Recommendation method and device of salary scheme, electronic equipment and storage medium

Abstract

The invention provides a salary scheme recommendation method and apparatus, an electronic device and a storage medium. The method comprises the steps of obtaining enterprise data of a first enterprise input by a user in a configuration interface; traversing a pre-constructed knowledge graph according to the enterprise data of the first enterprise to obtain a first salary formula, a system and a constraint condition corresponding to the second enterprise, the similarity between the second enterprise and the first enterprise being greater than a threshold, the knowledge graph at least comprising an enterprise entity, a system entity, a salary formula entity and a constraint condition; inputting the first salary formula, the system and the constraint condition corresponding to the second enterprise into the intelligent engine to obtain a second salary formula; and the second salary formula recommended to the user is displayed in the configuration interface, so that manual salary scheme making is avoided, and the salary scheme making efficiency and making accuracy are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a salary scheme recommendation method and device, electronic equipment and storage medium. BACKGROUND

[0002] Salary scheme formulation is essential for the operation of each company. Under the traditional salary formulation method, relevant business personnel need to track the dynamic adjustment of social security base, individual tax progressive tax rate and other systems in the company's territory in a timely manner, and then formulate a reasonable salary scheme.

[0003] However, companies operating across regions need to adapt to hundreds of system provisions (for example, social security bases in different regions are different) when formulating a salary scheme. It is difficult for manual work to track the dynamic adjustment of social security base, individual tax progressive tax rate and other systems in the company's territory in a timely manner, which will result in low efficiency and errors in formulating a salary scheme. SUMMARY

[0004] Therefore, the embodiments of the present application provide a salary scheme recommendation method, device, electronic equipment and storage medium to solve the problems of low efficiency and errors in formulating a salary scheme by relying on manual work.

[0005] To achieve the above-mentioned purpose, the embodiments of the present application provide the following technical solutions:

[0006] The first aspect of the embodiments of the present application discloses a salary scheme recommendation method, which comprises:

[0007] Obtaining enterprise data of a first enterprise input by a user in a configuration interface;

[0008] Traversing a pre-constructed knowledge graph according to the enterprise data of the first enterprise to obtain a first salary formula, system and constraint condition corresponding to a second enterprise, the similarity between the second enterprise and the first enterprise being greater than a threshold value, the knowledge graph comprising at least enterprise entities, system entities, salary formula entities and constraint conditions, and the knowledge graph being constructed based on enterprise data and system data;

[0009] Inputting the first salary formula, system and constraint condition corresponding to the second enterprise into an intelligent engine to obtain a second salary formula;

[0010] Displaying the second salary formula recommended to the user in the configuration interface.

[0011] Preferably, traversing the pre-constructed knowledge graph according to the enterprise data of the first enterprise to obtain the first salary formula, system and constraint condition corresponding to the second enterprise comprises:

[0012] Convert the enterprise data of the first enterprise into embedded representation features;

[0013] Similarity is calculated by using the embedded representation features corresponding to the first enterprise and the embedded representation features corresponding to each enterprise entity in the pre-constructed knowledge graph.

[0014] The enterprise entities whose similarity to the first enterprise is greater than a threshold are selected to determine the second enterprise;

[0015] Extract the first salary formula, system, and constraints corresponding to the second enterprise from the knowledge graph.

[0016] Preferably, the first salary formula, system, and constraints corresponding to the second enterprise are input into the intelligent engine to obtain the second salary formula, including:

[0017] The first salary formula is parsed into formula units;

[0018] The formula unit, the system and constraints corresponding to the second enterprise are combined and input into the intelligent engine to configure the salary formula, thereby obtaining the second salary formula output by the intelligent engine.

[0019] Preferably, after displaying the second salary formula recommended to the user in the configuration interface, the method further includes:

[0020] Obtain the third salary formula configured by the user in the configuration interface;

[0021] The second salary formula and the third salary formula are compared to obtain formula difference information;

[0022] The formula difference information is displayed in the configuration interface.

[0023] Preferably, the second salary formula and the third salary formula are compared to obtain formula difference information, including:

[0024] The second salary formula and the third salary formula are respectively converted into abstract syntax trees;

[0025] By comparing the node differences between the abstract syntax tree corresponding to the second salary formula and the abstract syntax tree corresponding to the third salary formula, formula difference information is obtained.

[0026] Preferably, after displaying the second salary formula recommended to the user in the configuration interface, the method further includes:

[0027] Regularly obtain updated policy information and the fourth salary formula currently used by the first company;

[0028] The knowledge graph is updated using the updated institutional information;

[0029] Check whether the fourth salary formula is compatible with the updated system information;

[0030] If the fourth salary formula is not compatible with the updated policy information, the updated policy information is used to generate and output a modification suggestion for the fourth salary formula.

[0031] Preferably, after displaying the second salary formula recommended to the user in the configuration interface, the method further includes:

[0032] When it is detected that the user has modified any variable in the second salary formula, the fifth salary formula affected by the modified variable is identified based on the knowledge graph.

[0033] A description of the scope of influence is generated based on the fifth salary formula, and the modified description of the scope of influence is displayed in the configuration interface.

[0034] A second aspect of this invention discloses a compensation scheme recommendation device, the device comprising:

[0035] The acquisition module is used to acquire the enterprise data of the first enterprise entered by the user in the configuration interface;

[0036] The traversal module is used to traverse the pre-built knowledge graph based on the enterprise data of the first enterprise to obtain the first salary formula, system and constraints corresponding to the second enterprise. The similarity between the second enterprise and the first enterprise is greater than a threshold. The knowledge graph includes at least enterprise entities, system entities, salary formula entities and constraints. The knowledge graph is built based on enterprise data and system data.

[0037] The generation module is used to input the first salary formula, system and constraints corresponding to the second enterprise into the intelligent engine to obtain the second salary formula;

[0038] The recommendation module is used to display the second salary formula recommended to the user in the configuration interface.

[0039] A third aspect of the present invention discloses a computer device, comprising: a processor and a memory, the processor and the memory being connected via a bus; wherein, the processor is configured to call and execute a program stored in the memory; the memory is configured to store the program, the program being configured to implement a method for recommending a compensation scheme as disclosed in the first aspect of the present invention.

[0040] A fourth aspect of the present invention discloses a storage medium storing computer-executable instructions for executing the method for recommending a compensation scheme disclosed in the first aspect of the present invention.

[0041] Based on the above embodiments of the present invention, a method, apparatus, electronic device, and storage medium for recommending a salary scheme are provided. The method involves: acquiring enterprise data of a first enterprise input by a user in a configuration interface; traversing a pre-constructed knowledge graph based on the enterprise data of the first enterprise to obtain a first salary formula, policy, and constraints corresponding to a second enterprise, wherein the similarity between the second enterprise and the first enterprise is greater than a threshold, and the knowledge graph includes at least enterprise entities, policy entities, salary formula entities, and constraints; inputting the first salary formula, policy, and constraints corresponding to the second enterprise into an intelligent engine to obtain a second salary formula; and displaying the second salary formula recommended to the user in the configuration interface. This scheme, for a first enterprise requiring salary scheme configuration, queries a second enterprise similar to the first enterprise from the knowledge graph, inputs the first salary formula, policy, and constraints corresponding to the second enterprise into an intelligent engine to obtain a second salary formula recommended for use by the first enterprise, avoiding reliance on manual salary scheme formulation and improving the efficiency and accuracy of salary scheme formulation. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0043] Figure 1 A flowchart illustrating a method for recommending a compensation scheme as provided in an embodiment of the present invention;

[0044] Figure 2 A flowchart for comparing differences in salary formulas provided in an embodiment of the present invention;

[0045] Figure 3 A flowchart for generating correction suggestions provided in embodiments of the present invention;

[0046] Figure 4 This is an example diagram illustrating an application scenario provided in an embodiment of the present invention;

[0047] Figure 5 A flowchart for generating a description of the scope of influence of modifications provided in an embodiment of the present invention;

[0048] Figure 6 This is an example diagram of the salary formula configuration process provided in an embodiment of the present invention;

[0049] Figure 7 This is a structural block diagram of a salary scheme recommendation device provided in an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] In this application, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0052] To better understand the following content, some terms will be explained first.

[0053] Drag-and-drop configuration interface: a low-code interactive tool that allows users to generate salary formulas by combining visual components.

[0054] Cold start: The recommendation mechanism for salary formulas.

[0055] Dynamic compliance monitoring; real-time tracking of system changes (such as changes in social security base), comparing existing systems and salary formulas in the knowledge graph, automatically detecting conflicts between existing salary formulas and the changed systems, and generating correction suggestions.

[0056] Impact prediction of changes: By utilizing the dependency relationships of salary formulas in the knowledge graph, we can analyze the scope and magnitude of the impact of modified variables (such as changes to the overtime pay coefficient) on related salary formulas (such as formulas for calculating total wages and individual income tax).

[0057] Cross-modal recommendation: Combines policy (text), constraints (such as parameters in the salary formula), and enterprise characteristics (such as enterprise industry and enterprise size) to generate configuration suggestions for the salary formula.

[0058] Abstract Syntax Tree (AST): Parses salary formulas into a tree-like logical structure, supporting semantic difference analysis, dependency extraction, and automatic correction.

[0059] Implicit dependencies: Variable relationships not explicitly declared in the salary formula, equivalent to multi-hop neighbor nodes in a knowledge graph.

[0060] Link prediction: This method uses deep learning to predict possible relationships in a knowledge graph. It is typically used to fill in missing knowledge in the knowledge graph after entity relationship extraction, entity recognition, and entity alignment.

[0061] Research has revealed significant shortcomings in the rule engines used by traditional payroll management systems for creating payroll plans, particularly in areas such as payroll formula configuration efficiency, cold start support, compliance tracking, and change impact control. Specifically, the drag-and-drop configuration method used by business personnel often results in payroll formula creation taking several hours, and the lack of historical data makes it difficult to quickly generate reasonable payroll formulas. Furthermore, manual monitoring of dynamic adjustments to local social security contribution bases and progressive individual income tax rates can easily lead to payroll calculation errors. The implicit dependencies between payroll formulas mean that modifications to a single formula can trigger large-scale calculation errors.

[0062] The reasons for this situation are as follows: First, companies operating across regions need to adapt to hundreds of policy clauses when formulating salary plans (for example, social security bases vary from region to region), and manually tracking and comparing policy changes is extremely inefficient; Second, frequent policy changes lead to a surge in the maintenance costs of the salary management system. For example, social security policies and minimum wages in different regions are adjusted dozens of times, and manually tracking policy changes takes 3-5 days; Third, the implicit dependencies between complex salary formulas create a technical black hole. When a company's salary plan involves hundreds or thousands of calculation items, modifying a variable in a certain salary formula will trigger a chain reaction of changes in multiple related salary formulas.

[0063] To address the aforementioned issues, this invention proposes a method, apparatus, electronic device, and storage medium for recommending salary schemes. For a first enterprise requiring salary scheme configuration, a second enterprise similar to the first enterprise is queried from a knowledge graph. The first salary formula, system, and constraints corresponding to the second enterprise are input into an intelligent engine to obtain a second salary formula recommended for use by the first enterprise. This avoids reliance on manual salary scheme formulation and improves the efficiency and accuracy of salary scheme formulation.

[0064] See Figure 1 The flowchart illustrates a method for recommending a compensation scheme according to an embodiment of the present invention. The method includes:

[0065] Step S101: Obtain the enterprise data of the first enterprise entered by the user in the configuration interface.

[0066] In the specific implementation step S101, the enterprise data of the first enterprise entered by the user in the drag-and-drop configuration interface is obtained. The first enterprise is the enterprise that needs to configure the salary formula. The enterprise data includes, but is not limited to: enterprise tag, salary type (such as bonus salary), enterprise size, industry type, local system parameters, employee position, etc.

[0067] Step S102: Based on the enterprise data of the first enterprise, traverse the pre-constructed knowledge graph to obtain the first salary formula, system and constraints corresponding to the second enterprise.

[0068] It should be noted that the similarity between the second company and the first company is greater than the threshold. The knowledge graph includes at least enterprise entities, institutional entities, salary formula entities, and constraints. The knowledge graph is constructed based on the enterprise data of each company and the institutional data adopted by each company.

[0069] In some embodiments, the specific way to construct a knowledge graph is as follows: for the structured information (such as industry type, employee position, local system parameters) in the enterprise data of each enterprise, the structured information is cleaned by using a standardized coding method, and then preprocessed by deduplication, standardization and missing value filling, thereby forming a knowledge graph prototype.

[0070] For unstructured information and unstructured institutional data in the enterprise data of various enterprises, data cleaning, preprocessing, named entity recognition (NER), relation extraction, entity alignment and link prediction are performed on the unstructured information and unstructured institutional data. The knowledge obtained after processing is integrated into the knowledge graph prototype, thereby obtaining a knowledge graph (RDF triple type knowledge graph) including enterprise entities, institutional entities, salary formula entities and constraints.

[0071] The constraints can be variable constraints in the salary formula. Named entity recognition and relation extraction enable deep semantic analysis, breaking down the information into relations and entities. Entity alignment uses a pre-trained word embedding model to capture the semantic similarity between entities with the same meaning, removing entities with duplicate meanings and concatenating related entities and relations in the form of "(head entity, relation, tail entity)" to form a knowledge graph. Link prediction uses a knowledge graph completion model to reason on existing "entity-relation pairs," completing missing relation vectors.

[0072] In the specific implementation step S102, the enterprise data of the first enterprise is converted into embedding representation features. Specifically, the enterprise data of the first enterprise is converted into low-dimensional embedding representation features through a pre-trained word embedding model, thereby obtaining the embedding representation features corresponding to the first enterprise.

[0073] The similarity between the first enterprise and each enterprise entity in the knowledge graph is calculated by using the embedded representation features corresponding to the first enterprise and the embedded representation features corresponding to each enterprise entity in the pre-constructed knowledge graph.

[0074] Specifically, the embedding representation features corresponding to all enterprise entities are extracted from the knowledge graph. The cosine similarity between the embedding representation features corresponding to the first enterprise and the embedding representation features corresponding to each enterprise entity is calculated to obtain the similarity between the first enterprise and each enterprise entity in the knowledge graph.

[0075] The second enterprise is determined by filtering out enterprise entities whose similarity to the first enterprise is greater than a threshold. In other words, the second enterprise is the enterprise corresponding to an enterprise entity whose similarity to the first enterprise is greater than a threshold. That is, the second enterprise is an enterprise similar to the first enterprise.

[0076] Extract the first salary formula, system, and constraints corresponding to the second company from the knowledge graph.

[0077] Step S103: Input the first salary formula, system and constraints corresponding to the second enterprise into the intelligent engine to obtain the second salary formula.

[0078] In the specific implementation step S103, the first salary formula is parsed into formula units. Specifically, the first salary formula is parsed into formula units through the abstract syntax tree parser.

[0079] The formula unit, the system and constraints corresponding to the second enterprise are combined and input into the intelligent engine (such as artificial intelligence) to configure the salary formula, and the second salary formula output by the intelligent engine is obtained.

[0080] Specifically, the formula unit obtained from the parsing of the first salary formula is equivalent to a salary template. The formula unit, the system and constraints corresponding to the second enterprise are merged and input into the intelligent engine to obtain the second salary formula output by the intelligent engine.

[0081] For example: The formula unit obtained by parsing the first salary formula is "sales volume × commission rate", the system corresponding to the first salary formula is "the local minimum wage standard is 2420 yuan / month", and the constraint condition corresponding to the first salary formula is "the commission rate must meet the total salary (basic salary + bonus salary) ≥ 2420 yuan". Then the second salary formula output by the intelligent engine is: bonus salary = sales volume × commission rate (total salary ≥ 2420 yuan).

[0082] Step S104: Display the second salary formula recommended to the user in the configuration interface.

[0083] In the specific implementation step S104, after obtaining the second salary formula, the second salary formula is displayed in the configuration interface to recommend the second salary formula to the user.

[0084] Understandably, when displaying the second salary formula in the configuration interface, the system on which the second salary formula is based can also be displayed.

[0085] For example, the second salary formula is: Bonus salary = Sales revenue × Commission rate (total salary must be ≥ 2420 yuan). The system upon which the second salary formula is based is "Article 12 of the 'Enterprise Wage Payment Regulations of XX Province'".

[0086] In this embodiment of the invention, for a first enterprise that needs to configure a salary scheme, a second enterprise similar to the first enterprise is queried from the knowledge graph. The first salary formula, system and constraints corresponding to the second enterprise are input into the intelligent engine to obtain a second salary formula recommended for use by the first enterprise. This avoids relying on manual formulation of the salary scheme and improves the efficiency and accuracy of the salary scheme formulation.

[0087] Understandably, after the second salary formula recommended to the user is displayed in the configuration interface, the user can use the second salary formula, or configure a third salary formula based on (or referencing) the second salary formula. At this time, the difference between the second and third salary formulas can be displayed in the configuration interface.

[0088] In some embodiments, see Figure 2 The flowchart illustrating the comparison of salary formula differences provided in an embodiment of the present invention is shown. Figure 2 Includes the following steps:

[0089] Step S201: Obtain the third salary formula configured by the user in the configuration interface.

[0090] Step S202: Compare the differences between the second salary formula and the third salary formula to obtain formula difference information.

[0091] In the specific implementation step S202, the second salary formula and the third salary formula are converted into abstract syntax trees (ASTs) respectively. The node differences between the ASTs corresponding to the second salary formula and the ASTs corresponding to the third salary formula are compared to obtain formula difference information.

[0092] In other words, the difference between the user-configured third salary formula and the recommended second salary formula is used to obtain information on the formula difference between the third salary formula and the second salary formula.

[0093] Step S203: Display formula difference information in the configuration interface.

[0094] In the specific implementation step S203, the formula difference information between the third salary formula and the second salary formula is displayed in the configuration interface, thereby realizing the differential recommendation of the salary formula.

[0095] It should be noted that the fourth salary formula currently used by the first company may need to be revised accordingly as the system changes.

[0096] In some embodiments, see Figure 3 The flowchart illustrating the generation of correction suggestions provided in an embodiment of the present invention is shown. Figure 3 Includes the following steps:

[0097] Step S301: Periodically obtain updated policy information and obtain the fourth salary formula currently used by the first company.

[0098] In the specific implementation step S301, the changes in the system are monitored, the updated system information is obtained periodically, and the fourth salary formula currently used by the first company is obtained.

[0099] Step S302: Update the knowledge graph using the updated institutional information.

[0100] In the specific implementation step S302, the updated institutional information is used to update the knowledge graph, thereby obtaining the latest knowledge graph.

[0101] Specifically, the updated institutional information is processed through named entity recognition, relation extraction, entity alignment, and link prediction, and then converted into structured data stored within the knowledge graph, thereby completing the knowledge graph update.

[0102] In practical applications, after updating the knowledge graph, if it is necessary to generate a second recommendation formula to recommend to the user, the updated knowledge graph is used to generate the second recommendation formula to recommend to the user.

[0103] Step S303: Check whether the fourth salary formula is compatible with the updated system information.

[0104] In the specific implementation step S303, it is checked whether the fourth salary formula is adapted to the updated system information, that is, the policy compliance check is performed on the fourth salary formula.

[0105] If the fourth salary formula does not match the updated system information, proceed to step S304.

[0106] Step S304: If the fourth salary formula is not compatible with the updated system information, use the updated system information to generate and output a modification suggestion for the fourth salary formula.

[0107] In the specific implementation step S304, if the fourth salary formula is not compatible with the updated system information, the updated system information is used to generate and output a correction suggestion for the fourth salary formula. The user corrects the fourth salary formula according to the correction suggestion so that the corrected fourth salary formula is compatible with the updated system information.

[0108] It should be noted that the proposed amendments include not only information on how to modify the fourth salary formula, but also details of the changes in the updated policy information and its impact on the fourth salary formula.

[0109] In practical applications, in conjunction with the above embodiments of the present invention Figures 1 to 3 Regarding the content, when creating a new salary formula, this solution has multiple application scenarios, such as... Figure 4 As shown in the example diagram of the application scenarios, this solution has at least three application scenarios: cold start, differential recommendation, and policy compliance check.

[0110] In the cold start application scenario, the system obtains the enterprise data of the first enterprise entered by the user in the configuration interface, traverses the knowledge graph based on the enterprise data of the first enterprise to obtain the first salary formula, system and constraints corresponding to the second enterprise, and inputs the first salary formula, system and constraints corresponding to the second enterprise into the intelligent engine to obtain the second salary formula recommended to the user.

[0111] In the differential recommendation scenario, after obtaining the second salary formula recommended to the user, the difference between the user-configured third salary formula and the recommended second salary formula is used to obtain the formula difference information between the third salary formula and the second salary formula, and this formula difference information is displayed to the user.

[0112] In the policy compliance inspection scenario, updated policy information is acquired periodically and converted into structured data stored in the knowledge graph to complete the knowledge graph update. If the fourth salary formula is not compatible with the updated policy information, the updated policy information is used to generate and output modification suggestions for the fourth salary formula.

[0113] In practical applications, when recommending a second salary formula to users, users may need to modify some variables in the second salary formula. When users modify any variable in the second salary formula, it is necessary to remind users which salary formulas will be affected by the "modified variable", that is, to remind users of the scope of the impact of the modification of the "modified variable".

[0114] In some embodiments, see Figure 5 The flowchart illustrating the generation of a description of the scope of influence of modifications provided in an embodiment of the present invention is shown. Figure 5 Includes the following steps:

[0115] Step S501: When it is detected that the user has modified any variable in the second salary formula, the fifth salary formula affected by the modified variable is identified based on the knowledge graph.

[0116] It should be noted that the drag-and-drop configuration interface provides a visual component library (including mathematical operators, conditional judgment blocks, etc.), which allows users to freely build and modify salary formulas.

[0117] In the specific implementation step S501, after recommending the second salary formula to the user, if it is detected that the user has modified any variable in the second salary formula, the fifth salary formula affected by the modified variable is identified based on the knowledge graph. The fifth salary formula is the salary formula affected by the modified variable in the second salary formula.

[0118] Step S502: Generate an impact range description based on the fifth salary formula, and display the modified impact range description in the configuration interface.

[0119] In the specific implementation step S502, an impact range description is generated based on the fifth salary formula. This impact range description at least represents the following information: which salary formulas will be affected by the modified variables in the second salary formula, and the magnitude of change of some variables in these affected salary formulas.

[0120] For example: Suppose the second salary formula is the bonus calculation formula. When a user modifies the bonus calculation formula, the scope of impact description indicates which salary formulas will be affected by the modified variables in the bonus calculation formula, as well as the magnitude of the change in individual income tax.

[0121] Understandably, if the company detects that a user has modified any variable in the fourth salary formula currently in use, the company can also identify the affected fifth salary formula and generate a corresponding description of the scope of impact.

[0122] In practical applications, one application scenario of this solution is as follows: Figure 6 The example diagram showing the configuration process for the salary formula is illustrated below. Figure 6 Includes the following steps:

[0123] Step S601: Create a new salary scheme.

[0124] In the specific implementation step S601, the user creates a new salary plan through a drag-and-drop configuration interface.

[0125] For example, when creating a "overtime pay calculation rule" compensation plan, users can drag and drop the "IF" condition block in the configuration interface to set: "Working hours > 8 hours" triggers the calculation of 1.5 times the hourly wage.

[0126] Step S602: Intelligent recommendation of the second salary formula.

[0127] In the specific implementation step S602, based on the enterprise data entered by the user in the configuration interface, a second salary formula that meets the user's needs is intelligently recommended to the user.

[0128] Step S603: Edit the second salary formula.

[0129] In the specific implementation step S603, the user can modify each variable in the second salary formula.

[0130] Step S604: Display the description of the scope of influence.

[0131] In the specific implementation step S604, when the user modifies any variable in the second salary formula, the fifth salary formula affected by the modified variable is identified based on the knowledge graph, and a description of the scope of influence is generated and dynamically displayed in the sidebar of the configuration interface.

[0132] This sidebar can also display the systems associated with the second salary formula, as well as the historical changes of those systems.

[0133] above Figures 1 to 6 This is an explanation of the plan. Overall, the plan can be divided into the following parts: formula recommendation, difference recommendation, and dynamic update of system and formula.

[0134] The formula recommendation section involves cleaning the structured information in the enterprise data of each company, followed by preprocessing such as deduplication, standardization, and missing value imputation to form a knowledge graph prototype. Unstructured information and unstructured institutional data from the enterprise data of each company undergo data cleaning, preprocessing, named entity recognition (NER), relation extraction, entity alignment, and link prediction. The resulting knowledge is then integrated into the knowledge graph prototype, resulting in a knowledge graph including enterprise entities, institutional entities, salary formula entities, and constraints.

[0135] It should be noted that named entity recognition can use the "RoBERTa" model. The "RoBERTa" model is trained using a general dataset and fine-tuned on annotated institutional text corpus to accurately annotate key entities such as "high temperature allowance" and "performance coefficient".

[0136] Relation extraction uses the "BiLSTM" model to extract relationships between entities from the text, such as the relationship between entities "high temperature allowance belongs to local subsidies".

[0137] Entity alignment and link prediction are trained on the dataset using the "GraphSAGE" graph neural network to align similar entities in the knowledge graph. For example, aligning "XX City" in the text to the node "2025 XX City" in the knowledge graph.

[0138] One example of the training parameters for “RoBERTa”, “BiLSTM”, and “GraphSAGE” is shown in Table 1.

[0139] Table 1:

[0140]

[0141] After constructing the enterprise-system knowledge graph, salary formulas can be recommended based on the knowledge graph. During the salary formula recommendation process, the enterprise data of the first enterprise input by the user is received through a natural language interface. This enterprise data includes: enterprise tags, salary type, enterprise size, industry type, local system parameters, employee positions, etc.

[0142] For example, it can receive data such as "bonus salary", "industry" and "region" input by users.

[0143] The knowledge graph's query interface is used to retrieve data related to the user's input. For example, data such as "retail industry," "minimum wage standard of 2420 yuan / month," "commission rate must meet total salary ≥ 2420 yuan," and "bonus salary = sales revenue × commission rate" are retrieved. Here, "bonus salary = sales revenue × commission rate" is the first salary formula corresponding to the second company. "commission rate must meet total salary ≥ 2420 yuan" is the constraint condition.

[0144] By leveraging the reasoning capabilities of the intelligent engine and injecting the retrieved data such as "retail industry", "minimum wage standard of 2420 yuan / month", "commission ratio must meet the requirement of total salary ≥ 2420 yuan", and "bonus salary = sales volume × commission ratio", a second salary formula recommended to users is finally generated.

[0145] During the training process, the intelligent engine focuses on improving its ability to understand and associate knowledge in multi-turn dialogues. For example, it enhances the recognition of intent to "combine industry attributes with local regulations" by fine-tuning dialogue data, while adjusting the weight of the attention mechanism to strengthen the ability to extract key constraints from long texts (such as institutional documents).

[0146] In practical applications, one scenario is as follows: When a user inputs "bonus salary" and selects "retail industry" and "XX city", the intelligent engine first analyzes the user's intent and calls the knowledge graph query interface to extract "bonus salary = sales revenue × commission rate" and "minimum wage standard of 2420 yuan / month". Then, based on "the commission rate must meet the total salary ≥ 2420 yuan", it generates a second salary formula: "bonus salary = sales revenue × commission rate (total salary ≥ 2420 yuan)". When recommending the second salary formula to the user, it also displays "Article 12 of the 'XX Province Enterprise Wage Payment Regulations'", the system on which the second salary formula is based.

[0147] The intelligent recommendation of salary formulas achieves a three-way matching of institutional compliance, enterprise personalized needs (industry attribute adaptation), and constraint injection, eliminating the need for manual collection of institutional documents or collation of historical salary formulas, thus improving the efficiency of salary formula generation.

[0148] Differential Recommendation: The differential recommendation mechanism analyzes the semantic differences between different salary formulas to accurately identify "formula difference information" and "the scope of influence of modifications". Its core part relies on the abstract syntax tree parsing and the reasoning ability of the intelligent engine to achieve high efficiency and accuracy in salary formula analysis.

[0149] In addition, the versions of each salary formula can be managed by hash values, and backtracking and comparison of each salary formula can be supported.

[0150] For the second and third salary formulas that need to be compared, the second and third salary formulas are converted into abstract syntax trees (nodes include operators, variables, and constants).

[0151] For example: Suppose the second salary formula is "total salary = basic salary + performance coefficient × 21.0", the abstract syntax tree obtained by converting the second salary formula is as follows: root node "+", left child node "basic salary", right child node "×" (its left child node is "performance coefficient", and its right child node is the constant "21.0").

[0152] The tree edit distance algorithm is used to calculate the node differences between the "abstract syntax tree corresponding to the second salary formula" and the "abstract syntax tree corresponding to the third salary formula". Modified nodes are identified (e.g., the constant node "21.0" is modified to "21.2") and the modification type (e.g., the value is modified), thereby obtaining the formula difference information between the second salary formula and the third salary formula.

[0153] Understandably, when identifying the scope of influence of modifications, a directed acyclic graph (DAG) can be used. Based on the knowledge graph, a DAG is constructed where the nodes are salary formulas or variables of salary formulas, and the edges are the computational dependencies of the salary formulas (e.g., "performance coefficient" affects "total salary", "total salary" affects "bonus cap"). The hierarchical relationships between nodes are stored through an adjacency list.

[0154] In practical applications, one possible scenario is that a user modifies the second salary formula (equivalent to the old version) to obtain the third salary formula (equivalent to the new version). After the user submits the second and third salary formulas, the intelligent engine analyzes the user's intent (e.g., "adjust the performance coefficient and check the scope of impact"), compares the node differences between the abstract syntax trees corresponding to the second and third salary formulas, confirms that the modification type of the second salary formula is a numerical adjustment, and then traverses the reverse path of the DAG based on the business rules described in natural language (e.g., "the bonus cap is positively correlated with the total salary") (e.g., "performance coefficient" depends on "total salary", and "total salary" depends on "bonus cap") to identify all the affected fifth salary formulas. The target value of the third salary formula is calculated based on the business rules (e.g., bonus cap = total salary × 1.2) after adjusting the second salary formula (e.g., "original total salary × 21.0 × 1.2" is adjusted to "new total salary × 21.2 × 1.2"). The bonus cap in the second salary formula is 5,000 yuan, and the bonus cap in the third salary formula is 6,000 yuan, that is, the bonus cap is adjusted from 5,000 yuan to 6,000 yuan.

[0155] It should be noted that during the training process of the intelligent engine, the focus of optimization is to enhance the ability to understand the context of business rules. For example, by fine-tuning the dialogue data of "scenarios involving salary formula adjustment" ("After adjusting parameter A, which parameters B need to be modified simultaneously"), the accuracy of reasoning about implicit dependencies can be improved.

[0156] Specifically, when a user adjusts the performance coefficient from 21.0 to 21.2, the system first analyzes the user's intent and generates formula difference information between the second and third salary formulas (performance coefficient adjusted from 21.0 to 21.2). After confirming the modification type, based on the DAG calculation dependencies (e.g., "performance coefficient" affects "total salary", "total salary" affects "bonus cap"), the system identifies the affected nodes (e.g., "total salary" and "bonus cap"), calculates the adjusted bonus cap according to business rules, and generates a natural language description for the adjusted bonus cap. For example, the generated natural language description is: "Due to a performance increase of 0.2, the total salary is expected to increase. It is recommended to increase the bonus cap from 5000 yuan to 6000 yuan to match the new calculation logic."

[0157] The mechanism provided by the differential recommendation component avoids the calculation of the full salary formula, improves the efficiency of salary formula adjustment, reduces the error rate, and thus improves the accuracy of salary formula configuration.

[0158] The system-formula dynamic update section parses the system text in real time, updates the knowledge graph and links to the salary formula library (a database containing salary formulas), and achieves automatic matching of salary formulas with the latest system.

[0159] The parsing of the policy text is achieved using a BERT-CRF entity extraction model and a relation extraction model. The BERT-CRF entity extraction model is trained on the labeled policy text, which includes: allowance type, amount, and binding object. Relationships include policy-effective date and policy-applicable enterprise type. Text features are extracted using pre-trained BERT, and entity labels are predicted by the CRF layer. The loss function is cross-entropy, the learning rate is 3e-5, the batch size is 32, and the training run is 10 epochs. This model can accurately extract entities such as "high-temperature allowance," "2300 yuan / month," and "outdoor work employees" from the policy "high-temperature allowance 2300 yuan / month."

[0160] The relation extraction model is based on SpanBERT, training a multi-classification head on the same corpus to identify relationships between entity pairs (e.g., "high-temperature allowance" belongs to "local subsidy"). The loss function is multi-class cross-entropy, and the model is trained for 8 epochs. Furthermore, a snapshot of the salary formula library and compliance verification are generated through hash value comparison (integrating the Drools rule engine, loading legal clauses, and checking formula variables in real time). The model automatically triggers parameter verification and salary formula correction processes, and generates natural language reminders and correction suggestions through an intelligent engine.

[0161] In practical applications, one scenario is as follows: After the updated policy information of "high-temperature allowance of 2300 yuan / month" is released, a scheduled task retrieves the updated policy information, calls the BERT-CRF entity extraction model to extract the new entity "high-temperature allowance (2300 yuan / month)" from the updated policy information, and associates it with the "local subsidy" type. Then, the new entity "high-temperature allowance (2300 yuan / month)" is written into the knowledge graph, and a relationship is established with the "summer total salary" salary formula. The semantics of the updated policy information are parsed (e.g., this policy is effective from August 2025 and applies to outdoor employees), and the associated fourth salary formula in the knowledge graph (e.g., summer total salary = basic salary + high-temperature allowance) is traversed. The formula parameters of the current fourth salary formula (e.g., the high-temperature allowance value is 1500 yuan / month) are checked to see if they match the updated policy information. If the formula parameters of the fourth salary formula conflict with the updated policy information, a reminder message and correction suggestions for the fourth salary formula are generated and output.

[0162] The reminder message states: A new system has been detected requiring a high-temperature allowance of ≥2300 yuan / month. The current formula “Total Summer Wage = Basic Wage + High-Temperature Allowance (1500 yuan / month)” needs to be adjusted.

[0163] The revised suggestion is to set the high-temperature allowance to max(1500, 2300) yuan / month.

[0164] It is worth noting that a real-time semantic check function is also embedded in the process of editing the salary formula. When the configuration information entered by the user for the salary formula does not conform to the system information, an early warning is triggered and a warning message is displayed in the form of a pop-up window, while providing correction suggestions.

[0165] For example, when a user enters "male-to-female ratio coefficient × 1.2", an alert is triggered and a pop-up window is generated according to the relevant system information, indicating that the current configuration information does not comply with the legal basis, and providing suggestions for correction.

[0166] The mechanism provided by the dynamic update section of the system-formula system can transform the compliance verification of the system from the "post-event remediation" of traditional manual verification to "pre-event prevention". Real-time semantic understanding and natural language interaction capabilities can improve the efficiency of compliance management, and the salary scheme always complies with the latest regulatory requirements.

[0167] In summary, the proposed salary scheme recommendation method, by constructing a knowledge graph and utilizing artificial intelligence technology for salary formula generation, comparison, and error detection, eliminates the need for users to frequently monitor policy changes offline, significantly shortening compliance response time and improving cold start configuration efficiency. Furthermore, it can accurately identify the impact of salary formula modifications, linking policy information within the knowledge graph to provide salary formula configuration verification functionality. When configuring salary formulas, it uses salary formulas from other companies as a reference, providing users with multifaceted data support.

[0168] Corresponding to the method for recommending a compensation scheme provided in an embodiment of the present invention, see [link to relevant documentation]. Figure 7 The present invention also provides a structural block diagram of a salary scheme recommendation device, which includes: an acquisition module 701, a traversal module 702, a generation module 703, and a recommendation module 704.

[0169] The acquisition module 701 is used to acquire the enterprise data of the first enterprise entered by the user in the configuration interface.

[0170] The traversal module 702 is used to traverse the pre-built knowledge graph based on the enterprise data of the first enterprise to obtain the first salary formula, system and constraints corresponding to the second enterprise. The similarity between the second enterprise and the first enterprise is greater than a threshold. The knowledge graph includes at least enterprise entities, system entities, salary formula entities and constraints. The knowledge graph is built based on enterprise data and system data.

[0171] The generation module 703 is used to input the first salary formula, system and constraints corresponding to the second enterprise into the intelligent engine to obtain the second salary formula.

[0172] In specific implementation, the generation module 703 is used to: parse the first salary formula into formula units; merge the formula units, the system and constraints corresponding to the second enterprise, and input them into the intelligent engine to configure the salary formula, thereby obtaining the second salary formula output by the intelligent engine.

[0173] The recommendation module 704 is used to display the second salary formula recommended to users in the configuration interface.

[0174] In this embodiment of the invention, for a first enterprise that needs to configure a salary scheme, a second enterprise similar to the first enterprise is queried from the knowledge graph. The first salary formula, system and constraints corresponding to the second enterprise are input into the intelligent engine to obtain a second salary formula recommended for use by the first enterprise. This avoids relying on manual formulation of the salary scheme and improves the efficiency and accuracy of the salary scheme formulation.

[0175] Preferred, combined Figure 7 The content shown indicates that the traversal module 702 includes a transformation submodule, a calculation submodule, a filtering submodule, and an extraction submodule; the execution principle of each submodule is as follows:

[0176] The transformation submodule is used to transform the enterprise data of the first enterprise into embedded representation features.

[0177] The calculation submodule is used to perform similarity calculations by utilizing the embedded representation features corresponding to the first enterprise and the embedded representation features corresponding to each enterprise entity in the pre-built knowledge graph.

[0178] The filtering submodule is used to filter out enterprise entities that have a similarity greater than a threshold with the first enterprise to determine the second enterprise.

[0179] The extraction submodule is used to extract the first salary formula, system, and constraints corresponding to the second company from the knowledge graph.

[0180] Preferred, combined Figure 7 The recommended device, as shown, also includes:

[0181] The difference comparison module is used to obtain the third salary formula configured by the user in the configuration interface; compare the second salary formula and the third salary formula to obtain formula difference information; and display the formula difference information in the configuration interface.

[0182] The process by which the difference comparison module obtains formula difference information is as follows: the second salary formula and the third salary formula are converted into abstract syntax trees respectively; the node differences between the abstract syntax trees corresponding to the second salary formula and the abstract syntax trees corresponding to the third salary formula are compared to obtain formula difference information.

[0183] Preferred, combined Figure 7 The recommended device, as shown, also includes:

[0184] The update module is used to periodically obtain updated policy information and the fourth salary formula currently used by the first company; and to update the knowledge graph using the updated policy information.

[0185] The correction module is used to detect whether the fourth salary formula is compatible with the updated system information; if the fourth salary formula is not compatible with the updated system information, it uses the updated system information to generate and output correction suggestions for the fourth salary formula.

[0186] Preferred, combined Figure 7 The recommended device, as shown, also includes:

[0187] The impact identification module is used to identify the fifth salary formula affected by the modified variable when it detects that the user has modified any variable in the second salary formula; it generates an impact range description based on the fifth salary formula and displays the impact range description in the configuration interface.

[0188] Preferably, the present invention also provides a computer device, including: a processor and a memory, the processor and the memory being connected via a bus; wherein, the processor is used to call and execute a program stored in the memory; the memory is used to store the program, the program being used to implement the method for recommending a compensation scheme provided in the above method embodiments.

[0189] Preferably, the present invention also provides a storage medium storing computer-executable instructions for executing the method of recommending a compensation scheme provided in the above-described method embodiments.

[0190] In summary, the embodiments of the present invention provide a method, apparatus, electronic device, and storage medium for recommending salary schemes. For a first enterprise that needs to configure a salary scheme, a second enterprise similar to the first enterprise is queried from a knowledge graph. The first salary formula, system, and constraints corresponding to the second enterprise are input into an intelligent engine to obtain a second salary formula recommended for use by the first enterprise. This avoids relying on manual salary scheme formulation and improves the efficiency and accuracy of salary scheme formulation.

[0191] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the descriptions in the method embodiments. The systems and system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0192] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0193] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for recommending a compensation scheme, characterized in that, The method comprises: obtaining enterprise data of a first enterprise input by a user in a configuration interface; traversing a pre-constructed knowledge graph according to the enterprise data of the first enterprise to obtain a first salary formula, a system and a constraint condition corresponding to a second enterprise, the similarity between the second enterprise and the first enterprise being greater than a threshold, the knowledge graph comprising at least enterprise entities, system entities, salary formula entities and constraint conditions, the knowledge graph being constructed based on enterprise data and system data; inputting the first salary formula, the system and the constraint condition corresponding to the second enterprise into an intelligent engine to obtain a second salary formula; displaying the second salary formula recommended to the user in the configuration interface.

2. The method of claim 1, wherein, The traversing of the pre-constructed knowledge graph according to the enterprise data of the first enterprise to obtain the first salary formula, the system and the constraint condition corresponding to the second enterprise comprises: converting the enterprise data of the first enterprise into embedded representation features; performing similarity calculation on the embedded representation features corresponding to the first enterprise and the embedded representation features corresponding to each enterprise entity in the pre-constructed knowledge graph; screening the enterprise entities similar to the first enterprise to determine the second enterprise, the similarity being greater than a threshold; extracting the first salary formula, the system and the constraint condition corresponding to the second enterprise from the knowledge graph.

3. The method of claim 1, wherein, The inputting of the first salary formula, the system and the constraint condition corresponding to the second enterprise into the intelligent engine to obtain the second salary formula comprises: parsing the first salary formula into formula units; merging the formula units, the system and the constraint condition corresponding to the second enterprise to input into the intelligent engine to configure the salary formula, and obtaining the second salary formula output by the intelligent engine.

4. The method of claim 1, wherein, After the displaying of the second salary formula recommended to the user in the configuration interface, the method further comprises: obtaining a third salary formula configured by the user in the configuration interface; performing difference comparison on the second salary formula and the third salary formula to obtain formula difference information; displaying the formula difference information in the configuration interface.

5. The method of claim 4, wherein, The performing of the difference comparison on the second salary formula and the third salary formula to obtain the formula difference information comprises: respectively converting the second salary formula and the third salary formula into abstract syntax trees; comparing the node difference between the abstract syntax tree corresponding to the second salary formula and the abstract syntax tree corresponding to the third salary formula to obtain the formula difference information.

6. The method of claim 1, wherein, After the displaying of the second salary formula recommended to the user in the configuration interface, the method further comprises: timely obtaining updated system information, and obtaining a fourth salary formula currently used by the first enterprise; updating the knowledge graph by using the updated system information; detecting whether the fourth salary formula is adapted to the updated system information; if the fourth salary formula is not adapted to the updated system information, generating and outputting a correction suggestion for the fourth salary formula by using the updated system information.

7. The method of claim 1, wherein, After the displaying of the second salary formula recommended to the user in the configuration interface, the method further comprises: When it is detected that the user modifies any variable in the second salary formula, a fifth salary formula influenced by the modified variable is identified according to the knowledge graph; An influence range description is generated according to the fifth salary formula, and the influence range description is displayed in the configuration interface.

8. A device for recommending a compensation package, characterized in that, The device comprises: An acquisition module is configured to acquire enterprise data of a first enterprise input by a user in a configuration interface; A traversal module is configured to traverse a pre-constructed knowledge graph according to the enterprise data of the first enterprise to obtain a first salary formula, a system, and a constraint condition corresponding to a second enterprise, the similarity between the first enterprise and the second enterprise being greater than a threshold, the knowledge graph comprising at least an enterprise entity, a system entity, a salary formula entity, and a constraint condition, the knowledge graph being constructed based on enterprise data and system data; A generation module is configured to input the first salary formula, the system, and the constraint condition corresponding to the second enterprise into an intelligent engine to obtain a second salary formula; A recommendation module is configured to display the second salary formula recommended to the user in the configuration interface.

9. A computer device, comprising: comprise: A processor and a memory connected through a bus; wherein the processor is configured to call and execute a program stored in the memory; The memory is configured to store a program, and the program is configured to implement the method for recommending a salary scheme according to any one of claims 1-7.

10. A storage medium, characterized by The storage medium stores computer executable instructions, and the computer executable instructions are configured to execute the method for recommending a salary scheme according to any one of claims 1-7.

Images