An AI analysis-based human resources management affair RPA automation task scheduling method

Abstract

The application discloses an AI analysis-based RPA automation task arrangement method for human resources and social security affairs, relates to the technical field of automatic processing, and comprises the following steps: collecting affair element data sets for AI semantic analysis, generating positive and negative evidence condition node sets, merging the conditions of the positive and negative evidence condition node sets, generating a human resource and social security affair positive and negative evidence semantic skeleton, extracting material field nodes in the human resource and social security affair positive and negative evidence semantic skeleton to perform a back negative evidence mirror image backflow positioning, generating a material field negative evidence anchor point chain, performing anchor point negative evidence gate control backfilling on the material field negative evidence anchor point chain, obtaining a negative evidence gate control mapping graph, judging an RPA action type by means of a negative evidence gate control phase arbitration method based on the negative evidence gate control mapping graph, sequentially performing type traction constraint homing and negative evidence priority weaving based on the RPA action type, generating RPA action arrangement constraint information, and constructing an RPA task arrangement graph. The application improves the self-adaptive capability by constructing a human resource and social security affair positive and negative evidence semantic skeleton.

Description

Technical Field

[0001] This invention relates to the field of automation processing technology, and in particular to an AI-based method for RPA-automated task scheduling of human resources and social security administrative matters. Background Technology

[0002] In existing technologies, RPA robots have been used to simulate page operations such as querying, inputting, uploading, saving, and submitting by staff in business systems. OCR, natural language processing, and rule engines are also used for application material recognition, service guide parsing, and policy condition matching. Building on this, some systems further introduce AI models to assist in the analysis of service guides, policy clauses, electronic certificates, and historical processing data, in order to improve the automation level of material pre-review, field filling, and process triggering.

[0003] However, existing RPA (Robotic Process Automation) technologies for human resources and social security administration still have two shortcomings. Firstly, current methods focus primarily on whether application materials are complete, fields match, and page actions are executable, leading to RPA tasks easily revealing anomalies only during the submission, upload, or verification stages. Secondly, existing RPA orchestration typically generates task chains linearly according to the workflow or page sequence, making it difficult to promptly categorize actions into direct execution, temporary storage and completion, rebuttal and blocking, or manual review. This results in insufficient controllability, traceability, and template adaptability in the automated processing process. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an AI-based method for automating RPA task scheduling for human resources and social security administrative matters, addressing the issues of delayed exposure of administrative anomaly risks and the lack of counter-evidence gating basis for RPA actions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides an AI-based method for automated task orchestration in human resources and social security administration, comprising: collecting a dataset of administration business elements for AI semantic analysis to generate a set of positive and negative evidence condition nodes; merging the conditions of the positive and negative evidence condition node set to generate a semantic skeleton of positive and negative evidence for human resources and social security administration; extracting material field nodes from the semantic skeleton of positive and negative evidence for human resources and social security administration, performing backflow and mirroring of backflow for backflow, generating a backflow anchor chain for material field backflow, and performing backflow of backflow for backflow of backflow for backflow of backflow for backflow of backflow, obtaining a backflow gated mapping map; based on the backflow gated mapping map, determining the RPA action type through the backflow gated phase arbitration method, and performing type-driven constraint alignment and backflow priority weaving based on the RPA action type to generate RPA action orchestration constraint information and construct an RPA task orchestration map; executing human resources and social security administration operations according to the RPA task orchestration map, recording administration execution feedback data, and writing the administration execution feedback data back to the semantic skeleton of positive and negative evidence for human resources and social security administration and the backflow gated mapping map for backflow correction, generating an RPA task orchestration template.

[0007] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the specific steps for collecting the dataset of administrative business elements and performing AI semantic analysis to generate a set of positive and negative evidence condition nodes are as follows: The service guide is subjected to hierarchical analysis of service elements, and the handling rules are subjected to polarity analysis of rule clauses. Positive qualification support semantics and negative blocking semantics are obtained and dual-channel discrimination is performed to generate a set of positive and negative proof semantic fragments. Based on the application materials, electronic certificates, and historical return records, the set of semantic fragments for positive and negative proofs is corrected to generate a set of conditional nodes for positive and negative proofs.

[0008] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security affairs, the set of positive and negative evidence condition nodes includes positive qualification support condition nodes, negative blocking condition nodes, historical return hit nodes, material field nodes, matter handling process nodes, and target system page action nodes. The aforementioned semantic skeleton for positive and negative evidence in human resources and social security matters refers to connecting positive qualification support condition nodes along the process nodes of the matter to form a processable support chain, connecting negative blocking condition nodes and historical return hit nodes along the process nodes of the matter to form a negative evidence blocking chain, and attaching material field nodes to the processable support chain and the negative evidence blocking chain respectively.

[0009] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the specific steps for extracting the material field nodes from the semantic skeleton of positive and negative evidence of human resources and social security matters, performing back-to-back evidence mirroring and reflow positioning, and generating a material field negative evidence anchor chain are as follows: Extract the material field nodes from the semantic skeleton of positive and negative evidence for human resources and social security matters, and use the process node of the matter as the mirror reference. The historical return hit nodes are reversed along the anti-evidence blocking chain to the material field nodes under the same process, forming a return anti-evidence mirror path. Material field nodes that are simultaneously located on the feasible support chain and the return counter-evidence mirror path are used as candidate counter-evidence anchors. A feasible blocking reversal judgment is performed on the candidate counter-evidence anchors. When the material field corresponding to the candidate counter-evidence anchor can trigger the counter-evidence blocking chain while supporting the feasible support chain, the field counter-evidence anchor is obtained and connected in the order of the event handling process nodes to generate a material field counter-evidence anchor chain.

[0010] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the specific steps for obtaining the counter-evidence gating mapping diagram are as follows: Extract the executable state and fallback state corresponding to the counter-proof anchor point in the material field counter-proof anchor point chain, and use the executable state as the initial execution state of the target system page action node, and use the fallback state as counter-proof perturbation to feed back to the corresponding target system page action node; When the return to the counter-proof state hits the reverse blocking condition node corresponding to the target system page action node, and causes the feasible support chain corresponding to the target system page action node to break, the gated sweep state is determined. The gated state is passed sequentially along the material field's counter-evidence anchor chain. The gated state includes blocking state, review state, temporary state, and release state arranged from high to low constraint level. When the constraint level of the passed gated state is higher than the constraint level corresponding to the existing execution state of the target system page action node, the execution state is updated with the current gated state to obtain the counter-evidence gated mapping map.

[0011] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the specific steps for determining the RPA action type based on the counter-evidence gating mapping graph and the counter-evidence gating phase arbitration method are as follows: The phase difference is calculated based on the release flag of the executable state and the suppression flag of the return to the counter-evidence state in the counter-evidence gating mapping diagram to obtain the counter-evidence suppression phase value. Based on the counter-evidence suppression phase value, determine whether the executable state has been suppressed by the counter-evidence state. In the corresponding target system page action node, if the counter-evidence blocking chain is not hit but there is a branch waiting for completion before submission, it is determined as a temporary phase. If the counter-evidence blocking chain is hit but the positive actionable support is not retained, it is determined as a blocking phase. If the counter-evidence blocking chain is hit but the positive actionable support is still retained, it is determined as a review phase. Obtain the counter-evidence gating phase type. The phase-driven action projection of the counter-proof gating phase type along the target system page action node in the counter-proof gating mapping diagram is performed to obtain the RPA action type.

[0012] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the specific steps for determining RPA action orchestration constraint information by weaving traction constraints on RPA action types and constructing an RPA task orchestration diagram are as follows: Based on RPA action types, according to the order of the nodes in the process of handling matters, the target system page action nodes in the counter-evidence gating mapping diagram are pulled to the processing path position corresponding to the RPA action type, and the target system page action nodes are bound with the pre-constraints, trigger constraints, execution positions and exception destinations to determine the RPA action orchestration constraint information. Based on RPA action orchestration constraint information, the abnormal destination and execution restrictions of target system page action nodes with counter-evidence constraints are determined, and the normal processing path of target system page action nodes that meet the preconditions is determined, thus constructing an RPA task orchestration diagram.

[0013] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the specific steps for executing human resources and social security administrative operations according to the RPA task orchestration diagram and recording administrative execution feedback data are as follows: Based on the RPA task orchestration diagram, the target system page action nodes are triggered according to the RPA action type and RPA action orchestration constraint information. The direct execution action is performed on the main chain handling page, the temporary completion action enters the completion waiting branch before submission, the counter-proof blocking action enters the termination flow branch along the reverse blocking condition node, and the manual review action inserts the field counter-proof anchor point between the target system page action node, generating a page action execution chain. Based on the page action execution chain, the target system page action nodes are verified by anchor points that leave traces along with the actions, forming the handling and execution feedback data.

[0014] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the administrative execution feedback data includes return reasons, correction opinions, and page execution anomalies. The specific steps for writing back the handling and execution feedback data to the semantic skeleton of positive and negative evidence for human resources and social security matters and the gated mapping diagram of negative evidence for negative evidence backflow correction, and generating a gated phase transition table are as follows: Map the reasons for the return to the reverse blocking condition node in the semantic skeleton of the positive and negative evidence of human resources and social security matters, map the correction opinions to the material field node, map the page execution exception to the target system page action node, and determine the earliest processing stage that can intercept the feedback event in advance according to the order of the processing stage nodes, and generate the pre-emptive negative evidence interception point. Based on the pre-reverse evidence interception point, the target system page action node corresponding to the reason for the rollback is migrated to the blocking phase, the target system page action node corresponding to the correction opinion is migrated to the temporary storage phase, and the target system page action node corresponding to the page execution exception is migrated to the review phase. The phase before migration, the phase after migration, and the corresponding pre-reverse evidence interception point are recorded to generate a gated phase migration table.

[0015] As a preferred embodiment of the AI-based RPA automated task orchestration method for human resources and social security administrative matters described in this invention, the RPA task orchestration template is generated by differentially rewriting the RPA action types and RPA action orchestration constraint information in the RPA task orchestration diagram based on a gated phase transition table.

[0016] The beneficial effects of this invention are as follows: by constructing a semantic skeleton for positive and negative evidence of human resources and social security matters, the handling rules, application evidence, historical return factors and page execution actions are continuously associated, and the historical return risks are brought forward to the material field through the mirror return and reflow positioning of return evidence, thereby improving the adaptive capability of human resources and social security handling automation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of an AI-based RPA (Robotic Process Automation) task orchestration method for human resources and social security administrative matters.

[0019] Figure 2 A flowchart for generating the semantic skeleton.

[0020] Figure 3 The flowchart for template adaptive optimization.

[0021] Figure 4 A flowchart for orchestrating RPA tasks.

[0022] Figure 5 This is a point map showing the distribution of action types.

[0023] Figure 6 This is a comparison chart of the return rates for multiple documents. Detailed Implementation

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0027] Reference Figures 1-6 This is one embodiment of the present invention, which provides an AI-based RPA automated task orchestration method for human resources and social security administrative matters, including the following steps: S1: Collect the data set of business elements to perform AI semantic parsing, generate a set of positive and negative evidence condition nodes, and merge the conditions of the positive and negative evidence condition node set to generate a semantic skeleton of positive and negative evidence for human resources and social security matters. S1.1: The data set of business processing elements includes service guides, processing rules, application materials, electronic certificates, historical return records, and target system page information.

[0028] Specifically, the service guide is retrieved from the service item database or the online service page to obtain the item name, service recipients, application conditions, application materials, processing procedures, and submission requirements; The operating rules are read from policy documents, business operating procedures and the rule database for items to obtain eligibility conditions, prohibition conditions, mutually exclusive conditions, correction rules and manual verification rules; The application materials are obtained from the files uploaded by the applicant during the online application process, including the application form, identity certificate, qualification certificate, and business certificate. Electronic certificates are retrieved from the electronic certificate database or shared data interface to obtain certificate data related to the applicant's identity, insurance status, eligibility, and account information; Historical return records are read from the historical case log of similar matters to obtain the return reason, correction opinions, return process and corresponding material fields; The target system page information is collected from the human resources and social security business system processing page, obtaining the page name, field controls, button actions, upload entry, save entry, submission entry, and receipt download entry.

[0029] All of the above content has been agreed to by the user and is used for legitimate purposes.

[0030] S1.2: Perform hierarchical analysis of service elements on the service guide and polarity analysis of rule clauses on the handling rules to obtain positive qualification support semantics and negative blocking semantics, and perform dual-channel discrimination to generate a set of positive and negative proof semantic fragments.

[0031] Specifically, the service guide is divided into semantic segments of service elements according to the service recipients, application conditions, application materials, processing procedures, and submission requirements. Based on the support relationship between each semantic segment of service elements and the acceptance of matters, the completeness of materials, and the submission of procedures, it is classified into positive qualification support semantics. Based on the restrictive relationship between each semantic segment of service elements and the requirements for missing materials, unmet preconditions, correction requirements, and manual verification requirements, it is classified into reverse blocking semantics.

[0032] The handling rules are divided into semantic segments of rule clauses according to eligibility requirements, prohibition conditions, mutually exclusive conditions, and correction rules. Then, based on the support relationship between each semantic segment of rule clauses and eligibility fulfillment and permitted processing, they are classified into positive eligibility support semantics. Based on the restriction relationship between each semantic segment of rule clauses and prohibition of processing, duplicate applications, mutually exclusive benefits, and inconsistent materials, they are classified into reverse blocking semantics. Using the process of handling a matter as the merging benchmark, positive qualification support semantics with the same meaning under the same process of handling a matter are merged into positive semantic fragments, and negative blocking semantics with the same meaning under the same process of handling a matter are merged into negative semantic fragments. A set of positive and negative proof semantic fragments is then generated from the positive and negative semantic fragments.

[0033] S1.3: Perform return and counter-proof node correction on the set of semantic fragments of positive and negative proofs based on the application materials, electronic certificates and historical return records, and generate a set of conditional nodes for positive and negative proofs.

[0034] Specifically, the material name, field name, field value, certificate status, validity period, and material source in the application materials and electronic certificates are organized into material field nodes.

[0035] The material field nodes are mapped to positive semantic fragments in the set of positive and negative proof semantic fragments. Positive semantic fragments that can prove that the applicant meets the eligibility conditions, material requirements, or submission requirements by the material field nodes are transformed into positive eligibility support condition nodes.

[0036] The material field nodes are mapped to the reverse semantic fragments in the set of positive and negative proof semantic fragments. The reverse semantic fragments that are hit due to the material field nodes having missing, inconsistent, invalid, conflicting or duplicate declaration risks are converted into reverse blocking condition nodes.

[0037] The return reasons, return stages, and return fields in the historical return records are mapped to the existing material field nodes and reverse blocking condition nodes. Historical return records that hit the same material field node or the same reverse blocking condition node are converted into historical return hit nodes.

[0038] The semantic fragments of the positive and negative proofs are organized into nodes for the processing steps of acceptance, verification, uploading, saving, submission and receipt download.

[0039] Then, organize the page fields, control names, and button actions corresponding to query, input, upload, save, submit, and receipt download in the target system page information into target system page action nodes.

[0040] The set of positive and negative evidence condition nodes includes positive qualification support condition nodes, negative blocking condition nodes, historical return hit nodes, material field nodes, matter handling process nodes, and target system page action nodes.

[0041] S1.4: Connect the positive qualification support condition nodes along the process nodes of the matter to form a processable support chain, connect the reverse blocking condition nodes and the historical return hit nodes along the process nodes of the matter to form a counter-evidence blocking chain, and attach the material field nodes to the processable support chain and the counter-evidence blocking chain respectively to generate the semantic skeleton of positive and negative evidence for human resources and social security matters.

[0042] Specifically, the processing nodes are arranged according to the order of processing, and the corresponding positive qualification support condition nodes under each processing node are connected in sequence, so that the output of the positive qualification support condition node under the previous processing node is used as the input of the positive qualification support condition node under the next processing node, thus forming a processing support chain.

[0043] Connect the corresponding reverse blocking condition node under each item processing node to the historical return hit node in sequence, so that the historical return hit node under the same item processing node points to the corresponding reverse blocking condition node, and connect the reverse blocking condition node under the previous item processing node to the reverse blocking condition node under the next item processing node, forming a reverse evidence blocking chain.

[0044] The material field nodes are connected to the feasible support chain positions that can be supported by the material field nodes according to the material source and field meaning, and connected to the counter-evidence blocking chain positions that can be triggered by the material field nodes, so that the same material field node has both a positive feasible support relationship and a reverse blocking counter-evidence relationship.

[0045] The target system page action nodes are mapped to the processing positions in the processing support chain and the counter-evidence blocking chain according to the processing links of the matter, forming a semantic skeleton of positive and negative evidence for human resources and social security matters.

[0046] It should be noted that the semantic skeleton for positive and negative evidence in human resources and social security matters connects positive eligibility support condition nodes along the process nodes to form a feasible support chain, while connecting negative blocking condition nodes and historical rejection hit nodes along the process nodes to form a negative evidence blocking chain. Material field nodes are then attached to both the feasible support chain and the negative evidence blocking chain, thus forming the overall skeleton structure. The function of the semantic skeleton for positive and negative evidence in human resources and social security matters is to establish a two-way logical path of positive support and negative blocking, enabling each material field node to verify the fulfillment of eligibility conditions and capture historical rejection and blocking conditions during execution, thereby achieving precise control over the triggering conditions of RPA actions.

[0047] S2: Extract the material field node from the semantic skeleton of the positive and negative proof of human resources and social security matters, perform backflow and positioning of the negative proof mirror, generate the anchor chain of the material field negative proof, and perform anchor point negative proof gating backflow on the anchor chain of the material field negative proof to obtain the negative proof gating mapping diagram.

[0048] S2.1: Extract the material field nodes from the semantic skeleton of the positive and negative evidence of human resources and social security matters, and use the event handling process node as the mirror reference. Then, backflow the historical return hit nodes along the negative evidence blocking chain to the material field nodes under the same event handling process to form a return negative evidence mirror path.

[0049] Specifically, extract the material field nodes from the semantic skeleton of the positive and negative evidence of human resources and social security matters, and record the field name, field meaning, and the relevant processing stage node of each material field node; extract the historical return hit nodes from the semantic skeleton of the positive and negative evidence of human resources and social security matters, and record the return reason, return field, return stage, and position in the negative evidence blocking chain for each historical return hit node.

[0050] Using the process node of the matter as a mirror reference, the historical return hit node that is consistent with the process node of the matter to which the return stage belongs is identified as the same stage return node, and the return field or return reason in the same stage return node is matched with the field name or field meaning of the material field node.

[0051] When the returned field or reason for return can be matched with the material field node, the historical returned hit node is flowed back to the corresponding material field node along the counter-evidence blocking chain, and a return counter-evidence mirror path is formed in the order of historical returned hit node, counter-evidence blocking chain, matter handling stage node and material field node.

[0052] It should be noted that the return-to-rebuttal mirror path is a path formed by mirroring the nodes in the process of handling a matter, and reversing the flow of historical return hit nodes along the rebuttal blocking chain back to the material field nodes under the same process of handling the matter. The purpose of the return-to-rebuttal mirror path is to transform historical return records into rebuttal evidence for the current material field through post-event feedback, so that subsequent attempts can identify candidate rebuttal anchors that are simultaneously located on the feasibility support chain and the return-to-rebuttal mirror path, thereby reducing the risk that RPA tasks will only expose anomalies during the submission or verification stage.

[0053] S2.2: Select material field nodes that are simultaneously located on the feasible support chain and the return counter-evidence mirror path as candidate counter-evidence anchors, and perform feasible blocking reversal judgment on the candidate counter-evidence anchors. When the material field corresponding to the candidate counter-evidence anchor can trigger the counter-evidence blocking chain while supporting the feasible support chain, obtain the field counter-evidence anchor and connect them in the order of the event handling process nodes to generate a material field counter-evidence anchor chain.

[0054] Specifically, by comparing the attachment position of the material field node in the feasible support chain and the return position in the return counter-evidence mirror path, the material field node that exists in both the feasible support chain and the return counter-evidence mirror path is identified as a candidate counter-evidence anchor point.

[0055] Verify whether the material fields corresponding to the candidate counter-evidence anchor point can meet the positive qualification support condition node under the same matter handling stage node in the available support chain. Then verify whether the material fields corresponding to the candidate counter-evidence anchor point can simultaneously hit the reverse blocking condition node or the historical return hit node under the same matter handling stage node in the counter-evidence blocking chain. When the material fields corresponding to the candidate counter-evidence anchor point can serve as the processing support basis for the positive qualification support condition node and the return trigger basis for the reverse blocking condition node or the historical return hit node, the candidate counter-evidence anchor point is determined as the field counter-evidence anchor point.

[0056] Arrange the field rebuttal anchors according to the order of the handling process nodes to which the field rebuttal anchors belong, and connect the field rebuttal anchors under the same handling process node according to the material source order of the material field node to generate a material field rebuttal anchor chain.

[0057] It should be noted that the material field rebuttal anchor chain is a field risk chain formed by connecting field rebuttal anchors that simultaneously have a positive actionable support relationship and a reverse return trigger relationship, according to the order of the nodes in the handling process; it is used to transmit the executable status and return rebuttal status of the material field to the corresponding target system page action node, thereby improving the accuracy of the handling action arrangement and the ability to control risks in advance.

[0058] S2.3: Extract the executable state and fallback state corresponding to the counter-proof anchor point in the material field counter-proof anchor point chain, and use the executable state as the initial execution state of the target system page action node, and use the fallback state as counter-proof disturbance to feed back to the corresponding target system page action node.

[0059] Specifically, based on the connection relationship of the field counter-evidence anchor point in the feasible support chain, it is verified whether the material field corresponding to the field counter-evidence anchor point meets the positive qualification support condition node. If the positive qualification support condition node is met, the field counter-evidence anchor point is marked as executable. Based on the hit relationship of the field counter-evidence anchor point in the return counter-evidence mirror path and the counter-evidence blocking chain, it is verified whether the material field corresponding to the field counter-evidence anchor point has missing fields, inconsistent fields, invalid certificates, material conflicts, or historical return reasons, and a return counter-evidence status is generated.

[0060] Based on the process node of the matter to which the field rebuttal anchor belongs, find the target system page action node under the same process node that corresponds to the field rebuttal anchor, and use the return to rebuttal status as a rebuttal disturbance to feed back to the target system page action node.

[0061] S2.4: When the return-to-the-contrast state hits the reverse blocking condition node corresponding to the target system page action node, and causes the feasible support chain corresponding to the target system page action node to break, determine the gated sweep state.

[0062] Specifically, the return of the counter-evidence status is matched with the reverse blocking condition node attached to the action node of the target system page. When the return of the counter-evidence status originates from missing fields, inconsistent fields, expired certificates, conflicting materials, or a historical return reason, and the reverse blocking condition node is used to restrict the continued processing of the same material field or the same matter, it is determined that the return of the counter-evidence status matches the reverse blocking condition node corresponding to the action node of the target system page.

[0063] If the process supports the positive eligibility support conditions node by checking the action node connected to the action node on the target system page, and if the process returns to the state of reversal, causing the field reversal anchor point to no longer meet the positive eligibility support conditions node, or causing the action node on the target system page to no longer be able to undertake subsequent processing actions, the process is deemed to be broken in the support chain corresponding to the action node on the target system page.

[0064] After the return of the counter-evidence status hits the reverse blocking condition node and causes the feasible support chain to break, the gated impact status is determined according to the degree of restriction of the return of the counter-evidence status. Among them, the return of the counter-evidence status for material correction forms a temporary impact status, the return of the counter-evidence status for prohibited processing forms a blocking impact status, and when the field has both feasible support relationship and return blocking relationship, a verification impact status is formed, thus obtaining the gated impact status corresponding to the action node of the target system page.

[0065] S2.5: Pass the gated state sequentially along the material field's counter-evidence anchor chain. The gated state includes blocking state, verification state, temporary state, and release state arranged from high to low constraint level. When the constraint level of the passed gated state is higher than the constraint level corresponding to the existing execution state of the target system page action node, update the execution state with the current gated state and obtain the counter-evidence gated mapping map.

[0066] Specifically, the rebuttal anchors in the material field rebuttal anchor chain are read sequentially according to the order of the handling process nodes, and the gating and impact status corresponding to the rebuttal anchors are passed to the target system page action node under the same handling process node.

[0067] Using the existing execution status of the target system page action node as the comparison object, the constraint levels of the blocking impact status, review impact status, temporary impact status and executable status are established according to the strength of the constraint (the blocking impact status is higher than the review impact status, the review impact status is higher than the temporary impact status, and the temporary impact status is higher than the executable status).

[0068] When the constraint level corresponding to the gated sweep state passed to the target system page action node is higher than the constraint level corresponding to the existing execution state of the target system page action node, the execution state of the target system page action node is updated with the gated sweep state passed to the target system page action node.

[0069] When the constraint level corresponding to the gating state passed to the target system page action node is lower than or equal to the constraint level corresponding to the existing execution state of the target system page action node, the existing execution state of the target system page action node is retained.

[0070] Bind the correspondence between the field proof anchor point, the gated impact state, the target system page action node, and the updated execution state to obtain the proof gate mapping graph.

[0071] It should be noted that the counter-evidence gating map is generated by passing the gating sweep status along the counter-evidence anchor chain of the material field and updating the execution status of the target system page action node. It is used to establish the dynamic constraint relationship of each page action node during the execution of RPA tasks. Through the counter-evidence gating map, abnormal constraints can be reflected in a timely manner in automated operation, improving the controllability, traceability and template self-adaptation capability of the human resources and social security operation process.

[0072] S3: Based on the proof-of-contrast gating mapping diagram, the RPA action type is determined by the proof-of-contrast gating phase arbitration method, and the type-driven constraint is aligned and proof-of-contrast priority is woven in sequence based on the RPA action type to generate RPA action orchestration constraint information and construct the RPA task orchestration diagram.

[0073] S3.1: Perform phase difference calculation based on the release flag of the executable state and the suppression flag of the return to the counter-proof state in the counter-proof gating mapping diagram to obtain the counter-proof suppression phase value.

[0074] Specifically, the content in the executable state that allows the target system page action node to continue processing is marked as a release identifier. The number of positive qualification support condition nodes associated with the target system page action node is counted, and the number of positive qualification support condition nodes that have been satisfied by the material fields corresponding to the field rebuttal anchor is counted. The ratio of the number of satisfied positive qualification support condition nodes to the number of associated positive qualification support condition nodes is used as the release level value.

[0075] The content that restricts the target system page action node from continuing to process in the return to the counter-evidence state is marked as a suppression identifier. The number of reverse blocking condition nodes and historical return hit nodes associated with the target system page action node are counted. The number of reverse blocking condition nodes and historical return hit nodes that are hit by the field counter-evidence anchor is also counted. The ratio of the number of hit reverse blocking condition nodes and historical return hit nodes to the number of associated reverse blocking condition nodes and historical return hit nodes is used as the suppression level value.

[0076] The number of field counter-evidence anchors that the target system page action node receives in the gating sweep status is counted. The number of field counter-evidence anchors in the material field counter-evidence anchor chain that are located before and within the handling stage of the matter to which the target system page action node belongs is counted. The ratio of the number of field counter-evidence anchors that receive the gating sweep status to the above-mentioned number of field counter-evidence anchors is used as the gating sweep constraint level value.

[0077] Using the release flag as the positive phase reference and the suppression flag as the negative phase reference, phase difference calculation is performed on the release and suppression flags under the same target system page action node to obtain the counter-suppression phase value, expressed as: ; in, To prove the suppression of phase value, The clearance level value, To suppress level values, This refers to the gated wave and constraint level value. This is the anchor point preconditioning coefficient for proof by contradiction.

[0078] It should be noted that the pre-evaluation coefficient of the counter-evidence anchor point is obtained by the ratio of the order position of the current field's counter-evidence anchor point in the material field's counter-evidence anchor point chain to the total number of anchor points in the chain, and its value range is [value missing]. .

[0079] S3.2: Based on the counter-evidence suppression phase value, determine whether the executable state has been suppressed by the counter-evidence state. If the corresponding target system page action node does not hit the counter-evidence blocking chain but has a pre-commit completion waiting branch, it is determined as a temporary phase. If it hits the counter-evidence blocking chain but does not retain positive actionable support, it is determined as a blocking phase. If it hits the counter-evidence blocking chain but still retains positive actionable support, it is determined as a verification phase. Obtain the counter-evidence gating phase type.

[0080] Specifically, when the counter-evidence suppression phase value is greater than the preset suppression judgment value (obtained by calibrating the distribution of counter-evidence suppression phase values ​​of normally released and returned suppression samples in the historical processing of similar human resources and social security matters, with a value range of [missing value]), the suppression phase value is determined. When the suppression phase value is less than or equal to the preset suppression judgment value, it is determined that the suppression flag has not been weakened or has not covered the release flag. When the suppression flag has not been weakened or has not covered the release flag, it is determined that the executable state has not been suppressed by the return to the proof-of-contrast state, and the corresponding target system page action node is determined as the release phase. When the suppression flag weakens or covers the release flag, and the corresponding target system page action node has been connected to the pre-commit completion waiting branch, it is determined that the executable state has been suppressed by the return to the proof-of-contrast state but still has a completion acceptance path, and the corresponding target system page action node is determined as the temporary storage phase. When the suppression flag weakens or covers the release flag, and the corresponding target system page action node hits the disproving blocking chain, the executable state is determined to be suppressed by the reverting disproving state and enters the blocking path, and the corresponding target system page action node is determined to be the blocking phase; when the suppression flag weakens or covers the release flag, and the corresponding field disproving anchor point is still connected to the actionable support chain, the same target system page action node is determined to have both positive actionable support and reverting disproving suppression, and the corresponding target system page action node is determined to be the review phase.

[0081] The gated phase types for counter-evidence include release phase, temporary phase, blocking phase, and verification phase.

[0082] S3.3: Project the phase type of the counter-proof gating along the target system page action node in the counter-proof gating mapping diagram to obtain the RPA action type.

[0083] Specifically, the target system page action node is used as the projection object, and the disproving gating phase type is used as the projection basis to convert the disproving gating phase type into the action control attribute of the target system page action node.

[0084] Convert the permission phase into an action control attribute that allows continued triggering of page actions, and determine the corresponding target system page action node as the action to be executed directly; The temporary phase is converted into an action control attribute that retains the current page input state and pauses submission, and the corresponding target system page action node is determined as the temporary completion action.

[0085] The blocking phase is converted into an action control attribute that stops the current page submission and cuts off subsequent page actions, and the corresponding target system page action node is identified as the blocking action.

[0086] The phase verification conversion adds a verification confirmation action control attribute between the field rebuttal anchor point and the target system page action node, and determines the corresponding target system page action node as a manual verification action.

[0087] RPA action types include direct execution actions, temporary completion actions, evidence-blocking actions, and manual review actions.

[0088] S3.4: Based on the RPA action type, according to the order of the task processing nodes, the target system page action node in the reverse evidence gating mapping diagram is pulled to the processing path position corresponding to the RPA action type, and the target system page action node is bound with the pre-constraints, trigger constraints, execution position and exception destination to determine the RPA action orchestration constraint information.

[0089] Specifically, based on RPA action types and the counter-evidence gating mapping, the basic position of the target system page action node in the process is determined according to the order of the task handling nodes; the target system page action node corresponding to the action to be executed directly is retained in the basic position of the task handling node, and the completion status of the previous task handling node is used as a prerequisite constraint.

[0090] The target system page action node corresponding to the temporary completion action is moved to the position before the submission action, and the material field completion status of the corresponding field counter-proof anchor point is used as the trigger constraint; the target system page action node corresponding to the counter-proof blocking action is moved to the position after the reverse blocking condition node, and the counter-proof blocking chain hit status is used as the termination flow constraint; the target system page action node corresponding to the manual review action is moved to the position between the field counter-proof anchor point and the original target system page action node, and the RPA action orchestration constraint information is determined.

[0091] S3.5: Based on RPA action orchestration constraint information, determine the abnormal destination and execution restrictions for target system page action nodes with counter-evidence constraints, and determine the normal processing path for target system page action nodes that meet the preconditions, and construct an RPA task orchestration diagram.

[0092] Specifically, the corresponding target system page action nodes are arranged according to the order of the task handling process, and the preconditions, trigger constraints, execution location, and exception destination of each target system page action node are included in the same arrangement relationship.

[0093] Connect the rebuttal blocking action that satisfies the termination flow constraint to the termination flow path corresponding to the reverse blocking condition node; connect the manual review action that satisfies the review confirmation constraint to the field rebuttal anchor point and the target system page action node; connect the temporary completion action that satisfies the material field completion status to the submission action; and connect the direct execution action that satisfies the preconditions to the processing path of the corresponding matter processing node.

[0094] After connecting all the action nodes of the target system page under the processing nodes of all matters, an RPA task orchestration diagram is formed, which includes the normal processing path, the waiting branch before submission, the termination of the process branch, and the manual review branch.

[0095] like Figure 5This section explains the differences in how page action nodes are assigned to different RPA action types under different RPA orchestration methods. The fixed-flow RPA group operates by triggering the target system's page action nodes sequentially based on the page flow, without considering positive qualification support relationships or negative blocking relationships. As long as page buttons are clickable, fields can be entered, and materials can be uploaded, subsequent submission actions continue. Therefore, most actions are concentrated on direct execution, with fewer actions involving temporary completion, rebuttal blocking, and manual review. The field validation RPA group checks for missing material fields, abnormal field formats, and valid certificate status before page submission. If common anomalies are found at the material field level, the relevant page actions are moved to temporary completion. If no obvious field anomalies are found, execution continues. However, it cannot convert factors such as historical rejection reasons, mutually exclusive benefits, duplicate applications, and manual verification rules into gating constraints on the rebuttal blocking chain. The specific operation of the RPA task orchestration group of this invention is as follows: based on the executable state and the reversal state of the material field's counter-evidence anchor point, the reversal state is fed back to the corresponding target system page action node; when the reversal state hits the reverse blocking condition node and affects the feasible support chain, a temporary impact state, a blocking impact state, or a review impact state is formed; the page action node is converted into a direct execution action, a temporary completion action, a counter-evidence blocking action, or a manual review action. In the figure, the distribution of the temporary completion, counter-evidence blocking, and manual review actions in this invention group is more comprehensive, enabling RPA task orchestration to change from fixed script execution to dynamic orchestration based on the positive and negative evidence relationship.

[0096] S4: Execute human resources and social security handling operations according to the RPA task orchestration diagram, record the handling execution feedback data, and write the handling execution feedback data back to the human resources and social security matter positive and negative proof semantic skeleton and negative proof gating mapping diagram for negative proof backflow correction, and generate RPA task orchestration template.

[0097] S4.1: Based on the RPA task orchestration diagram, trigger the target system page action node according to the RPA action type and RPA action orchestration constraint information. Directly execute the action to perform the main chain handling page operation, so that the temporary completion action enters the completion waiting branch before submission, so that the counter-evidence blocking action enters the termination flow branch along the reverse blocking condition node, so that the manual review action inserts the field counter-evidence anchor point between the target system page action node, and generates the page action execution chain.

[0098] Specifically, based on the pre-constraints, trigger constraints, execution location, and exception destination in the RPA action orchestration constraint information, the triggering order of the target system page action nodes is determined according to the order of the task handling process nodes. Then, according to the RPA action type corresponding to the target system page action node, the main chain handling page operation is completed for the target system page action nodes belonging to the execution action type in the order of page action sequence: query, input, upload, save, submit, and receipt download, and the completion status is passed to the next task handling process node.

[0099] The target system page action node corresponding to the temporary completion action pauses submission before the action is submitted, and enters the completion waiting branch based on the completion status of the material field corresponding to the field counter-evidence anchor point; the target system page action node corresponding to the counter-evidence blocking action stops subsequent page actions when the reverse blocking condition node is hit, and enters the termination flow branch along the reverse blocking condition node; the target system page action node corresponding to the manual review action adds a review confirmation action between the field counter-evidence anchor point and the target system page action node, and triggers the target system page action node only after the review confirmation is completed.

[0100] Generate a page action execution chain by connecting the actual triggering order of the target system page action nodes, RPA action types, entry branches, and completion states.

[0101] It should be noted that the page action execution chain refers to the sequence of operations formed by triggering page action nodes in the target system according to the RPA task orchestration diagram, the RPA action type, and the RPA action orchestration constraints, and linking different types of actions in the actual triggering order, entry branches, and completion status. Through the page action execution chain, the operation path and status changes can be completely recorded, enabling traceability of operations, clarity of operation logic, and immediate identification and subsequent processing of abnormal operations.

[0102] S4.2: Based on the page action execution chain, the target system page action nodes are verified by leaving traces of the anchor points along with the actions, forming the handling execution feedback data.

[0103] Specifically, according to the triggering order of the target system page action nodes, the processing status, material status, and page status are respectively bound to the fields of the same target system page action node: the counter-evidence anchor point, the RPA action type, and the counter-evidence gated phase type.

[0104] When the processing status indicates that the application is not accepted, the applicant is not qualified, the application is duplicated, the benefits are mutually exclusive, or the application is returned for modification, a reason for return is generated; when the material status indicates that the material is missing, the field is empty, the format is incorrect, the certificate is expired, or supplementary explanation is required, a correction opinion is generated; when the page status indicates that the page redirection failed, the control failed to trigger, the material upload failed, the form failed to save, or the receipt download failed, a page execution exception is generated.

[0105] The reasons for return, correction suggestions, and page execution exceptions are bound to the corresponding target system page action nodes, field rebuttal anchors, RPA action types, and rebuttal gate phase types to form the handling execution feedback data.

[0106] The data on feedback from the execution process includes the reason for the return, suggestions for correction, and page execution anomalies.

[0107] S4.3: Map the reason for return to the reverse blocking condition node in the semantic skeleton of human resources and social security matters, map the correction opinion to the material field node, map the page execution exception to the target system page action node, and determine the earliest processing stage that can intercept the feedback event in advance according to the order of the processing stage nodes, and generate the pre-reverse evidence interception point.

[0108] Specifically, based on the binding relationship between the reasons for return, the rectification opinions, and the page execution anomalies in the feedback data of the handling and execution, and the page action nodes of the target system, the handling process nodes of the matters in which the reasons for return, the rectification opinions, and the page execution anomalies occurred are determined.

[0109] Based on the meaning of the processing status corresponding to the reason for return, the reverse blocking condition node under the same processing stage node or the processing stage node of the preceding matter is matched in the reverse blocking chain of the semantic skeleton of positive and negative evidence of human resources and social security matters, and the reason for return is mapped to the matched reverse blocking condition node.

[0110] Based on the material name, field name, and field status corresponding to the correction opinion, match the material field node under the same matter handling stage node or the preceding matter handling stage node within the material field node in the semantic skeleton of human resources and social security matters, and map the correction opinion to the matched material field node.

[0111] Based on the page name, action name, and failure status corresponding to the page execution exception, match the target system page action node under the same matter handling stage node within the target system page action node in the semantic skeleton of human resources and social security matters, and map the page execution exception to the matched target system page action node.

[0112] Based on the order of the process nodes, the earliest processing stage is determined by identifying the first point where the reason for return, the correction opinion, or the page execution abnormality can be identified by the corresponding reverse blocking condition node, material field node, or target system page action node, and a pre-reverse evidence interception point is generated.

[0113] It should be noted that the pre-emptive counter-evidence interception point can determine the earliest interceptable position in the handling process of the return reason, correction opinion, or page execution abnormality in the handling feedback data. The role of the pre-emptive counter-evidence interception point is to enable potential abnormalities in the handling process to be identified and intervened at the earliest controllable position, thereby ensuring that subsequent RPA actions are executed according to the blocking, temporary storage, or review logic, and improving the controllability and adaptability of human resources and social security handling operations.

[0114] S4.4: Based on the pre-reverse evidence interception point, the target system page action node corresponding to the reason for the rollback is migrated to the blocking phase, the target system page action node corresponding to the correction opinion is migrated to the temporary storage phase, and the target system page action node corresponding to the page execution exception is migrated to the review phase. The phase before migration, the phase after migration and the corresponding pre-reverse evidence interception point are recorded to generate a gated phase migration table.

[0115] Specifically, the earliest processing step, feedback source, and corresponding node recorded in the pre-established counter-evidence interception point are used to determine the target system page action node that needs to be migrated, and the counter-evidence gating phase type currently corresponding to the target system page action node is determined in the counter-evidence gating mapping diagram.

[0116] When the current counter-evidence interception point originates from the reason for the return, and the reason for the return has been mapped to the reverse blocking condition node, the current counter-evidence gating phase type of the target system page action node is migrated to the blocking phase; when the current counter-evidence interception point originates from the correction opinion, and the correction opinion has been mapped to the material field node, the current counter-evidence gating phase type of the target system page action node is migrated to the temporary phase; when the current counter-evidence interception point originates from the page execution exception, and the page execution exception has been mapped to the target system page action node, the current counter-evidence gating phase type of the target system page action node is migrated to the review phase.

[0117] Record the target system page action nodes, the phase before migration, the phase after migration, the pre-proof interception point, and the feedback source accordingly to generate a gated phase migration table.

[0118] S4.5: Based on the gated phase transfer table, differentially rewrite the RPA action types and RPA action orchestration constraint information in the RPA task orchestration diagram to generate an RPA task orchestration template.

[0119] Specifically, using the target system page action node, pre-migration phase, and post-migration phase recorded in the gated phase migration table as comparison objects, the RPA action type and RPA action orchestration constraint information corresponding to the same target system page action node are located in the RPA task orchestration diagram.

[0120] When the migrated phase is a blocking phase, the RPA action type of the corresponding target system page action node is rewritten as a disproving blocking action, and the RPA action orchestration constraint information is rewritten as a termination flow constraint after connecting the reverse blocking condition node; when the migrated phase is a temporary phase, the RPA action type of the corresponding target system page action node is rewritten as a temporary completion action, and the RPA action orchestration constraint information is rewritten as a completion waiting constraint before the submission action; when the migrated phase is a verification phase, the RPA action type of the corresponding target system page action node is rewritten as a manual verification action, and the RPA action orchestration constraint information is rewritten as a verification confirmation constraint between the field disproving anchor and the target system page action node.

[0121] Save the rewritten RPA action types, RPA action orchestration constraints, pre-proof interception points, and target system page action nodes to generate an RPA task orchestration template.

[0122] like Figure 6 This section explains the different solutions' ability to control the risk of document returns as business volume continues to increase. The baseline solution operates as follows: the RPA robot only executes query, input, upload, save, submit, and receipt download actions according to the page order in the target system's page information. Service guides, handling rules, application materials, and electronic certificates are only used for routine filling; no semantic skeleton for positive and negative evidence of human resources and social security matters is constructed, and historical return records are not routed back to the material field node. Therefore, when material fields have conflicts, are invalid, missing, or inconsistent with policy blocking conditions, a return is only generated after submission or verification. The specific operation of this invention is as follows: the service guide, handling rules, application materials, electronic certificates, historical return records, and target system page information are parsed into a set of positive and negative evidence condition nodes. The positive qualification support condition nodes are linked together to form a feasible support chain, and the negative blocking condition nodes and historical return hit nodes are linked together to form a negative evidence blocking chain. The material field nodes are then attached to the two chains. Through the return negative evidence mirror backflow positioning, the historical return risk is brought forward to the material field nodes, generating a material field negative evidence anchor chain. Finally, the negative evidence gating mapping diagram is used to influence the target system page action nodes. Figure 6 The return rate of the intermediate baseline scheme increases with the increase of business volume, while the curve of the scheme of this invention remains at a low level, indicating that the risk of return can be detected in advance before submission, reducing the number of subsequent returns, and demonstrating the adaptive capability of human resources and social security administration automation.

[0123] In summary, this invention improves the adaptability of human resources and social security processing automation by constructing a semantic skeleton for positive and negative evidence of human resources and social security matters, which makes the handling rules, application evidence, historical return factors and page execution actions form a continuous association, and by positioning the historical return risk in the material field through the mirror return of negative evidence.

[0124] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for automated task scheduling of human resources and social security administrative matters based on AI analysis, characterized in that, include: Collect the data set of business elements and perform AI semantic parsing to generate a set of positive and negative evidence condition nodes. Then, merge the conditions of the positive and negative evidence condition node sets to generate a semantic skeleton of positive and negative evidence for human resources and social security matters. Extract the material field node from the semantic skeleton of the positive and negative proof of human resources and social security matters, perform backflow and positioning of the negative proof mirror, generate the anchor chain of the material field negative proof, and perform anchor point negative proof gating backflow on the anchor chain of the material field negative proof to obtain the negative proof gating mapping diagram. Based on the counter-evidence gating mapping diagram, the RPA action type is determined by the counter-evidence gating phase arbitration method. Based on the RPA action type, type-driven constraints are assigned to determine the RPA action orchestration constraint information. Based on the RPA action orchestration constraint information, the abnormal destination and execution restrictions of the target system page action nodes with counter-evidence constraints are determined, and the normal processing path of the target system page action nodes that meet the preconditions is determined, thus constructing the RPA task orchestration diagram. Based on the RPA task orchestration diagram, perform human resources and social security handling operations, record handling execution feedback data, and write the handling execution feedback data back to the human resources and social security matter positive and negative proof semantic skeleton and negative proof gating mapping diagram for negative proof backflow correction, and generate RPA task orchestration template.

2. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 1, characterized in that, The collected data set of business elements is subjected to AI semantic parsing to generate a set of positive and negative evidence condition nodes. The specific steps are as follows: The service guide is subjected to hierarchical analysis of service elements, and the handling rules are subjected to polarity analysis of rule clauses. Positive qualification support semantics and negative blocking semantics are obtained and dual-channel discrimination is performed to generate a set of positive and negative proof semantic fragments. Based on the application materials, electronic certificates, and historical return records, the set of semantic fragments for positive and negative proofs is corrected to generate a set of conditional nodes for positive and negative proofs.

3. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 1, characterized in that, The set of positive and negative proof condition nodes includes positive qualification support condition nodes, negative blocking condition nodes, historical return hit nodes, material field nodes, matter handling process nodes, and target system page action nodes. The aforementioned semantic skeleton for positive and negative evidence in human resources and social security matters refers to connecting positive qualification support condition nodes along the process nodes of the matter to form a processable support chain, connecting negative blocking condition nodes and historical return hit nodes along the process nodes of the matter to form a negative evidence blocking chain, and attaching material field nodes to the processable support chain and the negative evidence blocking chain respectively.

4. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 3, characterized in that, The process of extracting the material field nodes from the semantic skeleton of positive and negative evidence for human resources and social security matters, performing a backtracking and mirroring of the negative evidence, and generating a material field negative evidence anchor chain is as follows: Extract the material field nodes from the semantic skeleton of positive and negative evidence for human resources and social security matters, and use the process node of the matter as the mirror reference. The historical return hit nodes are reversed along the anti-evidence blocking chain to the material field nodes under the same process, forming a return anti-evidence mirror path. Material field nodes that are simultaneously located on the feasible support chain and the return counter-evidence mirror path are used as candidate counter-evidence anchors. A feasible blocking reversal judgment is performed on the candidate counter-evidence anchors. When the material field corresponding to the candidate counter-evidence anchor can trigger the counter-evidence blocking chain while supporting the feasible support chain, the field counter-evidence anchor is obtained and connected in the order of the event handling process nodes to generate a material field counter-evidence anchor chain.

5. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 4, characterized in that, The specific steps for obtaining the counter-evidence gating mapping are as follows: Extract the executable state and fallback state corresponding to the counter-proof anchor point in the material field counter-proof anchor point chain, and use the executable state as the initial execution state of the target system page action node, and use the fallback state as counter-proof perturbation to feed back to the corresponding target system page action node; When the return to the counter-proof state hits the reverse blocking condition node corresponding to the target system page action node, and causes the feasible support chain corresponding to the target system page action node to break, the gated sweep state is determined. The gated state is passed sequentially along the material field's counter-evidence anchor chain. The gated state includes blocking state, review state, temporary state, and release state arranged from high to low constraint level. When the constraint level of the passed gated state is higher than the constraint level corresponding to the existing execution state of the target system page action node, the execution state is updated with the current gated state to obtain the counter-evidence gated mapping map.

6. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 5, characterized in that, The specific steps for determining the RPA action type based on the proof-of-contrast gated mapping graph and the proof-of-contrast gated phase arbitration method are as follows: The phase difference is calculated based on the release flag of the executable state and the suppression flag of the return to the counter-evidence state in the counter-evidence gating mapping diagram to obtain the counter-evidence suppression phase value. Based on the counter-evidence suppression phase value, determine whether the executable state has been suppressed by the counter-evidence state. In the corresponding target system page action node, if the counter-evidence blocking chain is not hit but there is a branch waiting for completion before submission, it is determined as a temporary phase. If the counter-evidence blocking chain is hit but the positive actionable support is not retained, it is determined as a blocking phase. If the counter-evidence blocking chain is hit but the positive actionable support is still retained, it is determined as a review phase. Obtain the counter-evidence gating phase type. The phase-driven action projection of the counter-proof gating phase type along the target system page action node in the counter-proof gating mapping diagram is performed to obtain the RPA action type.

7. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 6, characterized in that, The process involves classifying and aligning RPA action types to determine RPA action orchestration constraints. Based on these constraints, abnormal paths and execution restrictions are identified for target system page action nodes with counter-evidence constraints. Normal processing paths are determined for target system page action nodes that meet pre-conditions. This process constructs an RPA task orchestration diagram. The specific steps are as follows: Based on RPA action types, according to the order of the nodes in the process of handling matters, the target system page action nodes in the counter-evidence gating mapping diagram are pulled to the processing path position corresponding to the RPA action type, and the target system page action nodes are bound with the pre-constraints, trigger constraints, execution positions and exception destinations to determine the RPA action orchestration constraint information. Based on RPA action orchestration constraint information, the abnormal destination and execution restrictions of target system page action nodes with counter-evidence constraints are determined, and the normal processing path of target system page action nodes that meet the preconditions is determined, thus constructing an RPA task orchestration diagram.

8. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 7, characterized in that, The specific steps for executing human resources and social security administrative operations based on the RPA task orchestration diagram and recording the operational feedback data are as follows: Based on the RPA task orchestration diagram, the target system page action nodes are triggered according to the RPA action type and RPA action orchestration constraint information. The direct execution action is performed on the main chain handling page, the temporary completion action enters the completion waiting branch before submission, the counter-proof blocking action enters the termination flow branch along the reverse blocking condition node, and the manual review action inserts the field counter-proof anchor point between the target system page action node, generating a page action execution chain. Based on the page action execution chain, the target system page action nodes are traced along with the action to obtain the counter-evidence anchor points, forming the handling and execution feedback data; the handling and execution feedback data includes the reason for return, the correction opinion, and the page execution anomaly.

9. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 8, characterized in that, The specific steps for writing back the handling and execution feedback data to the semantic skeleton of positive and negative evidence for human resources and social security matters and the gated mapping diagram of negative evidence for negative evidence backflow correction, and generating a gated phase transition table are as follows: Map the reasons for rejection to the reverse blocking condition nodes in the semantic skeleton of positive and negative evidence for human resources and social security matters, map the correction opinions to the material field nodes, map the page execution exceptions to the target system page action nodes, and determine the earliest processing stage that can intercept feedback events in advance according to the order of the processing stages of the matter, and generate the pre-emptive negative evidence interception point. Based on the pre-reverse evidence interception point, the target system page action node corresponding to the reason for the rollback is migrated to the blocking phase, the target system page action node corresponding to the correction opinion is migrated to the temporary storage phase, and the target system page action node corresponding to the page execution exception is migrated to the review phase. The phase before migration, the phase after migration, and the corresponding pre-reverse evidence interception point are recorded to generate a gated phase migration table.

10. The AI-based RPA automated task scheduling method for human resources and social security administrative matters as described in claim 9, characterized in that, The RPA task orchestration template is generated by differentially rewriting the RPA action types and RPA action orchestration constraint information in the RPA task orchestration diagram based on a gated phase transition table.

Images