of G). Through this integration, a set of association field elements containing all types of association relationships is formed.

[0049] Step S124: Integrate basic field elements and related field elements by adding field identifiers to both basic field elements and related field elements to form a set of declaration elements that includes independent field attributes and related attributes.

[0050] After obtaining the basic and related field elements, the system integrates them. During integration, it's crucial to ensure each field element is clearly identifiable; therefore, the system adds field identifiers to both basic and related field elements. Field identifiers can be unique strings composed of letters, numbers, and symbols. For example, basic field element identifiers can use the prefix "J-" followed by a sequence number (e.g., J-001, J-002), while related field element identifiers can use the prefix "G-" followed by a sequence number (e.g., G-001, G-002). After adding field identifiers, the system associates the independent attributes of the basic field elements (field name, data content, reporting scope, data source, etc.) with the related attributes of the related field elements (association type, related field combination, association relationship, etc.). This creates a complete set of reporting elements that includes both independent field attributes and related attributes.

[0051] Step S125: Extract scenario feature words from the declaration scenario description, establish a correspondence with the field identifiers in the declaration element set, add the scenario feature words as attribute tags to the corresponding field elements, and obtain the scenario-based declaration element set.

[0052] Finally, the system needs to bind the application element set with the application scenario description to generate a scenario-based application element set. First, the system performs text analysis on the application scenario description, extracting scenario-specific keywords. Scenario-specific keywords are key terms that reflect the application's business scenario. For example, if the application scenario description is "Application for Additional Deduction of R&D Activity Expenses for Manufacturing Enterprises in 2023," the scenario-specific keywords include "manufacturing," "2023," "R&D activities," and "additional deduction of expenses." Then, the system establishes a correspondence between these scenario-specific keywords and the field identifiers in the application element set. This correspondence is based on the business meaning of the field elements and the relevance of the scenario-specific keywords. For example, the scenario-specific keyword "R&D activities" is associated with the "R&D Project Name" field (identified by J-001) and the "R&D Expense Related Combinations" field (identified by G-001) in the application element set. After establishing the correspondence, the system adds the scenario-specific keywords as attribute tags to the corresponding field elements. For example, adding attribute tags such as "R&D activities" and "2023" to the "R&D Project Name" field. Through this processing, the application element set acquires scenario attributes, forming a scenario-based application element set.

[0053] Step S130: Import the scenario-based declaration element set and the policy element visualization set into the element association mechanism, lock the appropriate policy through scenario matching, perform attribute comparison between field elements and policy elements, establish a mapping link and generate element mapping results.

[0054] After generating the scenario-based declaration element set and the policy element representation set, the system needs to import these two datasets into the element association mechanism to achieve accurate matching and association between the declaration data and tax policies. The element association mechanism is a system module that integrates functions such as policy indexing, scenario matching, and attribute comparison. Its core objective is to find the corresponding policy element for each field element in the declaration data through multi-dimensional analysis and comparison.

[0055] Step S131: Import the concrete set of policy elements into the policy index module of the element association mechanism, construct a retrieval index with standard elements as the core, and associate the corresponding element association links and policy entry numbers with the index.

[0056] First, the system imports the visualized set of policy elements into the policy index module of the element association mechanism. The policy index module is a fundamental component of the element association mechanism, responsible for organizing and managing policy element data to facilitate efficient retrieval and matching. During the import process, the system constructs a retrieval index based on the standard elements in the visualized set of policy elements. When constructing the retrieval index, the system creates an index item for each standard element, containing the standard element's name, element code, and core attributes (such as policy constraints and applicable scenario characteristics). Simultaneously, the system associates and stores the corresponding element association link and policy entry number for each standard element with the index item. For example, for the standard element "R&D expenses," its index item contains the name "R&D expenses," the element code "QYDS-R&D-001," policy constraints (such as the conditions and proportions for additional deductions), and applicable scenario characteristics (such as R&D activities), and is associated with the element association link of this standard element and the corresponding policy entry number "Finance and Taxation

[2018] No. 99." In this way, the retrieval index can quickly locate the policy elements and their associated information related to the declared data.

[0057] Step S132: Extract scenario feature words from the scenario-based declaration element set, match the scenario feature words with the policy business type tags in the policy index module, and lock in the policy scope that is suitable for the declaration scenario.

[0058] Next, the system extracts scenario-specific vocabulary from the set of scenario-based declaration elements. These scenario-specific vocabulary are attribute tags previously added to the field elements, such as "manufacturing," "2023," "R&D activities," and "expense deduction." These terms accurately reflect the specific scenario of the declaration business. Then, the system matches the extracted scenario-specific vocabulary with policy business type tags in the policy index module. These policy business type tags are classification tags added to policy elements according to tax business types when constructing the concrete set of policy elements, such as "corporate income tax - R&D expense deduction" and "value-added tax - input tax deduction." The matching process uses a semantic similarity calculation method. The scenario-specific vocabulary and policy business type tags are converted into vector representations, and the similarity is calculated. When the similarity reaches a preset threshold, a successful match is determined. For example, matching the scenario-specific vocabulary "R&D activities" and "expense deduction" with the policy business type tag "corporate income tax - R&D expense deduction" is successful due to their high semantic relevance. Through this scenario matching, the system can pinpoint the policy scope suitable for the current declaration scenario, significantly reducing the search space for subsequent attribute comparisons and improving matching efficiency and accuracy.

[0059] Step S133: Extract individual field elements from the scenario-based declaration element set and generate field description text containing field name, data content, and scenario attributes.

[0060] After identifying the applicable policy scope, the system extracts individual field elements from the set of scenario-based declaration elements. The extraction process is based on field identifiers, with each identifier corresponding to an independent field element. For each extracted field element, the system generates a field description text. This description text provides a comprehensive overview of the key information of the field element, including the field name (e.g., "R&D personnel labor costs"), data content (i.e., the specific data information filled in by the taxpayer, represented by placeholders, such as "[specific amount]"), and scenario attributes (i.e., added scenario-specific keywords, such as "R&D activities" and "2023").

[0061] Step S134: Compare the field description text with the standard elements within the scope of the adaptation policy, determine the standard elements with the highest degree of fit by comparing the overlap of attribute dimensions, and establish a mapping link between field elements and standard elements.

[0062] Then, the system compares the generated field description text with the standard elements within the scope of the applicable policy. Attribute comparison is the core step in element association. By comparing the overlap between field elements and policy elements across multiple attribute dimensions, the system determines the degree of fit between the two, thereby finding the most matching policy element.

[0063] Step S1341: Extract the core attributes from the field description text, specifically covering the field data type, filling requirements, scenario adaptation features, and data format specifications, to form a set of field attributes.

[0064] The system first extracts attributes from the field description text to obtain the core attributes of the field elements. These core attributes specifically include the field data type (e.g., numeric, text, date), reporting requirements (e.g., required, optional, and whether null values ​​are allowed), scenario adaptation features (i.e., the specific business scenario to which the field applies, such as R&D expense reporting), and data format specifications (e.g., the unit of measurement for numeric values, decimal place requirements, and text length limits). For example, for the "R&D personnel labor costs" field, its core attributes might be: numeric data type, required reporting requirement, scenario adaptation feature of R&D activity expense reporting, and data format specifications of retaining two decimal places and using RMB as the unit. These core attributes are then integrated to form the field attribute set.

[0065] Step S1342: Extract the core attributes of each standard element included in the scope of the applicable policy, specifically covering policy constraint requirements, applicable scenario characteristics, data caliber standards, and compliance boundary definitions, to form a set of policy attributes.

[0066] Simultaneously, the system extracts the core attributes of each standard element within the scope of the applicable policy. The core attributes of standard elements differ from those of field elements, primarily revolving around policy provisions. Specifically, they include policy constraints (such as the conditions under which the element applies and what restrictions apply), applicable scenario characteristics (i.e., the business scenario to which the policy element applies, such as the scenario for additional deduction of R&D expenses for corporate income tax), data caliber standards (such as the policy's definition of the element's data and calculation methods), and compliance boundary definitions (such as the range of data values ​​and the allowable error range). For example, the core attributes of the policy element "R&D expenses" might be: policy constraints requiring that R&D expenses not formed into intangible assets and included in current profit and loss are eligible for additional deduction; applicable scenario characteristics being the declaration of R&D activities for corporate income tax; data caliber standards including personnel costs and direct input costs; and compliance boundaries defined as actual expenses related to R&D activities. These core attributes are integrated to form a set of policy attributes.

[0067] Step S1343: Compare the field attribute set with the policy attribute set dimension by dimension, and count the number of dimensions that are completely matched, partially matched, and completely mismatched.

[0068] Next, the system compares the set of field attributes with the set of policy attributes dimension by dimension. The comparison process is performed sequentially according to the attribute dimensions, and the comparison results for each dimension are divided into three categories: complete match, partial match, and complete mismatch. A complete match means that the content of the field attribute and the policy attribute is completely identical in that dimension; a partial match means that the two have some similarity or overlap, but are not completely identical; a complete mismatch means that the two have no correlation or obvious conflict in that dimension. For example, in the scenario adaptation feature dimension, if the field attribute is "Declaration of R&D Activity Expenses" and the policy attribute is "Enterprise Income Tax R&D Expense Additional Deduction Scenario", it is judged as a complete match; in the data caliber standard dimension, if the "R&D personnel labor costs" in the field attribute is included in the data caliber range of the "R&D expenses" in the policy attribute, it is judged as a partial match; in the data type dimension, if the field attribute is numeric and the policy attribute also requires it to be numeric, it is judged as a complete match; if the policy attribute requires it to be text, it is judged as a complete mismatch. The system records the comparison results for each dimension and counts the number of dimensions with complete matches, partial matches, and complete mismatches.

[0069] Step S1344: Statistically determine the number of matching dimensions to determine the attribute matching degree, select the standard element with the highest matching degree as the matching object, and if there are multiple standard elements with the same matching degree, further filter them by combining scene feature words.

[0070] After completing the dimension-by-dimensional comparison, the system determines the attribute fit based on the statistically analyzed number of dimension matches. The attribute fit is calculated by multiplying the number of fully matching dimensions by a higher weight (e.g., 2) and adding the number of partially matching dimensions by a lower weight (e.g., 1), then dividing by the total number of attribute dimensions. This method yields the attribute fit value for each standard element and the field element. The system selects the standard element with the highest attribute fit as the matching target for the field element. If multiple standard elements have the same highest attribute fit, the system further filters by incorporating scenario feature vocabulary. Specifically, it compares the matching degree between the applicable scenario features of these standard elements and the scenario feature vocabulary of the field element, selecting the standard element with the highest matching degree as the final matching target. For example, if two standard elements have the same fit, but one of them contains more terms in its applicable scenario features that are identical to the scenario feature vocabulary of the field element, then that standard element is selected.

[0071] Step S1345: Establish the correspondence between target field elements and matching standard elements, associate the attribute matching details of target field elements and matching standard elements, and generate a mapping link containing bidirectional association information between target field elements and matching standard elements.

[0072] After determining the matching standard elements, the system establishes a correspondence between the target field element and the matching standard element. This correspondence is represented in the form of a mapping link, which includes the field identifier of the target field element, the element code of the matching standard element, and details of the attribute matching between the two (such as which dimensions match completely and which dimensions match partially). For example, a mapping link is established between the "R&D personnel labor costs" field element with the field identifier J-001 and the "R&D expenses" standard element with the element code QYDS-R&D-001. The link records details of attribute matching, such as complete matching of scenario adaptation feature dimensions and partial matching of data caliber standard dimensions. By generating such a mapping link, a two-way association between the declaration data field elements and the tax policy standard elements is achieved.

[0073] Step S135: Record the attribute matching points, associated element links, and policy entry numbers of each mapping link, and integrate all mapping links to form an element mapping result containing matching information.

[0074] Finally, the system records and integrates the relevant information for each mapping link. Specifically, the system records the attribute matching points between field elements and standard elements in each mapping link (i.e., attribute dimensions and specific content of complete and partial matching), the element association links associated with the standard element in the policy element visualization set, and the corresponding policy entry number. For example, the attribute matching points of a certain mapping link include complete matching of scenario adaptation features (R&D activity scenario) and complete matching of data type (numerical type), the associated element association link is "R&D expenses (QYDS-R&D-001) → Additional deduction ratio (QYDS-R&D-003)", and the policy entry number is "Finance and Taxation

[2018] No. 99". The system integrates all generated mapping links and their associated information, organizes them according to the order of field elements or business type, and forms an element mapping result containing complete matching information.

[0075] Step S140: Filter deviation links from the element mapping results, extract the corresponding declaration field elements and policy elements, locate the root cause of the difference and track the chain effect of the difference on related fields, and sort out the transmission path to obtain the declaration difference transmission information.

[0076] After generating the element mapping results, the system needs to analyze the mapping links to identify the differences between the declaration data and tax policies, and track the possible chain effects of these differences, ultimately forming information on the transmission of declaration differences.

[0077] Step S141: Traverse all mapping links in the feature mapping results, filter out deviation links whose matching degree does not meet the preset attribute matching degree benchmark, and mark the declaration field elements corresponding to the deviation links.

[0078] First, the system traverses all mapping links in the element mapping results. The traversal process follows the storage order of the mapping links to ensure that each link is checked. During the traversal, the system evaluates the fit of each mapping link based on a preset attribute fit benchmark. The attribute fit benchmark is a pre-set threshold based on the strictness of tax policies and the importance of the declared data. For example, for key declaration fields, the fit benchmark might be set to 0.8, and for non-key fields, it might be set to 0.6. When the attribute fit of a mapping link is lower than the preset benchmark, the system classifies it as a deviation link. Simultaneously, the system marks the declaration field element corresponding to the deviation link and records information such as the field identifier and field name. For example, if the attribute fit of the "R&D personnel labor costs" field element with the policy element is 0.5, which is lower than the preset benchmark of 0.7, then the mapping link is marked as a deviation link, and the relevant information of that field element is recorded.

[0079] Step S142: Extract the standard elements and element association links associated with the deviation links, extract the attribute requirements of policy elements on the declaration fields from the element association links, and determine the attribute requirements of policy elements on the declaration fields extracted from the element association links as policy benchmark attributes.

[0080] For the selected deviation links, the system extracts their associated standard elements and element association links. Standard elements are the policy elements matched by the deviation links, while element association links reflect the relationships between these standard elements and other policy elements. The system extracts the policy element attribute requirements for the declared fields from the element association links. These attribute requirements include policy stipulations regarding field data types, data scope, reporting requirements, and compliance boundaries. For example, from the "R&D expenses" element association link, the system can extract the policy attribute requirements for "R&D personnel labor costs": the data type should be numeric, the data scope should include the wages, salaries, social insurance, and housing fund contributions of personnel directly engaged in R&D activities, the reporting requirement is mandatory, and the compliance boundary is the actual expenses related to R&D activities. The system determines these extracted attribute requirements as policy benchmark attributes, serving as the standard for judging whether the declared field elements are compliant.

[0081] Step S143: Compare the actual attributes of the declared field elements with the policy benchmark attributes, record the specific manifestations of attribute differences dimension by dimension, and combine the root causes of the differences in the element association links of the deviation link.

[0082] Next, the system compares the actual attributes of the declared field elements with the policy benchmark attributes. The comparison process proceeds sequentially along attribute dimensions (such as data type, data scope, and reporting requirements), recording the specific manifestations of attribute differences dimension by dimension. For example, in the data scope dimension, if the actual data scope of the declared field element only includes R&D personnel salaries, while the policy benchmark requires both salaries and social insurance and housing provident fund contributions, it is recorded as "Incomplete data scope, missing social insurance and housing provident fund contributions." In the reporting requirements dimension, if the declared field element is optional but the policy benchmark requires it to be mandatory, it is recorded as "Reporting requirements do not meet, should be mandatory." After recording the attribute differences, the system uses the element association links associated with the deviation links to pinpoint the root cause of the differences. The element association links contain the relationships and constraints between policy elements. By analyzing these relationships, the system can trace the specific policy clauses or logic that led to the attribute differences. For example, by analyzing the element association links, it was found that the policy requirement for the data scope of "R&D personnel labor costs" originates from a specific policy clause, while the declared data does not meet the requirements of that clause, thus pinpointing the root cause of the difference.

[0083] Step S144: Based on the related field element information in the declaration element set, track the correlation attributes of the difference field with other fields, and analyze the impact of the difference on the attributes of the related field elements.

[0084] After locating the root cause of the discrepancy, the system needs to analyze its impact on other related field elements. Based on the related field element information in the declaration element set, the system identifies related field elements that are directly or indirectly related to the discrepancy field. The related field element information records the numerical derivation, conditional dependency, hierarchical subordination, and other relationships between fields. For example, if "R&D personnel labor costs" is a numerically derived field of "Total R&D expenses" (Total R&D expenses = R&D personnel labor costs + direct input costs + ...), then a difference in "R&D personnel labor costs" may lead to a difference in "Total R&D expenses." The system analyzes the impact of the discrepancy field on the attributes of these related field elements, including the accuracy of the data content and compliance with reporting requirements. For example, if the value of "R&D personnel labor costs" is higher than the policy-allowed range, it may lead to an inflated calculation result for "R&D expense deduction," thereby affecting the attributes of related fields such as "taxable income."

[0085] Step S145: Record the initial difference field, derived difference field, and associated attributes in the order of difference transmission, record the hierarchical relationship and degree of impact of difference transmission, and form the declaration difference transmission information.

[0086] After analyzing the impact of differences on related fields, the system records relevant information according to the order of difference propagation, forming the declaration difference propagation information. The difference propagation order refers to the sequence in which differences are passed from the initial difference field to the derived difference fields, reflecting the diffusion path of differences in the declaration data.

[0087] Step S1451: Take the initial difference field as the starting point of the transmission, record its field identifier, difference performance and difference root cause, and use it as the first level node of the transmission link.

[0088] The system uses the initial difference field as the starting point of the transmission link, i.e., the first-level node. At this node, the system records the field identifier of the initial difference field (e.g., J-001), the nature of the difference (e.g., incomplete data caliber), and the root cause of the difference (e.g., excluding social insurance and housing fund contributions). This information forms the basis of the transmission link, clearly demonstrating the initial state of the difference.

[0089] Step S1452: Based on the information of related field elements, find the related fields that are directly related to the initial difference field, analyze the derived differences caused by the initial difference in the related fields, record the performance of the derived differences and the related attributes, and use them as the second-level nodes of the transmission link.

[0090] Next, based on the associated field element information, the system searches for related fields that are directly related to the initial difference field. Direct relationships include direct calculation relationships in numerical derivation, direct triggering relationships in conditional dependencies, and direct hierarchical relationships in hierarchical subordination. For each directly related field found, the system analyzes the derived differences caused by the initial difference. For example, if the value of the initial difference field "R&D personnel labor costs" is high, the value of the directly related field "Total R&D expenses" (related through numerical derivation) will also be high, resulting in a derived difference. The system records the manifestation of the derived difference (e.g., the total R&D expenses are high) and the association attributes (e.g., numerical derivation relationship: total R&D expenses = R&D personnel labor costs + direct input costs), using this information as the second-level nodes in the transmission chain.

[0091] Step S1453: Using the derived difference field in the second-level node as the new propagation starting point, continue to search for other fields that are related to the derived difference field, and repeat the derived difference analysis process until no new derived difference fields appear.

[0092] Then, the system uses the derived difference fields in the second-level nodes as the new starting point for propagation, continuing to search for other fields related to these derived difference fields. This process is similar to the previous step, but the starting point for propagation becomes the derived difference fields. The system repeats the search for related fields and the analysis of derived differences, continuously expanding the levels of the propagation link. For example, the derived difference field of "Total R&D Expenses" in the second-level nodes may have a numerical derived relationship with "R&D Expense Additional Deduction Amount" (Additional Deduction Amount = Total R&D Expenses × Additional Deduction Ratio). In this case, "R&D Expense Additional Deduction Amount" will generate a derived difference due to the higher "Total R&D Expenses," becoming a third-level node in the propagation link. The system continues this process until no new derived difference fields are found.

[0093] Step S1454: Use a hierarchical structure diagram to present the difference transmission process. Label each node with field identifiers, difference types, related attributes, and degree of influence. Use directional lines between nodes to indicate the direction of difference transmission.

[0094] To visually demonstrate the difference propagation process, the system uses a hierarchical structure diagram. Each node in the diagram represents a field element with a difference, labeled with the field identifier, difference type (e.g., initial difference, derived difference), related attributes (e.g., numerical derivation, conditional dependency), and degree of impact (e.g., high, medium, low). The degree of impact is assessed based on the difference's importance to the field and its influence on subsequent reporting data. For example, a difference that directly affects the tax amount is labeled as high. Directed lines connect nodes to indicate the direction of difference propagation, with arrows pointing from the source node to the target node, clearly showing the path of difference transmission. For example, an arrow pointing from the "R&D personnel labor costs" node to the "total R&D expenses" node indicates that the difference is propagated from the former to the latter.

[0095] Step S1455: Organize the information in the hierarchical structure diagram, write the difference transmission description according to the transmission order, mark the causal relationship of the differences in each field, and form a complete declaration difference transmission information.

[0096] Finally, the system organizes the information in the hierarchical structure diagram and writes a difference transmission description according to the transmission order. The difference transmission description is in the form of natural language text, describing in detail the field differences of each level node, the difference transmission path, and the causal relationship between the differences of each field. For example, "The initial difference field 'R&D personnel labor costs' (J-001) has a lower value due to incomplete data scope (excluding social insurance and housing fund) (root cause of difference); this difference is transmitted to 'total R&D expenses' (J-005) through numerical derivation association, causing its value to be lower (derived difference manifestation); furthermore, it affects 'R&D expense additional deduction' (J-008) through numerical derivation association, causing its calculation result to be lower (derived difference manifestation), ultimately affecting the accuracy of 'taxable income' (J-012)." Through this organization, a complete declaration difference transmission information is formed, which reflects the generation, transmission path, and scope of impact of differences in the declaration data.

[0097] Step S150: Integrate the information on differences in application submission with the set of scenario-based application elements, combine policy elements to generate field adaptation suggestions and policy basis, organize the content according to the application order to form an application logic verification report, and push the report to the application terminal.

[0098] After generating the information on the transmission of differences in declarations, the system needs to integrate it with the set of scenario-based declaration elements, and combine it with policy elements to generate specific field adaptation suggestions and policy basis. Finally, a declaration logic verification report is generated and pushed to the declaration terminal so that taxpayers can understand the problems in the declaration data and make corrections.

[0099] For example, in step S151: organize the scenario-based declaration element set according to the original filling order of the declaration data, add difference markers next to the field elements with differences, and associate them with the corresponding nodes in the declaration difference transmission information.

[0100] First, the system organizes the scenario-based declaration element set. The organization process follows the original order in which the declaration data was filled in, that is, arranging the field elements according to the order in which taxpayers filled in the declaration forms. This ensures that the subsequently generated verification report is consistent with taxpayers' filing habits and is easier for taxpayers to understand. During the organization process, the system adds difference markers next to field elements that show discrepancies. These difference markers can be special symbols (such as "..."). A label or color code is used to prominently alert taxpayers that there is a problem with the field. Simultaneously, the system associates the difference marker with the corresponding node in the difference transmission information, allowing users to quickly locate a detailed description of the field in the difference transmission information when viewing the difference field, such as the nature of the difference, its root cause, and its impact on related fields. For example, adding a label next to the "R&D personnel labor costs" field... The field is marked and associated with the first-level node information corresponding to that field in the declaration difference transmission information.

[0101] Step S152: For each difference field, combine the associated standard elements and policy entries to generate specific field adaptation suggestions, and mark the direction of field attribute adjustment and the basis for adjustment.

[0102] For each discrepancy in a field element, the system generates specific field adaptation suggestions based on its associated standard elements and policy entries. These suggestions are proposed modifications to the discrepancy field, aiming to help taxpayers adjust their declared data to comply with policy requirements. When generating suggestions, the system first analyzes the differences between the actual attributes of the discrepancy field and the policy benchmark attributes, and then determines the direction of attribute adjustment based on policy requirements. For example, if the discrepancy is due to incomplete data scope, the adaptation suggestion might be "Please supplement the social insurance and housing fund contributions included in the labor costs of R&D personnel"; if the discrepancy is due to values ​​exceeding compliance boundaries, the suggestion might be "Please verify the calculation basis of the labor costs of R&D personnel to ensure that it does not exceed the scope stipulated by policy." Simultaneously, the system marks the adjustment basis in the adaptation suggestions, i.e., the specific policy clauses or standard element attributes upon which the suggestion is based. For example, the adjustment basis could be "According to the provisions on the data scope of R&D expenses in the Ministry of Finance and State Taxation Administration document

[2018] No. 99, the labor costs of R&D personnel should include wages, salaries, social insurance and housing fund contributions, etc."

[0103] Step S153: Extract the complete text of the policy entries corresponding to the differences, mark them as policy basis next to the adaptation suggestions, and highlight the policy statements related to the differences.

[0104] To enhance the authority and persuasiveness of the field matching suggestions, the system extracts the complete text of the policy entries corresponding to the differences and marks them next to the matching suggestions as policy basis. The complete text of the policy entries is the original policy document content obtained from the tax policy knowledge base, ensuring the accuracy and completeness of the policy basis. During the annotation process, the system uses special formatting (such as bolding and underlining) to highlight the policy expressions related to the differences, so that taxpayers can quickly find the core content related to the current difference field in the policy basis. For example, for the matching suggestion of the "R&D personnel labor costs" field, the system extracts the complete clause on the composition of R&D expenses in the document Caishui

[2018] No. 99 and highlights the part involving personnel labor costs.

[0105] Step S154: In the declaration logic verification report, set up the declaration data display section, the difference details section, the adaptation suggestion section, and the policy basis section in sequence, and add a clear title to each section to form a declaration logic verification report with clear hierarchy.

[0106] The system organizes the content of the declaration logic verification report according to a specific structure to ensure a clear hierarchy and ease of reading. The report sequentially includes a declaration data display section, a difference details section, an adaptation suggestion section, and a policy basis section, each with a clear title. The declaration data display section shows all fields in the scenario-based declaration element set in the original submission order, with difference markers added next to fields showing discrepancies. The difference details section lists the field identifier, field name, difference manifestation, and root cause of each discrepancy field. The adaptation suggestion section provides specific adjustment suggestions for each discrepancy field. The policy basis section displays the complete text of the policy entries related to the differences. Through this structural arrangement, taxpayers can progressively understand the overall situation of the declaration data, existing problems, how to modify it, and the policy basis for modification.

[0107] Step S155: Send the declaration logic verification report to the corresponding declaration terminal through the encrypted data channel, configure the declaration terminal to support online viewing, downloading and printing functions, and synchronously record the declaration logic verification report push time and the declaration terminal receiving status.

[0108] Finally, the system sends the generated declaration logic verification report to the taxpayer's corresponding declaration terminal via an encrypted data channel. The encrypted data channel uses secure communication protocols (such as SSL / TLS) to encrypt the report data during transmission, ensuring data security and confidentiality and preventing information leakage. After receiving the report, the declaration terminal is configured to support multiple operation functions, including online viewing (allowing taxpayers to directly browse the report content on the terminal), downloading (allowing taxpayers to save the report to their local devices), and printing (allowing taxpayers to print the report for their records). Simultaneously, the system records the push time of the declaration logic verification report and the declaration terminal's receiving status (e.g., received, not received, read). This recorded information helps tax authorities track the report's delivery, ensuring taxpayers can obtain the verification results promptly and take appropriate action. For example, the system records the report push time as "2023-10-15 14:30:00" and the receiving status as "received (2023-10-15 14:32:15)".

[0109] Figure 2 The illustration shows exemplary hardware and software components of a tax policy knowledge base-based declaration logic verification system 100, which can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the tax policy knowledge base-based declaration logic verification system 100 and to perform the functions in this application.

[0110] The declaration logic verification system 100 based on a tax policy knowledge base can be a general-purpose server or a special-purpose server; both can be used to implement the declaration logic verification method based on a tax policy knowledge base of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0111] For example, the tax policy knowledge base-based declaration logic verification system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the tax policy knowledge base-based declaration logic verification system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The tax policy knowledge base-based declaration logic verification system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0112] For ease of explanation, only one processor is described in the tax policy knowledge base-based declaration logic verification system 100. However, it should be noted that the tax policy knowledge base-based declaration logic verification system 100 of this application may also include multiple processors. Therefore, the steps performed by one processor described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the tax policy knowledge base-based declaration logic verification system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0113] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned declaration logic verification method based on the tax policy knowledge base is implemented.

[0114] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for verifying declaration logic based on a tax policy knowledge base, characterized in that, The method includes: The system calls upon a pre-defined tax policy knowledge base, performs visualization processing on the policy entries within the base, extracts the elements related to the declaration in the policy description and the relationships between these elements, and generates a visualization set of policy elements. This visualization set of policy elements contains the corresponding information of policy entries and elements. Obtain the declaration data submitted by taxpayers and the associated declaration scenario descriptions, decompose the declaration data into basic field elements, analyze the correlation attributes between field elements to form a declaration element set, and bind the declaration element set with the declaration scenario descriptions to obtain a scenario-based declaration element set; The scenario-based application element set and the policy element representation set are imported into the element association mechanism. The appropriate policy is locked through scenario matching, the attribute comparison between field elements and policy elements is performed, and a mapping link is established to generate element mapping results. By filtering deviation links from the element mapping results, extracting the corresponding declaration field elements and policy elements, locating the root cause of the difference and tracking the chain effect of the difference on related fields, and recording the transmission path to obtain the declaration difference transmission information; Integrate the information on differences in the application process with the set of application elements for specific scenarios, combine policy elements to generate field adaptation suggestions and policy basis, organize the content according to the application order to form an application logic verification report, and push the report to the application terminal.

2. The declaration logic verification method based on a tax policy knowledge base according to claim 1, characterized in that, The process involves calling a pre-defined tax policy knowledge base, performing visualization processing on policy entries within the base, extracting elements related to the declaration from the policy descriptions and the relationships between these elements, and generating a set of visualized policy elements, including: Traverse all policy entries in the tax policy knowledge base, split each policy entry into natural language statements to obtain multiple independent policy statements, and add a corresponding policy entry number to each split policy statement. Semantic analysis is performed on policy statements to identify tax business keywords, declaration constraints, and applicable prerequisites contained in the statements, and these keywords, declaration constraints, and applicable prerequisites are used as initial elements. The initial elements are standardized by merging different initial elements that point to the same application requirements into the same standard element, and assigning a unique element code to each standard element. Analyze the association patterns of standard elements in the same policy statement, and combine tax policy semantic rules to identify and determine the preconditions, constraints, and execution order between each standard element, and generate element association links; The standard elements, element codes, element association links, and policy item numbers are linked together and categorized according to tax business types to form a structured set of concrete policy elements.

3. The declaration logic verification method based on a tax policy knowledge base according to claim 1, characterized in that, The process involves obtaining the taxpayer's submitted declaration data and associated declaration scenario descriptions, decomposing the declaration data into structured elements to obtain basic field elements, analyzing the correlation attributes between field elements to form a declaration element set, and binding the declaration element set with the declaration scenario descriptions to obtain a scenario-based declaration element set, including: Receive taxpayers' declaration data and declaration scenario descriptions submitted through the declaration terminal. The declaration scenario descriptions cover the industry to which the declaration business belongs, the time period in which the business occurred, and the characteristics of the economic behavior. The declared data is decomposed into multiple independent fields. The field name, data content, reporting scope and data source of each field are extracted as basic field elements. Analyze the association attributes of different basic field elements, identify field combinations with numerical derivation, conditional dependence, and hierarchical subordination based on the field data association characteristics, and record the association methods within the field combinations as associated field elements; Integrate basic field elements and related field elements by adding field identifiers to both basic field elements and related field elements to form a set of declaration elements that includes independent field attributes and related attributes; Extract scenario feature words from the application scenario description, establish a correspondence with the field identifiers in the application element set, and add the scenario feature words as attribute tags to the corresponding field elements to obtain a scenario-based application element set.

4. The declaration logic verification method based on a tax policy knowledge base according to claim 1, characterized in that, The process of importing the scenario-based declaration element set and the policy element representation set into the element association mechanism, locking in the appropriate policy through scenario matching, performing attribute comparison between field elements and policy elements, and establishing a mapping link to generate element mapping results includes: The policy index module of the element association mechanism is imported into the concrete set of policy elements. The retrieval index is built with standard elements as the core, and the index is associated with the corresponding element association links and policy item numbers. Extract scenario-specific vocabulary from the set of scenario-based application elements, match the scenario-specific vocabulary with the policy business type tags in the policy index module, and identify the policy scope that is suitable for the application scenario. Individual field elements are extracted from the set of scenario-based declaration elements, and field description text containing field name, data content and scenario attributes is generated; The field description text is compared with the standard elements within the scope of the adaptation policy. The overlap of attribute dimensions is compared to determine the standard elements with the highest degree of fit, and a mapping link between field elements and standard elements is established. Record the attribute matching points, associated element links, and policy entry numbers for each mapping link, and integrate all mapping links to form an element mapping result containing matching information.

5. The declaration logic verification method based on a tax policy knowledge base according to claim 1, characterized in that, The process of filtering deviation links from the element mapping results, extracting corresponding declaration field elements and policy elements, locating the root causes of differences and tracking the chain effects of differences on related fields, and sorting out the transmission path to obtain declaration difference transmission information includes: Traverse all mapping links in the element mapping results, filter out deviation links that do not meet the benchmark based on the preset attribute fit, and mark the declaration field elements corresponding to the deviation links. Extract the standard elements and element association links associated with the deviation links, extract the attribute requirements of policy elements on the declaration fields from the element association links, and determine the attribute requirements of policy elements on the declaration fields extracted from the element association links as policy benchmark attributes; Compare the actual attributes of the declared field elements with the policy benchmark attributes, record the specific manifestations of attribute differences dimension by dimension, and combine the root causes of the differences in the positioning of the element association links in the deviation link association. Based on the related field element information in the declaration element set, track the relationship attributes between the difference field and other fields, and analyze the impact of the difference on the attributes of the related field elements. Record the initial difference field, derived difference field, and associated attributes in the order of difference transmission, record the hierarchical relationship and degree of impact of difference transmission, and form the information on the transmission of differences in the declaration.

6. The declaration logic verification method based on a tax policy knowledge base according to claim 2, characterized in that, The analysis of the association patterns of standard elements within the same policy statement, combined with tax policy semantic rules, identifies and determines the preconditions, constraints, and execution order between each standard element, generating element association links, including: Position markers are set for standard elements in a single policy statement, recording the syntactic position and order of each standard element within the policy statement. By combining grammatical association rules, we can identify the grammatical associations between standard elements and distinguish between subject-verb associations, verb-object associations, and modifier associations. In line with the conventions of tax policy expression, we will strengthen the identification of special relationships between standard elements, specifically covering the relationship between preconditions and implementation content, the relationship between the scope of constraints and core requirements, and the relationship between process steps and their sequence. The relationships between standard elements are presented in the form of directed links. The link nodes are standard elements and element codes, and the connection between nodes is labeled with the association type and association description. Logical verification of the element association links is performed to correct grammatical conflicts or semantic contradictions in the associations, so that the element association links directly reflect the inherent logic of the policy statements.

7. The declaration logic verification method based on a tax policy knowledge base according to claim 3, characterized in that, The analysis examines the association attributes of different basic field elements. Based on the field data association characteristics, it identifies field combinations with numerical derivation, conditional dependency, and hierarchical subordinate relationships. The association methods within these field combinations are recorded as associated field elements, including: Select a basic field element from the declaration data as the target element, and retrieve other basic field elements in the declaration element set that have data correlation with the target element to form a candidate set of associations; Analyze the data content of the target element and each element in the associated candidate set, compare the numerical association characteristics of the target element and the associated elements to determine the source of the target element's values, and record the relevant association characteristics as numerical derived associations; Check whether the attributes of the target element are restricted by the attributes of other elements. When a certain element meets a specific attribute condition, the attribute of the target element must meet the corresponding requirements. Record the restriction relationship between the attribute condition of a certain element and the attribute of the target element as a conditional dependency association. Identify whether a feature attribute is a detailed description of the target feature attribute. When a feature attribute is a detailed description of the target feature attribute, the feature attribute belongs to the target feature attribute. Record the hierarchical relationship between the feature attribute and the target feature attribute as a hierarchical subordinate association. The numerical derivation association, conditional dependency association, and hierarchical subordinate association are integrated to annotate the field combinations and operation relationships corresponding to the numerical derivation association, the field combinations and restriction relationships corresponding to the conditional dependency association, and the field combinations and hierarchical relationships corresponding to the hierarchical subordinate association, forming a set of associated field elements.

8. The declaration logic verification method based on a tax policy knowledge base according to claim 4, characterized in that, The step of comparing the field description text with standard elements within the scope of the adaptation policy, determining the standard element with the highest degree of fit by comparing the overlap of attribute dimensions, and establishing a mapping link between field elements and standard elements includes: Extract the core attributes from the field description text, specifically covering the field data type, filling requirements, scenario adaptation features, and data format specifications, to form a set of field attributes; Extract the core attributes of each standard element included in the scope of the applicable policy, specifically covering policy constraints, applicable scenario characteristics, data standards, and compliance boundary definitions, to form a set of policy attributes; The set of field attributes is compared with the set of policy attributes dimension by dimension, and the number of dimensions that are completely, partially, and completely mismatched are counted. The number of matching dimensions determines the attribute matching degree. The standard element with the highest matching degree is selected as the matching object. If there are multiple standard elements with the same matching degree, further filtering is carried out by combining scene feature words. Establish the correspondence between target field elements and matching standard elements, associate the attribute matching details of target field elements and matching standard elements, and generate a mapping link containing bidirectional association information between target field elements and matching standard elements.

9. The declaration logic verification method based on a tax policy knowledge base according to claim 5, characterized in that, The process of recording the initial difference field, derived difference field, and associated attributes according to the order of difference transmission, recording the hierarchical relationship and degree of impact of difference transmission, forms the declaration difference transmission information, including: The initial difference field is used as the starting point of the transmission, and its field identifier, difference performance and difference root cause are recorded as the first level node of the transmission link. Based on the information of related field elements, find the related fields that are directly related to the initial difference field, analyze the derived differences caused by the initial difference in the related fields, record the performance of the derived differences and the related attributes, and use them as the second-level nodes of the transmission link; Using the derived difference field in the second-level node as the new starting point, continue to search for other fields that are related to the derived difference field, and repeat the derived difference analysis process until no new derived difference fields appear. A hierarchical structure diagram is used to present the difference transmission process. Each node at each level is labeled with field identifiers, difference types, related attributes, and degree of impact. Directional lines are used between nodes to indicate the direction of difference transmission. Organize the information in the hierarchical structure diagram, write the difference transmission description according to the transmission order, mark the causal relationship of the differences in each field, and form a complete information on the transmission of differences in the declaration.

10. A declaration logic verification system based on a tax policy knowledge base, characterized in that, The declaration logic verification system based on the tax policy knowledge base includes a processor and a memory, the memory and the processor are connected, the memory is used to store programs, instructions or code, and the processor is used to execute the programs, instructions or code in the memory to implement the declaration logic verification method based on the tax policy knowledge base as described in any one of claims 1-9.

Images