Method for checking consistency of cross-module reasoning results of ai-based centralized procurement governance hub

Abstract

This application discloses a method for cross-module reasoning result consistency verification in an AI-based centralized procurement governance hub. The method includes: acquiring reasoning results generated by AI models of multiple business modules within the centralized procurement governance hub based on procurement governance tasks; performing logical consistency verification on multiple reasoning results based on preset logical consistency verification rules or constraints, and performing temporal consistency verification on multiple reasoning results based on the generation timestamps of each reasoning result and the execution dependency order or path relationship between each business module; triggering a conflict handling mechanism when conflicts or inconsistencies are detected, and performing conflict rollback and result correction processing on conflicting reasoning results to obtain a verified and corrected consistent governance result; and integrating the various reasoning results and the consistent governance result to obtain a procurement governance decision. This method can improve the reliability, auditability, and traceability of procurement governance decisions.

Description

Technical Field

[0001] This application relates to the field of centralized procurement governance technology, and in particular to a method for cross-module reasoning result consistency verification of a centralized procurement governance hub based on AI. Background Technology

[0002] In centralized procurement governance scenarios, to achieve comprehensive analysis and decision-making capabilities, it is often necessary to deploy multiple functionally independent business modules within the procurement system, such as supplier evaluation modules, price compliance analysis modules, and contract risk identification modules. Within these business modules, artificial intelligence technology can be used to perform rule-based reasoning based on the data within each module, thus forming a valid basis for governance decisions.

[0003] However, the multiple business modules within the system also present challenges in ensuring consistency and coordination of results. Conventional methods require listing or merging the inference results from each module and relying on professionals for comparison and review. However, this review approach often fails to meet the complex rules and conditions required for procurement governance, and it struggles to adequately verify logical conflicts and data dependencies between the inference results of different modules, leading to erroneous decisions that are difficult to audit and trace. Summary of the Invention

[0004] In view of this, the embodiments of this application provide a cross-module reasoning result consistency verification method for a centralized procurement governance hub based on AI. The main purpose is to solve the technical problems of poor coordination and consistency of reasoning results among multiple modules in the procurement governance system, as well as poor reliability, auditability and traceability of decision results.

[0005] This application embodiment achieves consistent governance of cross-module reasoning results by constructing a logical consistency verification mechanism, a time sequence consistency verification mechanism, and a conflict rollback and result correction mechanism in the centralized procurement governance center, thereby ensuring the reliability, auditability, and traceability of procurement governance decisions.

[0006] According to one aspect of this application, a method for cross-module inference result consistency verification of an AI-based centralized procurement governance center is provided, the method being executed within the centralized procurement governance center, the method comprising:

[0007] The reasoning results generated by the artificial intelligence models of multiple business modules in the centralized procurement governance center based on procurement governance tasks are obtained, wherein each reasoning result includes decision suggestion information of the business module; Based on preset logical consistency verification rules or constraints, logical consistency verification is performed on multiple inference results to detect whether there are conflicts, mutual exclusions or duplicate executions between different inference results. Based on the generation timestamps of each inference result and the execution dependency order or path relationship between each business module, a time order consistency check is performed on multiple inference results to detect whether there is a mismatch between the generation order and the dependency relationship of different inference results, or whether the data dependency is not satisfied but has entered the execution of subsequent modules. When a conflict or inconsistency is detected in logical consistency verification or time sequence consistency verification, a conflict handling mechanism is triggered, and conflict rollback and result correction are performed on the conflicting inference results to obtain a verified and corrected consistency governance result. The inference results output by each of the business modules and the consistency governance results are integrated to obtain the procurement governance decision for the procurement governance task.

[0008] Optionally, the method for constructing the logical consistency verification rules or constraints includes: determining business modules related to the procurement governance task in the centralized procurement governance hub according to the business domain involved in the procurement governance task; obtaining a rule tree corresponding to each business module, wherein the rule tree includes upper-level governance target nodes, middle-level rule boundary nodes, and lower-level business rule nodes; extracting all business logic rules defined by lower-level business rule nodes in each rule tree, and merging the extracted business logic rules to obtain the logical consistency verification rules or constraints.

[0009] Optionally, the step of performing logical consistency verification on multiple inference results based on preset logical consistency verification rules or constraints to detect whether there is a logical conflict between different inference results includes: extracting logical rules related to the procurement governance task from the logical consistency verification rules or constraints; mapping the inference results output by each of the business modules to the input conditions of the logical rules; performing inference on the logical rules to determine whether there is a conflict, mutual exclusion, or repeated execution in the result set composed of the inference results output by each of the business modules; if so, outputting a logical consistency verification result, wherein the logical consistency verification result includes the business modules that have conflicts, mutual exclusion, or repeated execution and the existence of corresponding target logical rules.

[0010] Optionally, mapping the inference results output by each of the business modules to the input conditions of the logical rules includes: parsing the rule node identifier on which each inference result is based, wherein the rule node identifier is associated with a lower-level business rule node in the rule tree corresponding to the business module that generated the inference result; determining the scope of application of the rule corresponding to the inference result based on the hierarchical position of the rule node identifier in the rule tree; and mapping the conclusion and parameters of the inference result to the input condition variables in the logical rules that match the scope of application of the rule, so as to perform logical rule inference.

[0011] Optionally, the step of performing a time-order consistency check on multiple inference results based on the generation timestamps of each inference result and the execution dependency order or path relationship between each business module to detect whether there is a mismatch between the result generation order and the dependency relationship or a situation where data dependencies are not satisfied but have already entered the execution of subsequent modules includes: obtaining the generation timestamps of the inference results output by each business module, and the execution order dependency graph of each business module determined by the cross-module scheduling mechanism of the procurement governance center; determining the execution dependency order or path relationship between each business module according to the execution order dependency graph of each business module; and determining the execution dependency order or path relationship between each business module according to the execution order dependency graph of each business module. The timestamp of the inference result generation and the execution dependency order or path relationship are used to verify whether the result generation order matches the dependency relationship; for business module pairs with dependency relationships, it is verified whether the inference result of the first executed module has included the data or conditions required by the subsequent executed module; it is verified whether the inference result of the subsequent executed module contradicts the inference result of the first executed module in terms of data or conditions; if any verification fails, it is determined that there is a time sequence inconsistency conflict, and the time sequence consistency verification result is output, wherein the time sequence consistency verification result includes the business module pairs with inconsistent time sequences, the time point of the time sequence inconsistency conflict, and the dependency relationship.

[0012] Optionally, when a conflict or inconsistency is detected in logical consistency verification or chronological consistency verification, a conflict handling mechanism is triggered, and conflict rollback and result correction are performed on the conflicting inference results to obtain a verified and corrected consistency governance result. This includes: when a conflict or inconsistency is detected in logical consistency verification or chronological consistency verification, determining the conflict handling method according to the conflict type and a preset conflict handling strategy, wherein the conflict handling method includes at least one of automatic rollback and re-execution, adjusting or correcting the inference results, and requesting manual review; if the conflict handling method is determined to be automatic rollback and re-execution, a re-execution instruction is sent to the business module that caused the conflict. The system is configured to cause the business module to regenerate the inference results through the artificial intelligence model. If the conflict resolution method is determined to be adjusting or correcting the inference results, the conflicting inference results are weighted and fused or selectively output based on preset priority rules and confidence weights. If the conflict resolution method is determined to be requesting manual review, a conflict details report is generated and sent to the manual review terminal to receive the result after manual review confirmation. Based on at least one of the following results: the inference results regenerated by the business module, the result of weighted fusion or selective output of the inference results, and the result after manual review confirmation, a verified and corrected consistency governance result is obtained.

[0013] Optionally, when the conflict type is a logical conflict and the conflict handling method is automatic rollback and re-execution, sending a re-execution instruction to the business module that generated the conflict, so that the business module can regenerate the inference result through the artificial intelligence model, includes: locating the target logical rule with the logical conflict based on the logical consistency verification result; backtracking to find the lower-level business rule node that defines the target logical rule and the middle-level rule boundary node associated with the lower-level business rule node based on the rule tree corresponding to multiple business modules; generating rule boundary constraints based on the constraints defined by the middle-level rule boundary node, and adding the rule boundary constraints to the re-execution instruction; and sending the re-execution instruction to the business module to which the lower-level business rule node belongs, so as to guide the artificial intelligence model of the business module to regenerate the inference result within the rule boundary constraints.

[0014] Optionally, after obtaining the inference results generated by the artificial intelligence models of multiple business modules in the centralized procurement governance hub based on the procurement governance task, the method further includes: obtaining metadata corresponding to each inference result, wherein the metadata includes the confidence score of the inference result generated by the artificial intelligence model, the source data identifier on which the inference result is based, and the rule node identifier that triggers the inference of the business module; and storing the inference results output by each business module and the metadata corresponding to each inference result in association.

[0015] Optionally, the step of integrating the inference results output by each of the business modules and the consistency governance results to obtain the procurement governance decision for the procurement governance task includes: obtaining the metadata corresponding to each of the inference results, and performing confidence-weighted fusion on the inference results that have passed logical consistency verification and time sequence consistency verification based on the confidence score in the metadata; and integrating the confidence-weighted fusion inference results and the verified and corrected consistency governance results to obtain the procurement governance decision for the procurement governance task.

[0016] Optionally, the method further includes: generating a consistency verification report based on the logical consistency verification results, time sequence consistency verification results, and conflict handling process information of the inference results output by each of the business modules; mapping and storing the procurement governance decision and the consistency verification report to generate an audit traceability index.

[0017] By employing the above technical solutions, this application provides a cross-module reasoning result consistency verification method for an AI-based centralized procurement governance hub. This method acquires reasoning results from multiple business modules as verification objects and uses preset logical consistency verification rules or constraints to perform logical consistency verification on different reasoning results. It also performs temporal sequence consistency verification on different reasoning results based on their generation timestamps and execution dependency order or path relationships. This effectively identifies potential logical contradictions and mutual exclusions between the reasoning conclusions of different modules, as well as contradictions in temporal and causal aspects. Furthermore, by automatically triggering a conflict handling mechanism when logical conflicts or temporal sequence inconsistencies are detected, timely conflict rollback and result correction can be performed on conflicting reasoning results. This ensures logical and temporal consistency among the reasoning results generated by multiple business modules, thereby ensuring the reliability and effectiveness of procurement governance decisions. By constructing logical consistency verification, temporal sequence consistency verification, and conflict rollback mechanisms in the centralized procurement governance hub, this method effectively improves the collaborative consistency of multi-module reasoning results in the procurement governance hub, thereby ensuring the reliability, auditability, and traceability of procurement governance decisions.

[0018] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 The illustration shows a flowchart of a cross-module reasoning result consistency verification method for an AI-based centralized procurement governance hub provided in an embodiment of this application. Detailed Implementation

[0020] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0021] In one embodiment, such as Figure 1 As shown, this paper presents a method for cross-module reasoning result consistency verification in an AI-based centralized procurement governance center. This method effectively improves the collaborative consistency of multi-module reasoning results within the centralized procurement governance center by constructing three core governance mechanisms: a logical consistency verification mechanism, a time sequence consistency verification mechanism, and a conflict rollback and result correction mechanism. This ensures the reliability, auditability, and traceability of procurement governance decisions. The following steps illustrate the application of this method to a centralized procurement governance center: Step 101: Obtain the reasoning results generated by the artificial intelligence models of multiple business modules in the centralized procurement governance center based on the procurement governance tasks. Each reasoning result includes decision-making suggestions from the business modules.

[0022] Among them, artificial intelligence models refer to large language models (LLM), computer vision models (CV), large multimodal models (LMMs), and small models formed through model distillation techniques, which are capable of handling text generation tasks and a certain amount of data.

[0023] Among them, procurement governance tasks refer to specific business requests that need to be processed by the centralized procurement governance center, such as conducting risk assessments on a procurement application or conducting compliance reviews on a supplier list; business modules refer to independent units within the centralized procurement governance center responsible for specific governance functions, such as supplier evaluation modules or cost analysis modules; decision recommendation information refers to the structured output results generated by the artificial intelligence models of each business module after analyzing specific aspects of procurement governance tasks under certain rule constraints, such as a list of recommended suppliers or identified price risk points.

[0024] Specifically, the procurement governance hub can capture the inference results completed and published by various business modules scheduled by the rule-driven cross-module scheduling mechanism by monitoring the internal message bus or querying the result database. When acquiring inference results, it can read the inference result data packets output by each business module and parse out the business module identifier, procurement governance task identifier, and decision suggestion information contained within the data. Alternatively, it can actively retrieve inference results by calling the application programming interfaces provided by each business module, or read the result files asynchronously written by each business module from shared storage.

[0025] Step 102: Based on preset logical consistency verification rules or constraints, perform logical consistency verification on multiple inference results to detect whether there are conflicts, mutual exclusions or duplicate executions between different inference results.

[0026] Among them, the preset logical consistency verification rules or constraints refer to several business rules used to determine whether there are contradictory relationships, mutually exclusive relationships or repeated execution situations between different inference results.

[0027] Specifically, during logical consistency verification, pre-set logical consistency verification rules or constraints can be loaded. These rules or constraints can be formed by fusing business rules extracted from the rule trees corresponding to each business module. Then, the decision suggestion information in the inference results output by each business module can be obtained and mapped to logical propositions that the verification rules in the rule set can identify. Next, these verification rules can be used to infer whether these combinations of logical propositions violate any consistency constraints defined in the logical consistency verification rules or constraints. For example, checking whether different inference results simultaneously recommend and reject a supplier. If a violation of any verification rule or constraint is detected, details of the logical conflict and the type of conflict are recorded. Alternatively, logical consistency verification can also be implemented through relational reasoning using knowledge graphs or automatic theorem proof using assertion logic, among other methods. This embodiment, by constructing a logical consistency verification mechanism in the centralized procurement governance center, can automatically identify potential contradictions in the inference results from different business modules, thereby preventing the overall decision from failing due to internal logical inconsistencies and ensuring the reliability of the final procurement governance decision.

[0028] Step 103: Based on the generation timestamps of each inference result and the execution dependency order or path relationship between each business module, perform time order consistency verification on multiple inference results to detect whether there is a mismatch between the result generation order and the dependency relationship or a situation where the data dependency is not satisfied but has already entered the execution of subsequent modules.

[0029] Among them, the generation timestamp refers to the precise time recorded by the system when the inference result is generated; the execution dependency order or path relationship describes the order in which each business module is scheduled to be executed and the data dependency relationship between each business module during the completion of the procurement governance task.

[0030] Specifically, when performing time sequence inconsistency checks, the generation timestamp of each inference result can first be extracted from its metadata. Then, the execution path diagram generated for the current procurement governance task is obtained from the rule-driven cross-module scheduling module. This execution path diagram defines the execution order of each business module and the dependencies or path relationships between them. Next, the sequence of generation timestamps and the execution path diagram are compared to check for any violations of dependencies, such as a module that should have been executed later generating a result earlier. Furthermore, it can be verified whether the inference results of later-executed modules are consistent with the outputs of their dependent, earlier-executed modules at the data level to verify the correctness of the inference result timing. Alternatively, time sequence consistency checks can also be implemented by maintaining a global event log or using a distributed transaction clock. This embodiment, by constructing a time sequence consistency check mechanism in the centralized procurement governance hub, ensures the correctness of cross-module inference in terms of process and data evolution, avoiding causal errors caused by out-of-order execution or data asynchronization, thereby guaranteeing the accuracy of the inference process and the reliability of the final procurement governance decision.

[0031] Step 104: When a conflict or inconsistency is detected in the logical consistency check or the time sequence consistency check, the conflict handling mechanism is triggered, and the conflicted inference results are rolled back and corrected to obtain the verified and corrected consistency governance results.

[0032] Among them, the conflict handling mechanism refers to a set of strategies and processes predefined by the centralized procurement governance center to resolve logical conflicts or time sequence inconsistencies; the consistency governance result refers to the new or adjusted reasoning result after processing, which eliminates logical conflicts or time sequence inconsistencies.

[0033] Specifically, when logical conflicts or temporal inconsistencies are detected between different inference results, the processing flow can be automatically triggered based on the conflict type and preset conflict handling strategies. For example, for logical conflicts, a re-execution instruction can be sent to the business module that caused the conflict. This instruction can carry stricter rule boundary constraints to guide the AI ​​model in the business module to recalculate and generate new inference results. For temporal inconsistencies, the relevant business modules can be triggered to reschedule the AI ​​model according to the correct temporal order and data dependencies to generate new inference results. Under certain strategies, conflict results can also be automatically selected or weighted and fused based on priority rules or confidence weights. For complex or high-risk conflicts, a conflict report can be generated and submitted for manual review. The manually confirmed result is then used as the verified and corrected consistency governance result. In addition, conflict handling can also be implemented using multi-round negotiation reasoning or game theory-based decision fusion methods. This embodiment can automatically resolve various conflicts by constructing a conflict rollback and result correction mechanism in the centralized procurement governance center. Furthermore, by handling conflicts through methods such as re-execution under rule constraints, weighted fusion, or manual review, inconsistent intermediate states can be guided to consistency, thereby ensuring the reliability of the reasoning results.

[0034] Step 105: Integrate the inference results and consistency governance results output by each business module to obtain the procurement governance decision for the procurement governance task.

[0035] Integrating the inference results and the consistency governance results refers to summarizing and synthesizing the outputs of all business modules that have been verified and processed according to the structural requirements of governance decisions; procurement governance decisions refer to the final, unified, and automated decision recommendations given by the centralized procurement governance center for procurement governance tasks.

[0036] Specifically, after collecting all unconflicted original inference results and verified and corrected consistency governance results, conclusions from different modules can be summarized into corresponding sections according to a predefined report template. This is based on decision-making recommendations, confidence scores, and other metadata from the results. Then, the conclusions can be weighted based on the confidence scores or presented comprehensively under the template logic, ultimately generating a structurally complete and consistent procurement governance decision. Alternatively, the integration process can be generated by a large model summarizing all results, or by a rule engine automatically merging and formatting the conclusions. Through these methods, the inference results from dispersed and multi-source business modules can be integrated into a logically consistent and comprehensive procurement governance decision, thereby supporting the final execution of procurement governance tasks.

[0037] The above methodology will now be illustrated with a specific example. Assume the procurement governance task is to evaluate a server procurement request. In this scenario, the system first obtains three inferences from the centralized procurement governance center: the cost analysis module's suggested negotiation, the supplier module's recommendation of supplier A, and the compliance module's warning of energy efficiency non-compliance. Then, logical consistency and temporal consistency checks are performed on the inferences from these three modules. Logical consistency checks reveal a logical conflict between the recommended procurement and the energy efficiency non-compliance; temporal consistency checks show that the compliance module's execution depends on the supplier module's output, but the compliance module's results are generated earlier than the supplier module's, resulting in a temporal conflict. Subsequently, conflict handling can be automatically triggered. First, the compliance module is required to re-execute after obtaining the latest supplier information to obtain a new compliance result. Simultaneously, the logical conflict is submitted for manual review. After manual assessment, it is confirmed that supplier A will not be recommended until the energy efficiency issue is resolved. Finally, the system integrates the original results from the supplier module, the original results from the cost module, and the manually reviewed compliance conclusion to generate a final procurement governance decision: supplier A is not recommended at this time, along with suggestions for negotiation and energy efficiency rectification.

[0038] The technical solution provided in the above embodiments, by acquiring the inference results of multiple business modules as verification objects and using preset logical consistency verification rules or constraints to perform logical consistency verification on different inference results, and by performing temporal sequence consistency verification on different inference results based on the generation timestamps and execution dependency order or path relationships of the inference results, can effectively identify potential logical contradictions and mutual exclusion relationships between the inference conclusions of different modules, as well as contradictions in the temporal and causal aspects of the inference conclusions of different modules. Furthermore, by automatically triggering a conflict handling mechanism when logical conflicts or temporal sequence inconsistencies are detected, timely conflict rollback and result correction processing can be performed on the inference results that generate conflicts, thereby ensuring the logical consistency and temporal sequence consistency among the inference results generated by multiple business modules, and thus ensuring the reliability of procurement governance decisions. The above method, by constructing a logical consistency verification mechanism, a temporal sequence consistency verification mechanism, and a conflict rollback mechanism in the centralized procurement governance center, can effectively improve the collaborative consistency of the inference results of multiple modules in the procurement governance center, thereby ensuring the reliability, auditability, and traceability of procurement governance decisions.

[0039] In one embodiment, the preset logical consistency verification rules or constraints in step 102 can be constructed by the following method: First, based on the business domain involved in the procurement governance task, determine the business modules related to the procurement governance task in the centralized procurement governance hub. Then, obtain the rule tree corresponding to each business module, wherein the rule tree contains upper-level governance target nodes, middle-level rule boundary nodes, and lower-level business rule nodes. Finally, extract all business logic rules defined by the lower-level business rule nodes in each rule tree, and merge the extracted business logic rules to obtain logical consistency verification rules or constraints.

[0040] In this embodiment, when constructing logical consistency verification rules or constraints, firstly, based on the business areas covered by the current procurement governance task, such as cost control, supplier management, or compliance review, all business modules related to these areas are located and determined in the centralized procurement governance hub. Next, for each determined business module, its corresponding rule tree structure is accessed. The rule tree includes upper-level governance target nodes defining the macro-level direction, middle-level rule boundary nodes specifying the feasible scope, and lower-level business rule nodes describing specific operations. By traversing each rule tree, all formally defined business logic rules can be extracted from the lower-level business rule nodes. Finally, the business logic rules extracted from the rule trees corresponding to different business modules are aggregated, deduplicated, and logically merged to obtain logical consistency verification rules or constraints covering the business logic of multiple modules.

[0041] This embodiment ensures the reliability of each logical rule in the constructed logical consistency verification rules or constraints by identifying the relevant business modules based on the areas involved in the procurement governance task and locating the source of the rules. Furthermore, by extracting the business logic rules defined by the lower-level business rule nodes from the rule trees pre-built in each business module, the traceability of the verification rules and the final decision can be ensured.

[0042] In one embodiment, step 102 can be implemented by the following method: First, extract the logical rules related to the procurement governance task from the preset logical consistency verification rules or constraints, and map the reasoning results output by each business module to the input conditions of the logical rules. Then, perform reasoning on the logical rules to determine whether there are conflicts, mutual exclusions, or repeated executions in the result set composed of the reasoning results output by each business module. If any of the above logical conflicts exist, output the logical consistency verification result. The logical consistency verification result includes the business modules that have conflicts, mutual exclusions, or repeated executions, as well as the target logical rules that have conflicts, mutual exclusions, or repeated executions.

[0043] In this embodiment, when performing logical consistency verification on different inference results, the logical rules related to the current procurement governance task can first be filtered and extracted from the existing logical consistency verification rules or constraints based on the characteristics of the current procurement governance task. Then, the inference results output by each business module are obtained, mainly the decision suggestion information in the inference results. These are then parsed and converted into logical assertions, and the converted logical assertions are mapped to the input condition variables of the extracted logical rules. After mapping, the rule inference engine can be invoked to perform joint inference on all input condition variables to determine whether there are conflicts, mutual exclusions, or duplicate executions in the overall factual state constituted by these inference results from different modules. If the rule inference engine determines that there are conflicts, mutual exclusions, or duplicate executions, it generates and outputs logical consistency verification results. These results describe which business modules' inference results conflict with each other, and which specific target logical rule(s) are violated.

[0044] This embodiment ensures the reliability of logical consistency verification by extracting logical rules related to procurement governance tasks from logical consistency verification rules or constraints. Furthermore, by describing the business modules involved in logical conflicts and their corresponding target logical rules in the logical consistency verification results, the root cause of the logical conflicts can be accurately located, thereby ensuring the auditability and traceability of procurement governance decisions.

[0045] In one embodiment, the inference results output by each business module can be mapped to the input conditions of the logical rules using the following method: First, the rule node identifier on which each inference result is based is parsed, wherein the rule node identifier is associated with a lower-level business rule node in the rule tree corresponding to the business module that generated the inference result. Then, the scope of the rule corresponding to the inference result is determined according to the hierarchical position of the rule node identifier in the rule tree. Finally, the conclusion and parameters of the inference result are mapped to the input condition variables in the logical rules that match the scope of the rule, so as to perform logical rule inference.

[0046] In this embodiment, when mapping the inference result to the input conditions of the logical rules, the metadata carried by the inference result output by each business module can first be parsed, and the rule node identifier used to generate the inference result can be extracted. This rule node identifier uniquely associates with a lower-level business rule node in the rule tree corresponding to that business module. Next, based on this rule node identifier, its complete hierarchical position in its rule tree can be queried, including the upper-level middle-level rule boundary nodes and upper-level governance target nodes, thereby precisely defining the applicable business scope and logical context of the inference result, i.e., its rule scope. Finally, based on the determined rule scope, logical rules matching the rule scope can be selected from the extracted set of logical rules, and the core conclusions and key parameters from the inference result can be assigned to the input condition variables corresponding to these logical rules. This completes the adaptation mapping from business results to abstract inference conditions, thus preparing data for subsequent rule inference.

[0047] For example, in a server procurement task, by parsing the inference results output by the supplier evaluation module, the rule node identifier upon which the inference results are based can be extracted, pointing to a lower-level business rule node named "Supplier Historical Performance Rate Evaluation" in the rule tree corresponding to the supplier module. Through querying, it can be determined that the scope of this lower-level business rule node is "Supplier Access Qualification Evaluation". Subsequently, rules belonging to this scope can be found in the logical rules related to this procurement governance task, such as "Only suppliers that pass the access evaluation can be recommended," and the conclusion in the inference result "Supplier A passes the evaluation" can be mapped to the input condition of this rule "Supplier A_Access Evaluation = Pass".

[0048] This embodiment traces the origin of the business by parsing the rule node identifiers on which the reasoning results are based, and defines the logical boundaries by determining the scope of the rules. It can accurately match business conclusions from different sources and at different granularities to the condition variables of the corresponding logical rules, thereby ensuring that the cross-module logical verification is performed in the correct semantic context, avoiding mismatches of irrelevant rules, and ensuring the accuracy of the verification.

[0049] In one embodiment, step 103 can be implemented by the following method: First, obtain the generation timestamp of the inference results output by each business module, and the execution order dependency graph of each business module determined by the cross-module scheduling mechanism of the procurement governance center. Based on the execution order dependency graph of each business module, determine the execution dependency order or path relationship between each business module. Then, based on the generation timestamp of each inference result and the execution dependency order or path relationship, verify whether the result generation order matches the dependency relationship. For business module pairs with dependency relationships, verify whether the inference result of the first executed module has included the data or conditions required by the subsequent executed module. Verify whether the inference result of the subsequent executed module contradicts the inference result of the first executed module in terms of data or conditions. If any verification fails, it is determined that there is a time sequence inconsistency conflict, and the time sequence consistency verification result is output. The time sequence consistency verification result includes the business module pairs with inconsistent time sequences, the time point of the time sequence inconsistency conflict, and the dependency relationship.

[0050] In this embodiment, when performing time sequence consistency checks on different inference results, the generation timestamp of the inference result can first be extracted from the metadata of the inference result output by each business module. Then, in the rule-driven cross-module scheduling module of the centralized procurement governance hub, the execution order dependency graph generated for the current procurement governance task is obtained and parsed to clarify the execution order of each business module in this task and the data dependencies between them, i.e., the execution dependency order or path relationship. Subsequently, the sequence of generation timestamps of each inference result can be compared and analyzed with the theoretical execution dependency order or path relationship to verify whether the result generation order matches the dependency relationship. At the same time, for business module pairs with dependencies, it is verified whether the inference result of the first executed module has contained the data or conditions required by the second executed module, and whether the inference result of the second executed module contradicts the inference result of the first executed module in terms of data or conditions. If any check fails, a time sequence inconsistency conflict is determined, and a time sequence consistency check result is generated. The time sequence consistency check result describes which pair of business modules has a time sequence problem, the specific time point when the time sequence inconsistency conflict occurs, and the violated dependency relationship.

[0051] This embodiment can detect causal logic errors caused by out-of-order module execution or untimely data synchronization by acquiring and comparing the generation timestamps of the inference results and the task execution dependencies. Furthermore, by outputting the time sequence consistency verification results, which include conflicting module pairs, time points, and dependencies, it can accurately identify cases of inconsistent time sequences and pinpoint the source of the time sequence conflict.

[0052] In one embodiment, step 104 can be implemented as follows: When a conflict or inconsistency is detected in logical consistency verification or time sequence consistency verification, a conflict handling method is determined according to the conflict type and a preset conflict handling strategy. The conflict handling method includes at least one of automatic rollback and re-execution, adjusting or correcting the inference result, and requesting manual review. If the conflict handling method is determined to be automatic rollback and re-execution, a re-execution instruction is sent to the business module that generated the conflict, so that the business module regenerates the inference result through the artificial intelligence model. If the conflict handling method is determined to be adjusting or correcting the inference result, the conflicting inference result is weighted and fused or selectively output based on preset priority rules and confidence weights. If the conflict handling method is determined to be requesting manual review, a conflict detail report is generated and sent to the manual review terminal to receive the result after manual review confirmation. Finally, a consistency governance result can be obtained based on at least one of the following: the inference result regenerated by the business module, the result of weighted fusion or selective output of the inference result, and the result after manual review confirmation.

[0053] In this embodiment, a conflict handling mechanism can be automatically triggered when logical conflicts or inconsistencies in time sequence are detected between different inference results. After triggering, the specific type of conflict is first analyzed, and a conflict handling method is determined based on a preset strategy configuration library. The available conflict handling methods include at least one of automatic rollback and re-execution, adjustment or correction of the inference results, and requesting manual review. If the processing method is determined to be automatic rollback and re-execution, a re-execution instruction can be sent to the business module identified as having a conflict. This instruction will trigger the business module to recalculate and generate a new inference result under its artificial intelligence model and related rule constraints. If the processing method is determined to be adjusting or correcting the inference result, the conflicting inference results are weighted and fused according to the preset module priority rules and the confidence weights attached to each inference result, or one of the results is directly selected and output according to priority. If automatic rollback and re-execution, adjustment or correction of the inference result are not suitable, the processing method can be determined to request manual review. In this case, an analysis report containing conflict details, involved business modules and the logical rules that caused the conflict can be automatically generated and sent to the manual review terminal to wait for authorized personnel to review and input the final decision result, and receive this result as the corrected output.

[0054] For example, in a server procurement task, when a logical conflict is detected between the recommendation conclusion output by the supplier module and the non-compliance conclusion output by the compliance module, according to the preset strategy, for such conflicts involving compliance bottom lines, the first step can be to determine the handling method as automatic rollback and re-execution, that is, to send an instruction to the compliance module, requiring it to re-execute based on the latest supplier information; if the conflict still exists after re-execution, the conflict handling method is switched to requesting manual review, and a report is generated for the procurement personnel to decide whether to make the final recommendation.

[0055] This embodiment selects conflict handling methods based on conflict type, enabling precise location and efficient processing of different types of conflicts. Furthermore, by setting up multi-layered conflict handling mechanisms such as automatic rollback and re-execution, adjustment or correction of inference results, and requesting manual review, the coordination and recovery capabilities of the centralized procurement governance center in the face of internal inconsistencies and conflicts can be significantly improved, thereby ensuring the reliability of the final decision.

[0056] In one embodiment, when the conflict type is a logical conflict and the conflict handling method is automatic rollback and re-execution, a re-execution instruction can be sent to the business module that caused the conflict in the following way, so that the business module can regenerate the inference result through the artificial intelligence model: First, based on the logical consistency verification result, the target logical rule with logical conflict is located. Then, based on the rule tree corresponding to multiple business modules, the lower-level business rule node that defines the target logical rule and the middle-level rule boundary node associated with the lower-level business rule node are traced back. Next, based on the constraint conditions defined by the middle-level rule boundary node, rule boundary constraints are generated and added to the re-execution instruction. Finally, the re-execution instruction is sent to the business module to which the lower-level business rule node belongs, so as to guide the artificial intelligence model of the business module to regenerate the inference result within the rule boundary constraints.

[0057] In this embodiment, when the conflict type is determined to be a logical conflict and the selected conflict handling method is automatic rollback and re-execution, the target logical rule(s) causing the contradiction between multiple inference results can be located based on the information provided by the logical consistency verification result. Then, a backtracking search can be performed according to the rule tree structure corresponding to each business module. First, the lower-level business rule node defining the target logical rule is located, and then the intermediate-level rule boundary node directly associated with and with which the applicable scope is defined is found. Next, the constraints defined by the intermediate-level rule boundary node can be parsed, and key limiting parameters and boundary values ​​can be extracted and encapsulated into a clear rule boundary constraint description. This rule boundary constraint can then be added as core guiding information to the re-execution instruction to be sent. Finally, a re-execution instruction can be sent to the business module to which the logical conflict originated, guiding the business module to generate new inference results strictly within the scope defined by the rule boundary constraints when calling its artificial intelligence model to re-perform inference.

[0058] For example, in a server procurement task, the logical consistency check results showed that the supplier module's "Recommended Supplier A" and the compliance module's "Equipment Energy Efficiency Does Not Meet Standards" violated the target logical rule "If the equipment's energy efficiency does not meet standards, it should not be recommended." By tracing back the rule tree, it was found that this rule originated from a lower-level business rule node named "Energy Efficiency Standard Compliance Judgment" in the compliance module's rule tree. Its upper-level node is associated with a middle-level rule boundary node that defines "The inspection scope is National Standard Level 1 and above." Based on this, the rule boundary constraint "Verification Standard: National Standard Level 1" was generated. Subsequently, a re-execution instruction was sent to the compliance review module. The re-execution instruction included this constraint to guide its artificial intelligence model to re-evaluate the energy efficiency of Supplier A's equipment under the premise of strictly adhering to the "National Standard Level 1" standard and generate inference results.

[0059] This embodiment, by tracing back to the source of the conflicting rules before re-execution and extracting its upper-level boundary constraints, ensures that the re-executed instructions are no longer simple repetitive calculations, but targeted calculations with clear and reinforced guidance. This ensures that the business module can be effectively constrained within the correct business scope during re-reasoning, thereby more effectively resolving discovered logical contradictions and improving the effectiveness of re-execution results.

[0060] In one embodiment, after step 101, the cross-module inference result consistency verification method of the AI-based centralized procurement governance center may further include the following steps: obtaining the metadata corresponding to each inference result, wherein the metadata includes the confidence score of the inference result generated by the artificial intelligence model, the source data identifier on which the inference result is based, and the rule node identifier that triggers the inference of the business module; and then, associating and storing the inference results output by each business module and the metadata corresponding to each inference result.

[0061] In this embodiment, after each business module outputs its inference results, while reading the decision suggestion information from each inference result, several key elements can be extracted from the metadata area attached to the inference result. These elements include the confidence score assigned by the AI ​​model when generating the inference result, the unique source data identifier of the original data on which the inference result is based, and the rule node identifier in the rule tree corresponding to the rule that triggered the business module to execute this inference. Subsequently, the extracted metadata elements can be bound to the corresponding inference results, and the relationship between these data can be established through database operations. These associated data records are then persistently stored in a dedicated data storage area.

[0062] This embodiment collects and associates key metadata such as confidence scores, source data identifiers, and rule node identifiers for each inference result, which can provide rich contextual information and quantitative basis for the consistency verification process and the final procurement governance decision. This can enhance the interpretability of the inference results and lay an effective data foundation for subsequent conflict analysis, weight fusion, and full-process audit traceability.

[0063] In one embodiment, step 105 can be implemented by: obtaining the metadata corresponding to each inference result, and performing a confidence-weighted fusion of the inference results that have passed the logical consistency check and the time sequence consistency check based on the confidence score in the metadata; then, integrating the confidence-weighted fusion of the inference results and the verified and corrected consistency governance results to obtain the procurement governance decision.

[0064] In this embodiment, when generating procurement governance decisions, the system first acquires all inference results that have passed logical consistency and temporal sequence consistency checks, along with their corresponding metadata. From the metadata, the confidence score given by the AI ​​model for each inference result is extracted. Subsequently, based on the integration logic defined in the report generation template, a confidence-weighted fusion calculation is performed on the inference conclusions from different business modules. For example, a weighted average is applied to numerical conclusions, while classification or recommendation conclusions are sorted or selected based on their confidence levels. After the weighted fusion is completed, the confidence-weighted fusion inference results and the verified and corrected consistency governance results are organized and arranged according to a preset report structure and format, ultimately generating the procurement governance decision for the procurement governance task.

[0065] For example, in a server procurement task, assuming the confidence level of the cost module's "Recommended Negotiation" is 0.9 and the confidence level of the supplier module's "Recommended Supplier A" is 0.8, and both have passed the consistency check, then when integrating to generate the final procurement governance decision, "Recommended Negotiation" and "Recommended Supplier A" can be included as strong recommendations in the core conclusions of the report based on the higher confidence level weight, and the report can indicate that these conclusions have high confidence support.

[0066] This embodiment utilizes the confidence scores of each reasoning result to perform confidence-weighted fusion of the verified reasoning results. This ensures logical consistency in the conclusions of procurement governance decisions and reflects the differences in the reliability of judgments from different business modules. As a result, the generated governance decision report can further improve the refinement of conclusions and the reference value for decision support on the basis of consistency.

[0067] In one embodiment, after step 105, the above method may further include the following steps: generating a consistency verification report based on the logical consistency verification results, time sequence consistency verification results, and conflict handling process information of the inference results output by each business module; then, mapping and storing the procurement governance decision and the consistency verification report to generate an audit traceability index.

[0068] In this embodiment, after generating the procurement governance decision, a consistency verification report of the inference results can be further generated. When generating the consistency verification report, the logical consistency verification results and chronological consistency verification results of each inference result are first summarized. Simultaneously, detailed process information recorded during conflict resolution is integrated, and this information is structured and formatted according to a preset report template to generate a complete consistency verification report. Subsequently, this newly generated consistency verification report can be logically associated with the procurement governance decision, and the two are mapped and bound through database operations. A unique audit traceability index identifier is generated for this set of associated reports. This index will serve as the key basis for subsequent queries, retrieval, and auditing of the entire governance task process.

[0069] This embodiment generates a consistency verification report for procurement governance decisions, maps and stores the two reports, and creates a unified index. This establishes a complete and independently searchable audit archive for each governance task, thereby effectively enhancing the auditability and traceability of the entire decision-making outcome.

[0070] It should be noted that the user information (including but not limited to device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the labels corresponding to each step in the above embodiments are only for identification purposes and are not intended to limit the order of execution of the steps. The order of execution of the steps in each embodiment can be set according to the actual situation.

[0071] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, graphics processors, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0072] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0073] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for verifying the consistency of cross-module inference results in a centralized procurement governance center based on AI, wherein the method is executed in the centralized procurement governance center, characterized in that, The method includes: The reasoning results generated by the artificial intelligence models of multiple business modules in the centralized procurement governance center based on procurement governance tasks are obtained, wherein each reasoning result includes decision suggestion information of the business module; Based on preset logical consistency verification rules or constraints, logical consistency verification is performed on multiple inference results to detect whether there are conflicts, mutual exclusions or duplicate executions between different inference results. Based on the generation timestamps of each inference result and the execution dependency order or path relationship between each business module, a time order consistency check is performed on multiple inference results to detect whether there is a mismatch between the generation order of the results and the dependency relationship or a situation where the data dependency is not satisfied but has already entered the execution of subsequent modules. When a conflict or inconsistency is detected in logical consistency verification or time sequence consistency verification, a conflict handling mechanism is triggered, and conflict rollback and result correction are performed on the conflicting inference results to obtain a verified and corrected consistency governance result. The inference results output by each of the business modules and the consistency governance results are integrated to obtain the procurement governance decision for the procurement governance task.

2. The method according to claim 1, characterized in that, The method for constructing the logical consistency verification rules or constraints includes: Based on the business areas involved in the procurement governance task, determine the business modules related to the procurement governance task in the centralized procurement governance center; Obtain the rule tree corresponding to each of the business modules, wherein the rule tree includes upper-layer governance target nodes, middle-layer rule boundary nodes, and lower-layer business rule nodes; Extract all business logic rules defined by lower-level business rule nodes in each rule tree, and merge the extracted business logic rules to obtain logical consistency verification rules or constraints.

3. The method according to claim 2, characterized in that, The step of performing logical consistency checks on multiple inference results based on preset logical consistency verification rules or constraints to detect whether there are conflicts, mutual exclusions, or duplicate executions between different inference results includes: Extract the logical rules related to the procurement governance task from the logical consistency verification rules or constraints; The inference results output by each of the aforementioned business modules are mapped to the input conditions of the logical rules; Reasoning is performed on the logical rules to determine whether there are conflicts, mutual exclusions or duplicate executions in the result set composed of the reasoning results output by each of the business modules. If they exist, the logical consistency verification result is output. The logical consistency verification result includes the business modules that have conflicts, mutual exclusions, or repeated executions, as well as the corresponding target logical rules.

4. The method according to claim 3, characterized in that, The step of mapping the reasoning results output by each of the business modules to the input conditions of the logical rules includes: Parse the rule node identifier on which each inference result is based, wherein the rule node identifier is associated with a lower-level business rule node in the rule tree corresponding to the business module that generated the inference result; The scope of the rule corresponding to the reasoning result is determined based on the hierarchical position of the rule node in the rule tree. The conclusions and parameters of the reasoning results are mapped to input condition variables in the logical rules that match the scope of the rules, in order to perform logical rule reasoning.

5. The method according to claim 1, characterized in that, Based on the generation timestamps of each inference result and the execution dependency order or path relationship between each business module, a time order consistency check is performed on multiple inference results to detect whether there are cases where the result generation order does not match the dependency relationship or where data dependencies are not satisfied but have already entered the execution of subsequent modules, including: Obtain the generation timestamp of the inference results output by each of the business modules, and the execution order dependency graph of each of the business modules determined by the cross-module scheduling mechanism of the procurement governance center; Based on the execution order dependency graph of each business module, determine the execution dependency order or path relationship between each business module; Based on the generation timestamps of each inference result and the execution dependency order or path relationship, verify whether the result generation order matches the dependency relationship; for business module pairs with dependency relationships, verify whether the inference result of the first executed module has already contained the data or conditions required by the subsequent executed module; verify whether the inference result of the subsequent executed module contradicts the inference result of the first executed module in terms of data or conditions. If any check fails, it is determined that there is a time sequence inconsistency conflict, and the time sequence consistency check result is output. The time sequence consistency check result includes the business module pair with inconsistent time sequence, as well as the time point and dependency relationship that caused the time sequence inconsistency conflict.

6. The method according to any one of claims 3 to 5, characterized in that, When a conflict or inconsistency is detected during logical consistency verification or chronological consistency verification, a conflict handling mechanism is triggered. This mechanism performs conflict rollback and result correction on the conflicting inference results, yielding a verified and corrected consistency governance result, including: When a conflict or inconsistency is detected in a logical consistency check or a time sequence consistency check, a conflict handling method is determined according to the conflict type and a preset conflict handling strategy. The conflict handling method includes at least one of automatic rollback and re-execution, adjustment or correction of the reasoning result, and requesting manual review. If the conflict resolution method is determined to be automatic rollback and re-execution, a re-execution instruction is sent to the business module that caused the conflict, so that the business module can regenerate the reasoning result through the artificial intelligence model. If the conflict resolution method is determined to be adjusting or correcting the reasoning results, then the conflicting reasoning results are weighted and fused or selectively output based on the preset priority rules and confidence weights. If the conflict resolution method is determined to be requesting manual review, a conflict details report is generated and sent to the manual review terminal to receive the result after manual review confirmation. Based on at least one of the following results: the inference results regenerated by the business module, the results of weighted fusion or selective output of the inference results, and the results after manual review and confirmation, a verified and corrected consistency governance result is obtained.

7. The method according to claim 6, characterized in that, When the conflict type is a logical conflict and the conflict handling method is automatic rollback and re-execution, sending a re-execution instruction to the business module that caused the conflict, so that the business module regenerates the inference result through the artificial intelligence model, includes: Based on the logical consistency verification results, the target logical rule with logical conflict is located. Based on the rule trees corresponding to the multiple business modules, the lower-level business rule nodes that define the target logic rules and the middle-level rule boundary nodes associated with the lower-level business rule nodes are traced back. Based on the constraints defined by the mid-level rule boundary nodes, rule boundary constraints are generated and added to the re-execution instruction. The re-execution instruction is sent to the business module to which the lower-level business rule node belongs, so as to guide the artificial intelligence model of the business module to regenerate the reasoning result within the rule boundary constraints.

8. The method according to claim 1, characterized in that, After obtaining the inference results generated by the artificial intelligence models of multiple business modules in the centralized procurement governance center based on the procurement governance tasks, the method further includes: Obtain the metadata corresponding to each of the inference results, wherein the metadata includes the confidence score of the inference result generated by the artificial intelligence model, the source data identifier on which the inference result is based, and the rule node identifier that triggers the inference of the business module; The inference results output by each of the aforementioned business modules and the corresponding metadata for each inference result are stored together.

9. The method according to claim 8, characterized in that, The process of integrating the inference results output by each of the business modules and the consistency governance results to obtain the procurement governance decision for the procurement governance task includes: Obtain the metadata corresponding to each of the inference results, and perform a confidence-weighted fusion of the inference results that have passed the logical consistency check and the temporal sequence consistency check based on the confidence score in the metadata; The procurement governance decision for the procurement governance task is obtained by integrating the reasoning results obtained through confidence-weighted fusion and the consistency governance results obtained after verification and correction.

10. The method according to claim 1, characterized in that, The method further includes: Based on the logical consistency verification results, temporal sequence consistency verification results, and conflict handling process information of the inference results output by each of the aforementioned business modules, a consistency verification report is generated. The procurement governance decisions and the consistency verification reports are mapped and stored to generate an audit traceability index.

Images