Unified semantic encoding cross-model business decision and self-explanation method and system

Abstract

This invention discloses a unified semantic coding method and system for cross-model business decision-making and self-explanation, belonging to the field of business decision modeling and intelligent decision-making. It addresses the technical problems of existing BPMN, UML, and state machine models, which suffer from different modeling standards, semantic isolation, and runtime fragmentation, leading to reliance on manual context piecing together for business decision generation and the inability to form interpretable decision chains. The system includes a unified semantic coding engine module, a ternary binding mapping module, a decision causal chain management module, and a visualization module. The method achieves semantic fusion and linked reasoning of process, data, and state models through a unified semantic ternary binding coding mechanism, while simultaneously achieving interpretability of the decision chain through automatic backtracking and visualization of the decision causal chain. This invention constructs a cross-model unified semantic operation mechanism, supporting forward impact propagation and reverse dependency tracing of decisions, thus solving the problem of uninterpretable AI decision-making.

Description

Technical Field

[0001] This invention relates to the field of business decision modeling and intelligent decision-making, and in particular to a cross-model business decision-making and self-explanatory method and system with unified semantic coding. Background Technology

[0002] In the business decision-making process of finance, power grid, state-owned enterprises and other fields, BPMN (Business Process Modeling and Annotation), UML (Unified Modeling Language) and state machine models are often used for business modeling and process design. However, the various models in the existing technologies have technical problems such as different modeling standards, semantic isolation and runtime fragmentation. As a result, the context information of various models needs to be manually pieced together in the business decision generation process, which not only reduces the efficiency of business decision generation, but also fails to form an interpretable decision link.

[0003] Meanwhile, most existing AI decision-making models are black-box models, with poor interpretability of the decision-making process, which cannot meet the strict audit compliance requirements of fields such as finance, power grids, and state-owned enterprises. Therefore, there is a need for a cross-model business decision-making and self-explanatory method and system that can achieve cross-model semantic fusion, automatically generate business decisions, and form an interpretable decision-making chain. Summary of the Invention

[0004] Purpose of the invention: The invention provides a cross-model business decision-making and self-explanatory method and system with unified semantic coding, which solves the problems mentioned in the background art.

[0005] Technical Solution: To solve the above-mentioned technical problems, according to one aspect of the present invention, more specifically, a cross-model business decision-making and self-explanatory system with unified semantic coding, comprising: The unified semantic coding engine module is used to generate unique semantic codes for process nodes, data objects, and status nodes, and store the coding binding relationship in the form of a computable graph structure; the unique semantic code includes at least a business domain identifier, an object type identifier, a behavior / status role identifier, and a lifecycle sequence number. The ternary binding mapping module communicates with the unified semantic encoding engine module and is used to construct a process-data-state ternary binding structure. When any model node changes, the change is mapped to the other two types of model nodes through semantic encoding, and a linkage reasoning path is automatically formed. The linkage reasoning path supports forward influence propagation and reverse dependency tracing. The decision causal chain management module communicates with the ternary binding mapping module and is used to generate a decision causal chain snapshot for each business decision, and supports reverse reconstruction of decisions based on the decision causal chain snapshot; the decision causal chain snapshot records the trigger data encoding, the encoding of the participating process nodes, the state transition path, and the version of the rule / model called; The visualization module, which communicates with the decision causal chain management module, is used to visualize the decision causal chain snapshot and the reverse-reconstructed decision link.

[0006] Furthermore, the unified semantic coding engine module includes a coding generation unit and a computable graph storage unit; the coding generation unit is used to generate unique semantic codes for process nodes of the BPMN model, data objects of the UML model, and state nodes of the state machine model according to preset coding rules; the computable graph storage unit is used to convert the coding binding relationship of process-data-state into a computable graph structure and store it, wherein the nodes of the computable graph structure are each semantic code, and the edges are the binding association relationship between codes.

[0007] Furthermore, the ternary binding mapping module includes a node change monitoring unit and a mapping inference unit; the node change monitoring unit is used to monitor the attribute or state changes of process nodes, data objects, and state nodes in real time; the mapping inference unit is used to match and map the associated nodes of the other two types of models in the computable graph structure according to the semantic code corresponding to the monitored node changes, and at the same time generate a linkage inference path from the changed node to the associated node according to the coding binding relationship.

[0008] Furthermore, the decision causal chain management module includes a snapshot generation unit and a reverse reconstruction unit. The snapshot generation unit is used to collect the trigger data encoding, participating process node encoding, state transition path, and called rule / model version in real time when a business decision is generated, and generate a decision causal chain snapshot according to a preset format. The reverse reconstruction unit is used to reverse deduce the data source changes, state changes, and process conditions corresponding to the decision based on the decision causal chain snapshot corresponding to the target decision, according to the user's decision tracing request, forming a reverse reconstruction link of "decision ← data change ← state change ← process condition".

[0009] Furthermore, the visualization module includes a link rendering unit and an interactive display unit; the link rendering unit is used to transform the decision causal chain snapshot and reverse reconstruction link into a visualized graphical link, including the visualization rendering of node identification, correlation relationship and flow direction; the interactive display unit is used to provide human-computer interaction functions for the visualized link, supporting users to query the details of link nodes, expand the link step by step and perform backtracking operations.

[0010] According to one aspect of the present invention, and more specifically, a cross-model business decision-making and self-explanatory method based on unified semantic coding, using the cross-model business decision-making and self-explanatory system based on unified semantic coding as described, the method includes the following steps: S1. Unified semantic code generation: The unified semantic code engine module generates unique semantic codes for the process nodes of the BPMN model, the data objects of the UML model, and the state nodes of the state machine model. The unique semantic code includes at least the business domain identifier, the object type identifier, the behavior / state role identifier, and the lifecycle sequence number. S2. Construction and storage of ternary binding structure: The process-data-state ternary binding structure is constructed through the ternary binding mapping module, and the coded binding relationship of process-data-state is stored in the form of a computable graph structure. The computable graph structure supports forward influence propagation and reverse dependency tracing. S3. Cross-model linkage reasoning: When a change is detected in any model node, the node change is mapped to the associated nodes of the other two types of models through semantic encoding, and a linkage reasoning path is automatically formed. Based on the linkage reasoning path, forward influence propagation or reverse dependency tracing across models is realized, providing full contextual information for business decision generation. S4. Decision Causal Chain Snapshot Generation: When a business decision is generated, the decision causal chain management module collects the trigger data code, the code of the participating process nodes, the state transition path, and the rule / model version called for the decision, and automatically generates and stores the decision causal chain snapshot. S5. Decision Reverse Reconstruction and Visualization: Based on the user's decision tracing request, and using the snapshot of the decision causal chain of the target decision as a basis, the reverse reconstruction link of "decision ← data change ← state change ← process conditions" is derived in reverse. Then, the decision causal chain snapshot and the reverse reconstruction link are graphically visualized through the visualization module.

[0011] Furthermore, in step S1, the generation rule for the unique semantic code is as follows: the code is encoded according to the hierarchical structure of business domain identifier → object type identifier → behavior / state role identifier → lifecycle sequence number. Each level identifier uses a combination of characters or numbers of a preset length, and the semantic code unique identifier within the same business domain corresponds to the process node, data object, or state node.

[0012] Furthermore, in step S2, the specific process of constructing the process-data-state ternary binding structure is as follows: establishing the association relationship between the semantic encoding of process nodes and the semantic encoding of data objects, the association relationship between the semantic encoding of data objects and the semantic encoding of state nodes, and the association relationship between the semantic encoding of process nodes and the semantic encoding of state nodes, so that the three types of models of process, data, and state form a ternary structure that is mutually bound and mutually related through semantic encoding.

[0013] Furthermore, in step S3, the specific process of cross-model linkage reasoning is as follows: S31, real-time monitoring of the attribute or state changes of process nodes, data objects, and state nodes to determine the target node that has changed and its corresponding target semantic code; S32, matching the target semantic code to obtain the association semantic codes of the other two types of models in the computable graph structure, and determining the association node; S33, generating a linkage reasoning path from the target node to the association node based on the binding relationship between the target semantic code and the association semantic code, and realizing forward influence propagation or reverse dependency tracing based on the path.

[0014] Furthermore, in step S5, the specific process of decision reverse reconstruction is as follows: starting from the snapshot of the decision causal chain of the target decision, the data source change corresponding to the trigger data code of the decision, the process conditions corresponding to the process node code, and the state change corresponding to the state transition path are traced in sequence to form a complete decision reverse reconstruction link and realize the interpretability traceability of the decision.

[0015] Beneficial effects: (1) By using a unified semantic coding mechanism, unique semantic codes are generated for BPMN process nodes, UML data objects, and state machine state nodes, and a process-data-state ternary binding structure is constructed, breaking down the semantic barriers and runtime fragmentation between different models, realizing the linkage reasoning of the three types of models, eliminating the need for manual splicing of context information, and fundamentally eliminating cross-model information silos in business decision-making. Leveraging the forward influence propagation and backward dependency tracing capabilities of the computable graph structure, when any model node changes, it can automatically match related nodes and generate linked reasoning paths, providing comprehensive and accurate cross-model context information for business decisions. This replaces the cumbersome process of manually integrating multi-model data, significantly improving the efficiency of business decision generation while reducing information bias caused by manual operations and enhancing decision accuracy.

[0016] (2) This invention uses decision causal chain snapshot generation and reverse reconstruction technology to completely record the trigger data, participating process nodes, state transition paths, and calling rules / model versions of each business decision. It can also reverse deduce the complete chain of "decision ← data change ← state change ← process conditions" according to user requests, transforming the black-box AI decision into a transparent decision that can be clearly traced and logically verified, so that the reasons for the decision can be found. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] To make the technical solution of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Reference Figure 1 The unified semantic coding cross-model business decision-making and self-explanatory system includes a unified semantic coding engine module, a ternary binding mapping module, a decision causal chain management module, and a visualization module. Each module realizes data interaction based on a distributed communication architecture and is deployed on a cloud server or a local server, supporting business decision-making scenarios in fields such as finance, power grid, and state-owned enterprises.

[0020] The unified semantic coding engine module includes a coding generation unit and a computable graph storage unit. The coding generation unit generates unique semantic codes for process nodes in the BPMN model, data objects in the UML model, and state nodes in the state machine model, according to the rule of "business domain identifier (2 digits) + object type identifier (1 letter, L for process node, D for data object, S for state node) + behavior / state role identifier (3 digits) + lifecycle sequence number (4 digits)". For example, the semantic code for a customer information data object in the financial business domain is 01D0010001. The computable graph storage unit transforms the coding binding relationship between process, data, and state into a computable graph structure. The nodes of the computable graph are the semantic codes, and the edges represent the binding relationships between codes. It uses a graph database for storage, supporting efficient node matching and link retrieval.

[0021] The ternary binding mapping module includes a node change monitoring unit and a mapping inference unit. The node change monitoring unit collects node attributes and state changes in BPMN, UML, and state machine models in real time through an interface, such as changes in the execution state of process nodes, changes in the values ​​of data objects, and changes in the migration of state nodes. After detecting node changes, the mapping inference unit matches the associated semantic codes in the computable graph structure to generate a linkage inference path.

[0022] For example, when the status of the "Loan Application Submission" process node (code 01L0020003) in the financial business is detected to change to "Submitted", the associated "Loan Application Information" data object (code 01D0030005) and "Application Pending Review" status node (code 01S0010002) are matched through the computable graph, and an automatic linkage reasoning path from the process node to the data object and status node is formed.

[0023] The decision causal chain management module includes a snapshot generation unit and a reverse reconstruction unit. When a business decision is generated (e.g., loan approval / rejection decision), the snapshot generation unit collects in real-time the trigger data code (e.g., applicant credit data code 01D0050008), the participating process node codes (e.g., document review 01L0040006, credit check 01L0050007), the status transition path (e.g., application pending review → credit check in progress → review completed), and the called rule / model version (e.g., loan approval rule V2.1). It then generates a decision causal chain snapshot in JSON format and stores it in the database. The reverse reconstruction unit, based on the user's traceability request and the snapshot of the target decision, reverse-engineers the changes in the decision's data source (e.g., overdue credit data), process conditions (e.g., document review failed), and status changes (e.g., credit check failed), forming a complete reverse reconstruction chain.

[0024] The visualization module includes a link rendering unit and an interactive display unit. The link rendering unit transforms the decision causal chain snapshot and reverse reconstruction link into a visualized directed graph, where nodes are labeled with semantic encoding and node names, and edges are labeled with relationships and flow directions. The interactive display unit provides a human-computer interaction interface for web and client applications, allowing users to click on nodes to query details (such as the specific values ​​of data objects, the execution time of process nodes), expand the link step by step, and perform backtracking operations, thus realizing the visualization of the decision link.

[0025] refer to Figure 1 A unified semantic coding cross-model business decision-making and self-explanatory method, applied to the above system, includes the following steps: S1. Unified Semantic Code Generation: Through the coding generation unit of the unified semantic coding engine module, unique semantic codes are generated for all process nodes of the BPMN model, all data objects of the UML model, and all state nodes of the state machine model. The codes include business domain identifiers, object type identifiers, behavior / state role identifiers, and lifecycle sequence numbers, and the codes are unique within the same business domain. S2. Construction and storage of ternary binding structure: The semantic coding relationship between process nodes, data objects and state nodes is established through the ternary binding mapping module to build a process-data-state ternary binding structure. Then, the structure is transformed into a computable graph structure and stored in the graph database through the computable graph storage unit. This computable graph supports forward influence propagation and reverse dependency tracing. S3, Cross-model Linked Reasoning: The node change monitoring unit monitors the changes of various model nodes in real time, determines the target node and target semantic code, and the mapping reasoning unit matches the associated semantic code and associated node in the computable graph to generate a linked reasoning path. Based on this path, it provides full cross-model context information for business decision generation without the need for manual splicing. S4. Decision Causal Chain Snapshot Generation: When business decisions are generated, the snapshot generation unit collects the trigger data codes, participating process node codes, state transition paths, and called rule / model versions related to the decision, automatically generates and stores the decision causal chain snapshot; S5. Decision Reverse Reconstruction and Visualization: Based on the user's decision tracing request, the reverse reconstruction unit starts with a snapshot of the target decision and reversely derives the reverse reconstruction link of "decision ← data change ← state change ← process conditions". The link rendering unit renders the snapshot and the reverse reconstruction link graphically, and the interactive display unit displays the visualized link to the user, realizing the interpretable traceability of the decision link.

[0026] In the financial loan approval scenario of this embodiment, the method and system of this invention realize cross-model semantic fusion of BPMN process model, UML data model and state machine model. The generation of loan approval decisions does not require manual splicing of context, and the reasons for the formation of approval decisions can be clearly traced through decision causal chain snapshots and reverse reconstruction. For example, loan rejection decisions can be traced back to "applicant's credit data is overdue" (data change) → "credit check status is not passed" (status change) → "credit check process node not passed" (process condition), realizing the full-link interpretability of decisions and meeting the audit compliance requirements of the financial field.

[0027] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A cross-model business decision-making and self-explanatory system with unified semantic coding, characterized in that, include: The unified semantic coding engine module is used to generate unique semantic codes for process nodes, data objects, and status nodes, and store the coding binding relationship in the form of a computable graph structure; the unique semantic code includes at least a business domain identifier, an object type identifier, a behavior / status role identifier, and a lifecycle sequence number. The ternary binding mapping module communicates with the unified semantic encoding engine module and is used to construct a process-data-state ternary binding structure. When any model node changes, the change is mapped to the other two types of model nodes through semantic encoding, and a linkage reasoning path is automatically formed. The linkage reasoning path supports forward influence propagation and reverse dependency tracing. The decision causal chain management module communicates with the ternary binding mapping module and is used to generate a decision causal chain snapshot for each business decision, and supports reverse reconstruction of decisions based on the decision causal chain snapshot; the decision causal chain snapshot records the trigger data encoding, the encoding of the participating process nodes, the state transition path, and the version of the rule / model called; The visualization module is connected in communication with the decision causal chain management module and is used to visualize the decision causal chain snapshot and the reverse reconstructed decision link.

2. The cross-model business decision-making and self-explanatory system with unified semantic coding according to claim 1, characterized in that: The unified semantic coding engine module includes a coding generation unit and a computable graph storage unit. The coding generation unit is used to generate unique semantic codes for process nodes of the BPMN model, data objects of the UML model, and state nodes of the state machine model according to preset coding rules. The computable graph storage unit is used to convert the coding binding relationship of process-data-state into a computable graph structure and store it. The nodes of the computable graph structure are the semantic codes, and the edges are the binding relationships between the codes.

3. The cross-model business decision-making and self-explanatory system with unified semantic coding according to claim 1, characterized in that: The ternary binding mapping module includes a node change monitoring unit and a mapping inference unit. The node change monitoring unit is used to monitor the attribute or state changes of process nodes, data objects, and state nodes in real time. The mapping inference unit is used to match and map the semantic codes corresponding to the monitored node changes to the associated nodes of the other two types of models in the computable graph structure, and generate a linkage inference path from the changed node to the associated node based on the coding binding relationship.

4. The cross-model business decision-making and self-explanatory system with unified semantic coding according to claim 1, characterized in that: The decision causal chain management module includes a snapshot generation unit and a reverse reconstruction unit. The snapshot generation unit is used to collect the trigger data codes, participating process node codes, state transition paths, and called rule / model versions related to the decision in real time when the business decision is generated, and generate a decision causal chain snapshot according to a preset format. The reverse reconstruction unit is used to deduce the data source changes, state changes, and process conditions corresponding to the decision based on the decision causal chain snapshot corresponding to the target decision according to the user's decision tracing request, forming a reverse reconstruction link of "decision ← data change ← state change ← process conditions".

5. The cross-model business decision-making and self-explanatory system with unified semantic coding according to claim 1, characterized in that: The visualization module includes a link rendering unit and an interactive display unit. The link rendering unit is used to transform the decision causal chain snapshot and reverse reconstruction link into a visualized graphical link, including the visualization rendering of node identification, correlation, and flow direction. The interactive display unit is used to provide human-computer interaction functions for the visualized link, supporting users to query the details of link nodes, expand the link step by step, and perform backtracking operations.

6. A cross-model business decision-making and self-explanatory method based on unified semantic coding, using the cross-model business decision-making and self-explanatory system based on unified semantic coding as described in any one of claims 1-5, characterized in that, Includes the following steps: S1. Unified semantic code generation: The unified semantic code engine module generates unique semantic codes for the process nodes of the BPMN model, the data objects of the UML model, and the state nodes of the state machine model. The unique semantic code includes at least the business domain identifier, the object type identifier, the behavior / state role identifier, and the lifecycle sequence number. S2. Construction and storage of ternary binding structure: The process-data-state ternary binding structure is constructed through the ternary binding mapping module, and the coded binding relationship of process-data-state is stored in the form of a computable graph structure. The computable graph structure supports forward influence propagation and reverse dependency tracing. S3. Cross-model linkage reasoning: When a change is detected in any model node, the node change is mapped to the associated nodes of the other two types of models through semantic encoding, and a linkage reasoning path is automatically formed. Based on the linkage reasoning path, forward influence propagation or reverse dependency tracing across models is realized, providing full contextual information for business decision generation. S4. Decision Causal Chain Snapshot Generation: When a business decision is generated, the decision causal chain management module collects the trigger data code, the code of the participating process nodes, the state transition path, and the rule / model version called for the decision, and automatically generates and stores the decision causal chain snapshot. S5. Decision Reverse Reconstruction and Visualization: Based on the user's decision tracing request, and using the snapshot of the decision causal chain of the target decision as a basis, the reverse reconstruction link of "decision ← data change ← state change ← process conditions" is derived in reverse. Then, the decision causal chain snapshot and the reverse reconstruction link are graphically visualized through the visualization module.

7. The cross-model business decision-making and self-explanatory method with unified semantic coding according to claim 6, characterized in that: In step S1, the generation rule for the unique semantic code is as follows: the code is encoded according to the hierarchical structure of business domain identifier → object type identifier → behavior / status role identifier → lifecycle sequence number. Each level identifier uses a combination of characters or numbers of a preset length, and the semantic code uniquely identifies the process node, data object or status node within the same business domain.

8. The cross-model business decision-making and self-explanatory method with unified semantic coding according to claim 6, characterized in that: In step S2, the specific process of constructing the process-data-state ternary binding structure is as follows: establishing the association relationship between the semantic encoding of process nodes and the semantic encoding of data objects, the association relationship between the semantic encoding of data objects and the semantic encoding of state nodes, and the association relationship between the semantic encoding of process nodes and the semantic encoding of state nodes, so that the three types of models of process, data and state form a ternary structure that is mutually bound and mutually related through semantic encoding.

9. The cross-model business decision-making and self-explanatory method with unified semantic coding according to claim 6, characterized in that: In step S3, the specific process of cross-model linkage reasoning is as follows: S31. Monitor the changes in attributes or status of process nodes, data objects, and status nodes in real time, and determine the target node that has changed and its corresponding target semantic code. S32. In the computable graph structure, obtain the associated semantic codes of the other two types of models based on the target semantic code matching, and determine the associated nodes; S33. Based on the binding relationship between the target semantic encoding and the associated semantic encoding, generate a linkage reasoning path from the target node to the associated node, and realize forward influence propagation or reverse dependency tracing based on the path.

10. The cross-model business decision-making and self-explanatory method with unified semantic coding according to claim 6, characterized in that: In step S5, the specific process of decision reverse reconstruction is as follows: starting from the snapshot of the decision causal chain of the target decision, the data source change corresponding to the trigger data code of the decision, the process conditions corresponding to the process node code, and the state change corresponding to the state transition path are traced in sequence to form a complete decision reverse reconstruction link and realize the interpretability traceability of the decision.

Images