Intelligent processing method, device and system applied to cross-department decision-making

Abstract

The application relates to the technical field of cross-department decision analysis, and discloses an intelligent processing method, device and system applied to cross-department decision, which triggers cross-department decision demand through natural language, accurately captures and efficiently disassembles the decision demand, obtains two types of decision tasks of structured query and semantic retrieval, and respectively performs structured decision case retrieval and unstructured decision document retrieval on a knowledge decision graph based on the structured query decision task and the semantic retrieval decision task, so that the information dispersion and retrieval complexity in cross-department decision are reduced, the retrieval efficiency and accuracy of the structured decision case and the unstructured decision document are improved, the structured decision case and the decision document are fused to obtain a structured decision information package, the information fragmentation is integrated, the generation efficiency and accuracy of the structured decision information package are improved, and therefore the efficiency and accuracy of cross-department decision are improved.

Description

Technical Field

[0001] This invention relates to the field of cross-departmental decision analysis technology, and in particular to an intelligent processing method, apparatus and system for cross-departmental decision making. Background Technology

[0002] In modern enterprise management, cross-departmental decision-making is a core element in ensuring efficient operation and achieving strategic goals. The quality and efficiency of cross-departmental decisions directly impact an enterprise's market competitiveness and sustainable development capabilities. As enterprises expand and business scenarios become more complex, cross-departmental decision-making often requires integrating multi-dimensional information such as business data, historical cases, and rules and regulations from multiple departments to obtain accurate decisions.

[0003] Currently, when cross-departmental decision integration is required, decision-makers often need to manually retrieve data sources from various departments, which is time-consuming and labor-intensive, and makes it difficult to quickly uncover potential relationships between data, resulting in low decision-making efficiency and accuracy. Furthermore, the knowledge assets within enterprises are fragmented and dispersed; business data, decision cases, rules and regulations from various departments are mostly stored in independent databases or document systems, lacking unified integration and correlation, forming data silos, which further reduces the efficiency and accuracy of cross-departmental decision-making. Therefore, proposing a new cross-departmental decision analysis method to improve the efficiency and accuracy of cross-departmental decision-making, thereby improving the efficiency and accuracy of inter-departmental collaboration, and ultimately enhancing the enterprise's management efficiency, security, and accuracy, is particularly important. Summary of the Invention

[0004] This invention provides an intelligent processing method, device, and system for cross-departmental decision-making, which can improve the efficiency and accuracy of cross-departmental decision-making, thereby improving the efficiency and accuracy of collaboration between departments, and ultimately improving the management efficiency, security, and accuracy of enterprises.

[0005] The first aspect of this invention discloses an intelligent processing method applied to cross-departmental decision-making, the method comprising: The system detects cross-departmental decision-making needs triggered by target users through natural language descriptions in the current decision-making scenario. Based on a preset decision analysis model, the cross-departmental decision requirements are subjected to intent parsing and problem decomposition operations to obtain a first decision task and a second decision task corresponding to the cross-departmental decision requirements; wherein, the first decision task is used to perform a structured query operation, and the second decision task is used to perform a semantic retrieval operation; Based on the first decision task, a graph path retrieval operation is performed on the pre-constructed knowledge decision graph to obtain a structured decision case that matches the first decision task; Based on the second decision task, a vector semantic retrieval operation is performed on the knowledge decision graph to obtain a decision document that matches the second decision task; A data fusion operation is performed on the structured decision case and the decision document to obtain a structured decision information package that matches the current dialogue of the cross-departmental decision-making needs. The structured decision information package is used as the basis for the target user's decision in the current decision scenario.

[0006] As an optional implementation, in a first aspect of the present invention, the step of performing a graph path retrieval operation on a pre-constructed knowledge decision graph based on the first decision task to obtain a structured decision case matching the first decision task includes: Perform graph node analysis on the first decision task to obtain all decision nodes of the first decision task, and analyze the node association relationship between all decision nodes based on the pre-constructed knowledge decision graph. Based on the node association relationship, a central decision node is determined from all the decision nodes, and the central decision node is used as the node center. Based on the node association relationship, a graph retrieval path for the first decision task is generated. Based on the graph retrieval path of the first decision task, a graph path retrieval operation is performed on the knowledge decision graph to obtain a structured decision case that matches the first decision task.

[0007] As an optional implementation, in a first aspect of the present invention, the step of performing a vector semantic retrieval operation on the knowledge decision graph according to the second decision task to obtain a decision document matching the second decision task includes: The second decision task is vectorized to obtain the third decision task, and the decision filtering parameters of the third decision task are generated. According to the third decision task, a vector semantic retrieval operation is performed on the knowledge decision graph to obtain the decision node parameters that match the third decision task, and according to the third decision task, the first decision document that matches the third decision task is selected from the pre-determined document decision database. The first decision document is correlated with the decision node parameters corresponding to the third decision task to obtain a second decision document that matches the decision node parameters. Based on the decision filtering parameters, a filtering operation is performed on the second decision document to obtain a decision document that matches the second decision task.

[0008] As an optional implementation, in a first aspect of the invention, the data fusion operation performed on the structured decision case and the decision document to obtain a structured decision information package matching the current dialogue of the cross-departmental decision-making needs includes: For any target decision node in the structured decision case, in the decision documents, filter the decision documents associated with the target decision node; wherein, all decision nodes include all target decision nodes; Analyze the importance of each target decision node to the current decision scenario and the correlation strength among all target decision nodes; Based on the decision importance of each target decision node and the correlation strength between all target decision nodes, the decision documents corresponding to each target decision node are sorted to obtain document sorting parameters. Based on the document sorting parameters, a corresponding case-document association relationship is established for each target decision node and its associated first decision document. Analyze the decision documents to obtain multiple document decision parameters, and set decision basis parameters for the structured decision cases based on all the document decision parameters; obtain the data decision sources for the structured decision cases and the data decision sources for each decision document; Based on the case-document relationship, the structured decision cases, all the decision basis parameters, and all the data decision sources, a structured decision information package matching the current dialogue of the cross-departmental decision-making needs is generated.

[0009] As an optional implementation, in a first aspect of the present invention, the method further includes: Based on the cross-departmental decision-making requirements, a short-term memory cache database and a long-term memory database corresponding to the current session of the cross-departmental decision-making requirements are obtained; wherein, the short-term memory cache database stores decision parameters corresponding to multiple rounds of dialogue that already exist for the cross-departmental decision-making requirements; the long-term memory database stores a historical structured decision case set and business parameters corresponding to each historical structured decision case in the historical structured decision case set; Determine whether there is scenario decision data matching the target decision task in the short-term memory cache database and the long-term memory database; wherein, the target decision task includes the first decision task and the second decision task; When it is determined that the scenario decision data does not exist, the operation of performing a graph path retrieval operation on the pre-constructed knowledge decision graph based on the first decision task is executed to obtain a structured decision case that matches the first decision task. When the existence of the scenario decision data is determined, the scenario decision data is identified from the short-term memory cache database and the long-term memory database, and a structured decision information package matching the current dialogue of the cross-department decision requirement is determined from the scenario decision data based on the first decision task and the second decision task.

[0010] As an optional implementation, in a first aspect of the present invention, the method further includes: Based on the structured decision information package, all intermediate decision path nodes are determined from the detection of the cross-departmental decision requirement to the generation of the structured decision information package; wherein each intermediate decision path node has a corresponding decision result; wherein the decision result of the intermediate decision path node that occurs earlier in two related intermediate decision path nodes is the input parameter of the intermediate decision path node that occurs later. Determine the decision confidence level for each of the intermediate decision path nodes; The decision root path node is the decision reasoning path diagram required by the target user for detecting the cross-departmental decision-making needs, and the decision end path node is the decision reasoning path diagram generated by generating the structured decision information package. According to the decision path generation parameters, the decision root path node, all the intermediate decision path nodes and the decision end path node are used to generate a multi-turn dialogue decision reasoning path for the cross-departmental decision-making needs. Based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node, a reasoning decision path diagram for multi-round dialogues targeting the cross-departmental decision-making needs is generated.

[0011] As an optional implementation, in a first aspect of the present invention, determining the decision confidence level of each of the intermediate decision path nodes includes: For any of the intermediate decision path nodes, the data decision source of the intermediate decision path node is determined according to the decision parameters of the intermediate decision path node, wherein the decision parameters of each intermediate decision path node include the structured decision case and / or the decision document; Based on the data decision sources and decision parameters of the intermediate decision path nodes, the decision confidence level of the intermediate decision path nodes is generated; The step of generating a reasoning decision path diagram for multi-turn dialogues targeting cross-departmental decision-making needs based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node includes: From the multiple nodes of the knowledge decision graph, select the target knowledge decision node that matches each of the intermediate decision path nodes; Generate the decision-making relationship among the intermediate decision-making path nodes, the target knowledge decision-making nodes corresponding to the intermediate decision-making path nodes, the data decision-making source, and the decision confidence level; Based on the decision reasoning path and the decision association relationship corresponding to each intermediate decision path node, a reasoning decision path diagram for multi-round dialogues targeting the cross-departmental decision-making needs is generated.

[0012] A second aspect of the present invention discloses an intelligent processing device for cross-departmental decision-making, the device comprising: The detection module is used to detect cross-departmental decision-making needs triggered by target users through natural language expressions in the current decision-making scenario; The analysis module is used to perform intent parsing and problem decomposition operations on the cross-departmental decision-making needs according to a preset decision analysis model, so as to obtain a first decision task and a second decision task corresponding to the cross-departmental decision-making needs; wherein, the first decision task is used to perform a structured query operation, and the second decision task is used to perform a semantic retrieval operation; The retrieval module is used to perform a graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task, and obtain a structured decision case that matches the first decision task. The retrieval module is further configured to perform vector semantic retrieval operations on the knowledge decision graph according to the second decision task to obtain a decision document that matches the second decision task; The analysis module is further configured to perform data fusion operations on the structured decision case and the decision document to obtain a structured decision information package that matches the current dialogue of the cross-departmental decision-making needs; wherein, the structured decision information package is used as the decision basis for the target user in the current decision scenario.

[0013] As an optional implementation, in a second aspect of the present invention, the retrieval module performs a graph path retrieval operation on a pre-constructed knowledge decision graph based on the first decision task to obtain a structured decision case matching the first decision task, including: Perform graph node analysis on the first decision task to obtain all decision nodes of the first decision task, and analyze the node association relationship between all decision nodes based on the pre-constructed knowledge decision graph. Based on the node association relationship, a central decision node is determined from all the decision nodes, and the central decision node is used as the node center. Based on the node association relationship, a graph retrieval path for the first decision task is generated. Based on the graph retrieval path of the first decision task, a graph path retrieval operation is performed on the knowledge decision graph to obtain a structured decision case that matches the first decision task.

[0014] As an optional implementation, in a second aspect of the present invention, the specific method by which the retrieval module performs a vector semantic retrieval operation on the knowledge decision graph according to the second decision task to obtain a decision document matching the second decision task includes: The second decision task is vectorized to obtain the third decision task, and the decision filtering parameters of the third decision task are generated. According to the third decision task, a vector semantic retrieval operation is performed on the knowledge decision graph to obtain the decision node parameters that match the third decision task, and according to the third decision task, the first decision document that matches the third decision task is selected from the pre-determined document decision database. The first decision document is correlated with the decision node parameters corresponding to the third decision task to obtain a second decision document that matches the decision node parameters. Based on the decision filtering parameters, a filtering operation is performed on the second decision document to obtain a decision document that matches the second decision task.

[0015] As an optional implementation, in a second aspect of the invention, the specific method by which the analysis module performs a data fusion operation on the structured decision case and the decision document to obtain a structured decision information package matching the current dialogue of the cross-departmental decision-making needs includes: For any target decision node in the structured decision case, in the decision documents, filter the decision documents associated with the target decision node; wherein, all decision nodes include all target decision nodes; Analyze the importance of each target decision node to the current decision scenario and the correlation strength among all target decision nodes; Based on the decision importance of each target decision node and the correlation strength between all target decision nodes, the decision documents corresponding to each target decision node are sorted to obtain document sorting parameters. Based on the document sorting parameters, a corresponding case-document association relationship is established for each target decision node and its associated first decision document. Analyze the decision documents to obtain multiple document decision parameters, and set decision basis parameters for the structured decision cases based on all the document decision parameters; obtain the data decision sources for the structured decision cases and the data decision sources for each decision document; Based on the case-document relationship, the structured decision cases, all the decision basis parameters, and all the data decision sources, a structured decision information package matching the current dialogue of the cross-departmental decision-making needs is generated.

[0016] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The acquisition module is used to acquire a short-term memory cache database and a long-term memory database corresponding to the current session of the cross-departmental decision-making requirement, based on the cross-departmental decision-making requirement; wherein, the short-term memory cache database stores decision parameters corresponding to multiple rounds of dialogue that have already existed for the cross-departmental decision-making requirement; the long-term memory database stores a historical structured decision case set and business parameters corresponding to each historical structured decision case in the historical structured decision case set; The judgment module is used to determine whether there is scenario decision data matching the target decision task in the short-term memory cache database and the long-term memory database; wherein, the target decision task includes the first decision task and the second decision task; when it is determined that there is no scenario decision data, the retrieval module is triggered to perform the graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task to obtain the structured decision case matching the first decision task; The first determining module is used to determine the scene decision data from the short-term memory cache database and the long-term memory database when it is determined that the scene decision data exists; The analysis module is further configured to determine, based on the first decision task and the second decision task, a structured decision information package that matches the current dialogue of the cross-departmental decision-making needs from the scenario decision data.

[0017] As an optional implementation, in a second aspect of the invention, the apparatus further includes: The second determining module is used to determine, based on the structured decision information package, all intermediate decision path nodes between detecting the cross-departmental decision requirement and generating the structured decision information package; wherein each intermediate decision path node has a corresponding decision result; wherein the decision result of the intermediate decision path node that occurs earlier in two associated intermediate decision path nodes is the input parameter of the intermediate decision path node that occurs later. The second determining module is further configured to determine the decision confidence level of each of the intermediate decision path nodes; The generation module is used to generate a decision reasoning path for a multi-turn dialogue for the cross-departmental decision-making needs, based on the decision root path node of the decision reasoning path diagram required by the detected cross-departmental decision-making needs and the decision end path node of the decision reasoning path diagram, according to the decision path generation parameters, by taking the decision root path node, all the intermediate decision path nodes and the decision end path node. The generation module is further configured to generate a reasoning decision path diagram for multi-turn dialogues targeting the cross-departmental decision-making needs, based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node.

[0018] As an optional implementation, in a second aspect of the invention, the second determining module determines the decision confidence level of each intermediate decision path node in a specific manner, including: For any of the intermediate decision path nodes, the data decision source of the intermediate decision path node is determined according to the decision parameters of the intermediate decision path node, wherein the decision parameters of each intermediate decision path node include the structured decision case and / or the decision document; Based on the data decision sources and decision parameters of the intermediate decision path nodes, the decision confidence level of the intermediate decision path nodes is generated; The specific method by which the generation module generates a reasoning decision path diagram for the multi-turn dialogue required for the cross-departmental decision-making needs, based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node, includes: From the multiple nodes of the knowledge decision graph, select the target knowledge decision node that matches each of the intermediate decision path nodes; Generate the decision-making relationship among the intermediate decision-making path nodes, the target knowledge decision-making nodes corresponding to the intermediate decision-making path nodes, the data decision-making source, and the decision confidence level; Based on the decision reasoning path and the decision association relationship corresponding to each intermediate decision path node, a reasoning decision path diagram for multi-round dialogues targeting the cross-departmental decision-making needs is generated.

[0019] A third aspect of the present invention discloses a decision processing system, the decision processing system comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in any of the methods described in the first aspect of the present invention.

[0020] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in any of the methods described in the first aspect of the present invention.

[0021] Compared with the prior art, the present invention has the following beneficial effects: In this embodiment of the invention, the system detects that a target user triggers a cross-departmental decision-making need for the current decision-making scenario through natural language expression; it performs intent parsing and problem decomposition operations on the cross-departmental decision-making need according to a preset decision analysis model to obtain a first decision task and a second decision task corresponding to the cross-departmental decision-making need; wherein, the first decision task is used to perform a structured query operation, and the second decision task is used to perform a semantic retrieval operation; it performs a graph path retrieval operation on a pre-constructed knowledge decision graph according to the first decision task to obtain a structured decision case matching the first decision task; it performs a vector semantic retrieval operation on the knowledge decision graph according to the second decision task to obtain a decision document matching the second decision task; it performs a data fusion operation on the structured decision case and the decision document to obtain a structured decision information package matching the current dialogue of the cross-departmental decision-making need, wherein the structured decision information package is used as the decision basis for the target user in the current decision-making scenario. As can be seen, implementing this invention triggers cross-departmental decision-making needs through natural language, accurately captures and efficiently decomposes these needs, resulting in two types of decision tasks: structured query and semantic retrieval. Based on these structured query and semantic retrieval tasks, structured decision case retrieval and unstructured decision document retrieval are performed on the knowledge decision graph. This reduces information fragmentation and cumbersome retrieval in cross-departmental decision-making, improves the retrieval efficiency and accuracy of both structured and unstructured decision documents, and integrates structured decision cases and documents to obtain structured decision information packages. This consolidates fragmented information and improves the generation efficiency and accuracy of structured decision information packages, thereby enhancing the efficiency and accuracy of cross-departmental decision-making, improving inter-departmental collaboration efficiency and accuracy, and ultimately enhancing enterprise management efficiency, security, and accuracy. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating an intelligent processing method for cross-departmental decision-making disclosed in an embodiment of the present invention; Figure 2 This is a flowchart illustrating another intelligent processing method for cross-departmental decision-making disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an intelligent processing device for cross-departmental decision-making disclosed in an embodiment of the present invention; Figure 4This is a schematic diagram of another intelligent processing device for cross-departmental decision-making disclosed in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a decision processing system disclosed in an embodiment of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0025] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or end that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or ends.

[0026] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will understand, explicitly and implicitly, that the embodiments described herein can be combined with other embodiments.

[0027] This invention discloses an intelligent processing method, device, and system for cross-departmental decision-making. It triggers cross-departmental decision-making needs through natural language, accurately captures and efficiently decomposes these needs, resulting in two types of decision tasks: structured query and semantic retrieval. Based on these structured query and semantic retrieval tasks, it performs structured decision case retrieval and unstructured decision document retrieval on a knowledge decision graph. This reduces information fragmentation and cumbersome retrieval in cross-departmental decision-making, improving the retrieval efficiency and accuracy of both structured and unstructured decision documents. Furthermore, it fuses structured decision cases and decision documents to obtain structured decision information packages, integrating fragmented information and improving the generation efficiency and accuracy of these packages. This, in turn, enhances the efficiency and accuracy of cross-departmental decision-making, improves inter-departmental collaboration, and ultimately enhances enterprise management efficiency, security, and accuracy. Detailed explanations follow.

[0028] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating an intelligent processing method for cross-departmental decision-making disclosed in an embodiment of the present invention. Figure 1 The described method can be applied to any scenario requiring cross-departmental decision-making, such as product return decision analysis, new user acquisition campaign decision-making, and raw material price increase scenarios. Figure 1 As shown, the method may include the following operations: 101. Detect the target user's cross-departmental decision-making needs triggered by natural language expression in the current decision-making scenario.

[0029] In this embodiment of the invention, natural language can be text and / or voice and / or images.

[0030] 102. Based on the preset decision analysis model, perform intent parsing and problem decomposition operations on cross-departmental decision requirements to obtain the first decision task and the second decision task corresponding to the cross-departmental decision requirements; wherein, the first decision task is used to perform structured query operations, and the second decision task is used to perform semantic retrieval operations.

[0031] In this embodiment of the invention, a basic decision analysis model is trained using sample decision data. This allows the model to learn from sample decision data representing different cross-departmental decision-making needs across various decision-making scenarios. It automatically distinguishes the boundaries between two types of decision tasks. All retrieval conditions for structured decision tasks are quantifiable and precisely matchable fields, while semantic retrieval tasks match unstructured needs. This supplements relevant decision data beyond structured decision cases, such as decision guidelines. The basic decision analysis model includes, but is not limited to, one or more models capable of analyzing natural language, such as LLM and BERT. The sample decision data includes sample decision data representing different cross-departmental decision-making needs across various decision-making scenarios. The specific content of the sample decision data is detailed in the relevant content of structured decision cases and decision documents in this invention, and will not be repeated here.

[0032] In this embodiment of the invention, a preset decision analysis model is used to first analyze the decision intent in cross-departmental decision-making needs. Then, based on the decision intent, the cross-departmental decision-making needs are further broken down and analyzed to obtain the first decision task and the second decision task corresponding to the cross-departmental decision-making needs. For example, the cross-departmental decision-making need is: "Please help me find out why the return rate of 3C products in South China has increased by 12% in the past 3 weeks, and also find out how similar problems were handled in the past, as well as the corresponding system and regulations." Through the preset decision analysis model, the intent is analyzed as follows: cross-departmental operational problem decision category, locating the reasons for the increase in the return rate of 3C products and obtaining feasible handling reference cases, requiring coordination among multiple departments such as supply chain, warehousing, customer service, and operations. Then, the problem is broken down and analyzed to obtain the first decision task: region = South China, category = 3C digital products, time range = the past 3 weeks, core indicator = return rate increase of 12%, and the second decision task: experience in handling similar problems, system and regulations / guidelines, etc.

[0033] 103. Perform a graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task to obtain a structured decision case that matches the first decision task, and perform a vector semantic retrieval operation on the knowledge decision graph according to the second decision task to obtain a decision document that matches the second decision task.

[0034] In this embodiment of the invention, the knowledge decision graph consists of various decision graph nodes and the decision relationships between each decision graph node. These decision graph nodes include, but are not limited to, business scenario nodes (e.g., South China region, 3C category, etc.), decision case nodes, decision document nodes, decision-making process object nodes (e.g., organization department, merchandise department, warehousing department, customer service department, etc.), and rule / guideline nodes (e.g., the 0.3% damage rate threshold for 3C category sorting). Decision relationships include, but are not limited to, scenario-based decision relationships (e.g., the applicable scenario for the 2024 rectification case is the 3C category in South China), object-based decision relationships (e.g., the responsible department for the warehousing rectification process is the warehousing department), rule-based decision relationships (e.g., the execution standard for the warehousing sorting process is the "Warehousing Service Quality Assessment Standard"), causal decision relationships (e.g., a 0.7% warehousing damage rate leads to an increase in the 3C category return rate), temporal decision relationships (e.g., warehousing rectification must precede packaging optimization), and source-based decision relationships (e.g., the reference rules for the 2024 rectification case are derived from the "Cross-Departmental Collaboration Specification"), etc.

[0035] 104. Perform data fusion operations on structured decision cases and decision documents to obtain a structured decision information package that matches the current dialogue of cross-departmental decision-making needs. The structured decision information package is used as the basis for the target user's decision-making in the current decision-making scenario.

[0036] It is evident that implementation Figure 1 The described method triggers cross-departmental decision-making needs through natural language, accurately captures and efficiently decomposes these needs, resulting in two types of decision tasks: structured query and semantic retrieval. Based on these structured query and semantic retrieval tasks, the method performs structured decision case retrieval and unstructured decision document retrieval on the knowledge decision graph. This reduces information fragmentation and cumbersome retrieval in cross-departmental decision-making, improving the retrieval efficiency and accuracy of both structured and unstructured decision documents. Furthermore, the method integrates structured decision cases and decision documents to obtain structured decision information packages, consolidating fragmented information and improving the generation efficiency and accuracy of these packages. This, in turn, enhances the efficiency and accuracy of cross-departmental decision-making, improves inter-departmental collaboration, and ultimately enhances enterprise management efficiency, security, and accuracy.

[0037] In this embodiment of the invention, optionally, a graph path retrieval operation is performed on the pre-constructed knowledge decision graph according to the first decision task to obtain a structured decision case matching the first decision task, including: Perform graph node analysis on the first decision task to obtain all decision nodes of the first decision task, and analyze the node association relationship between all decision nodes based on the pre-constructed knowledge decision graph; Based on the node association relationships, the central decision node is determined from all decision nodes, and the graph retrieval path for the first decision task is generated based on the node association relationships, with the central decision node as the node center. Based on the graph retrieval path of the first decision task, a graph path retrieval operation is performed on the knowledge decision graph to obtain a structured decision case that matches the first decision task.

[0038] In this embodiment of the invention, the knowledge decision graph contains each decision node and the relationships between nodes in all historical structured decision cases. Continuing with the example of a product return decision analysis scenario, the first decision task is to query historical structured decision cases corresponding to the increase in return rates for the 3C category in South China over the past three weeks. Graph node analysis yields all decision nodes: South China region, 3C category, increase in return rate, time range of the past three weeks, and a 12% increase. Through knowledge decision graph analysis, the relationships between all decision nodes are obtained. Then, the central decision node is analyzed; for example, an increase in return rate yields a graph retrieval path of: increase in return rate → increase of 10%-15% → 3C category → South China region → time range of the past month. The knowledge decision graph first retrieves all historical structured decision cases corresponding to the central decision node. Then, based on the graph retrieval path, it retrieves historical structured decision cases corresponding to the next less important decision node from among all these historical structured decision cases. This process continues sequentially for subsequent decision nodes, and the historical structured decision case corresponding to the last decision node is selected as the structured decision case matching the first decision task.

[0039] As can be seen, implementing the embodiments of the present invention can also construct a precise graph retrieval path by sequentially analyzing graph nodes, graph node relationships, and central graph nodes for structured decision-making tasks, and perform graph path retrieval on the knowledge decision graph based on the graph retrieval path, thereby improving the retrieval efficiency and accuracy of complete structured decision-making cases, which is conducive to further improving the analysis efficiency and accuracy of structured decision-making information packages, and thus conducive to improving the efficiency and accuracy of cross-departmental decision-making.

[0040] In this embodiment of the invention, optionally, performing a vector semantic retrieval operation on the knowledge decision graph according to the second decision task to obtain a decision document matching the second decision task includes: The second decision task is vectorized to obtain the third decision task, and the decision filtering parameters of the third decision task are generated. These decision filtering parameters are used to perform matching filtering operations. Based on the third decision task, a vector semantic retrieval operation is performed on the knowledge decision graph to obtain the decision node parameters that match the third decision task. Based on the third decision task, the first decision document that matches the third decision task is selected from the pre-determined document decision database. By performing correlation analysis between the first decision document and the decision node parameters corresponding to the third decision task, a second decision document matching the decision node parameters is obtained. Based on the decision filtering parameters, a filtering operation is performed on the second decision document to obtain a decision document that matches the second decision task.

[0041] In this embodiment of the invention, different third decision tasks correspond to different decision filtering parameters. For example, following the previous example, the decision filtering parameters can be that the document type is a solution type, the publication time is within the last 2 years, and the applicable scope is e-commerce retail scenario. After performing vector semantic retrieval operation on the knowledge decision graph, the obtained decision node parameters can be 3C category return management, regional abnormal operation problems, and cross-departmental collaboration process. At the same time, the first decision document related to return processing is screened out, and then the first decision document and the vector of the aforementioned obtained decision node parameters are analyzed for similarity to obtain the second decision document with a similarity greater than the preset similarity (such as 90%). Finally, the second decision document is filtered based on the aforementioned decision filtering parameters to obtain the final required decision document.

[0042] As can be seen, implementing the embodiments of the present invention can also accurately capture users' deep implicit needs by converting unstructured decision-making tasks into semantic vectors, reduce information loss caused by keyword omissions, and analyze corresponding decision filtering parameters, obtain corresponding decision node parameters by semantic retrieval of the knowledge decision graph, and perform decision document analysis to initially screen matching decision documents. Then, combined with the decision node parameters obtained from the knowledge decision graph, the initially screened decision documents are further refined to ensure the semantic relevance of the decision documents. Finally, the refined decision documents are filtered based on the decision filtering parameters to exclude irrelevant decision documents, so that the final selected decision documents are highly matched with the current decision scenario, which greatly improves the accuracy and reliability of decision document retrieval. This is conducive to further improving the analysis efficiency and accuracy of structured decision information packages, and thus conducive to improving the efficiency and accuracy of cross-departmental decision-making.

[0043] In this embodiment of the invention, optionally, a data fusion operation is performed on the structured decision-making cases and decision documents to obtain a structured decision information package that matches the current dialogue regarding cross-departmental decision-making needs, including: For any target decision node in a structured decision case, filter the decision documents associated with the target decision node in the decision documents; where all decision nodes include all target decision nodes; Analyze the importance of each target decision node to the current decision scenario and the correlation strength between all target decision nodes. The sum of the decision importance of all target decision nodes is equal to 1. The higher the correlation strength, the stronger the correlation between target decision nodes. Based on the decision importance of each target decision node and the correlation strength between all target decision nodes, the decision documents corresponding to each target decision node are sorted to obtain document sorting parameters. Based on the document sorting parameters, a corresponding case-document relationship is established for each target decision node and its first associated decision document. The higher the ranking, the higher the decision importance and correlation strength. Analyze decision documents to obtain multiple document decision parameters, and set decision basis parameters for structured decision cases based on all document decision parameters; obtain the data decision sources for structured decision cases and the data decision sources for each decision document; Based on case-document relationships, structured decision-making cases, all decision-making basis parameters, and all data decision sources, generate a structured decision information package that matches the current dialogue with cross-departmental decision-making needs.

[0044] In this embodiment of the invention, the importance of each target decision node to the current decision scenario and the correlation strength between all target decision nodes are analyzed in conjunction with the central decision node. Using the previous example, considering that the key reason for the current increase in return rates is warehousing issues, the analysis shows that the decision importance of investigating warehousing and sorting problems is 0.4, the decision importance of upgrading 3C product packaging is 0.3, and the decision importance of customer service compensation scripts and cross-departmental collaboration processes are 0.15 respectively. Simultaneously, the analysis shows that the correlation strength between investigating warehousing and sorting problems and upgrading 3C product packaging reaches 0.8, and the correlation strength between investigating warehousing and sorting problems and customer service compensation scripts reaches 0.1.

[0045] In this embodiment of the invention, document decision parameters are used to indicate that if the current decision scenario reaches the corresponding decision parameters, a decision adjustment is required. For example, the parameter extracted from the "Warehouse Service Quality Assessment Standard" is that the threshold for the sorting damage rate of 3C products is 0.3%. Warehouse outlets that exceed the threshold need to complete the sorting process rectification within 3 working days. After obtaining the document decision parameters, a binding relationship is established between the document decision parameters and the corresponding structured decision cases, which serve as their corresponding decision basis parameters. As can be seen, implementing the embodiments of the present invention can also achieve the systematic integration of discrete decision-making information by constructing an association mapping between case nodes and documents. Then, based on the importance and association strength of decision nodes, the decision documents corresponding to the decision nodes are sorted to achieve the ranking of decision priorities. At the same time, combined with the data decision sources of documents and cases, a structured decision information package matching the current decision-making scenario is generated. This not only provides clear institutional basis for cases, but also ensures the traceability of the decision-making process, and improves the accuracy and reliability of the generation of structured decision information packages, thereby helping to improve the efficiency and accuracy of cross-departmental decision-making.

[0046] In this embodiment of the invention, optionally, a structured decision information package matching the current dialogue regarding cross-departmental decision-making needs is generated based on the case-document association, structured decision cases, all decision basis parameters, and all data decision sources. This package includes: Based on the case-document relationship, obtain all decision-making process objects (such as warehousing, product, customer service, operation, etc.) involved in the structured decision-making case and decision-making document, as well as the execution order of each decision-making process object. The earlier the execution order, the higher the execution priority. Filter cases and documents related to each decision-making stage from structured decision-making cases and documents, and establish the relationship between the decision-making stage object, cases, and documents for each decision-making stage object; Based on all decision-making parameters and all data decision-making sources, the corresponding decision-making parameters and data decision-making sources are set in the documents and cases of the corresponding decision-making objects to obtain a structured decision information package for each decision-making object. Based on the execution order of objects in each decision-making stage and the structured decision information package of each decision-making stage object, a structured decision information package matching the current dialogue of cross-departmental decision-making needs is generated.

[0047] As can be seen, implementing the embodiments of the present invention can further decompose the decision-making process based on the division of responsibilities and authority, clarify the execution priority, bind corresponding decision cases and decision documents, set corresponding data decision sources, generate a structured decision information package for each decision-making process, and finally generate a final hierarchical structured decision information package based on the execution order of the decision-making process. This eliminates the need for repeated cross-departmental verification, further improving the accuracy and reliability of the generated structured decision information package. It reduces the situation of unclear responsibility boundaries and chaotic execution order in cross-departmental decision-making, as well as the situation of conflict between cases and decisions and the lack of verifiable decision basis. This is conducive to improving the efficiency and accuracy of cross-departmental decision execution, while also improving the traceability of the entire decision-making process.

[0048] In an optional embodiment, the method may further include the following steps: Based on the structured decision information package, identify multiple intermediate decision path nodes between the detection of cross-departmental decision needs and the generation of the structured decision information package; Determine the decision confidence level for each intermediate decision path node; The decision root path node is the decision reasoning path diagram required by the target user to detect cross-departmental decision-making needs, and the decision end path node is the decision reasoning path diagram to generate structured decision information packages. According to the decision path generation parameters, the decision root path node, all intermediate decision path nodes and decision end path nodes are used to generate a multi-turn dialogue decision reasoning path for cross-departmental decision-making needs. Based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node, a reasoning decision path diagram for multi-round dialogues targeting cross-departmental decision-making needs is generated.

[0049] In this optional embodiment, each intermediate decision path node has a corresponding decision result; wherein, among two associated intermediate decision path nodes, the decision result of the earlier intermediate decision path node is the input parameter of the later intermediate decision path node. All intermediate decision path nodes can be important decision path nodes from detecting cross-departmental decision needs to generating a structured decision information package, such as each step in steps 102-104 corresponding to an important decision path node. Each important decision path node can consist of multiple sub-decision path nodes, such as the two steps in step 103 each corresponding to a sub-decision path node. Continuing with the previous examples, such as the structured decision-making case retrieval node: the matched historical cases match the current decision-making scenario in terms of region, category, growth rate, and time. The data source is the officially archived ledger from the operations department, with a case matching degree of 92%, i.e., a confidence level of 92%. Another example is the decision-making document detection node: the four matched decision documents are all the latest and currently effective systems issued by each department, all sourced from the company's official knowledge base, with a document scenario matching degree of 98%, i.e., a confidence level of 98%. Yet another example is the data fusion decision-making node: the breakdown of responsibility links conforms to the company's cross-departmental collaborative responsibility norms, the parameter binding logic is consistent with similar past decisions, and the result accuracy is 95%, i.e., a confidence level of 95%.

[0050] As can be seen, after obtaining the structured decision information package matching the current decision-making scenario, this optional embodiment analyzes the intermediate decision-making path nodes between detecting cross-departmental decision-making needs and generating the structured decision information package, sequentially generates a decision reasoning path, and then combines the confidence level corresponding to each intermediate decision-making path node to generate a visualized and traceable reasoning decision-making path diagram of the current decision-making scenario. This can restore the logic of intermediate decision-making nodes in the entire decision-making chain, quantify the decision confidence level of each link, and make the entire decision-making path clear at a glance. This is conducive to improving the transparency of cross-departmental decision-making, reducing the trust cost caused by doubts about the decision-making logic among various participants, and quickly locating weak links in decision derivation through confidence level labeling. It is also beneficial to directly trace back to the corresponding decision-making path node for verification and optimization when deviations occur in the implementation of the decision, which greatly improves the trust alignment efficiency and accuracy of the decision-making scheme and greatly reduces the cost of decision correction.

[0051] In this optional embodiment, determining the decision confidence level of each intermediate decision path node may include: For any intermediate decision path node, the data decision source of the intermediate decision path node is determined according to the decision parameters of the intermediate decision path node, wherein the decision parameters of each intermediate decision path node include structured decision cases and / or decision documents. Based on the data sources and decision parameters of the intermediate decision path nodes, generate the decision confidence level of the intermediate decision path nodes; In this optional embodiment, optionally, a reasoning decision path diagram for multi-turn dialogues targeting cross-departmental decision-making needs is generated based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node, including: From multiple nodes in the knowledge decision graph, select the target knowledge decision node that matches each intermediate decision path node. For example, the target knowledge decision node of the structured decision case retrieval node is the 2024 South China 3C return rectification case node. Generate the decision association relationships among the four elements: intermediate decision path nodes, target knowledge decision nodes corresponding to intermediate decision path nodes, data decision sources, and decision confidence. Each decision association relationship serves as the basis for jumping between these four data elements. Based on the decision-making reasoning path and the decision-making relationships corresponding to each intermediate decision-making path node, a reasoning decision-making path diagram for multi-round dialogues targeting cross-departmental decision-making needs is generated, such as: User Requirement Root Node → Requirement Analysis Node (labeled with 96% confidence, associated with intent nodes, and user input source) → Case Retrieval Node (labeled with 92% confidence, associated with historical case nodes, and operational ledger source) → Document Retrieval Node (labeled with 97% confidence, associated with system nodes, and official knowledge base source) → Fusion and Decomposition Node (labeled with 95% confidence, associated with rights and responsibilities rules nodes, and collaborative norms source) → Decision Information Package Final Node.

[0052] As can be seen, implementing this optional embodiment comprehensively analyzes the data decision sources and decision parameters of intermediate decision path nodes to obtain the decision confidence of intermediate decision path nodes, improving the accuracy and reliability of decision confidence generation. Then, by combining this with the knowledge graph nodes corresponding to the intermediate decision path nodes in the knowledge decision graph, the data decision sources, and the decision reasoning path, a reasoning decision path diagram for multi-round dialogues targeting cross-departmental decision needs is generated. This improves the accuracy and reliability of the reasoning decision path diagram generation, making the entire reasoning path transparent. This, in turn, helps improve the transparency of cross-departmental decision-making, reduces the trust costs caused by doubts about the decision logic among participating parties, and allows for rapid identification of weak links in decision derivation through confidence level annotation. This facilitates direct backtracking to the corresponding decision path nodes for verification and optimization when deviations occur during decision implementation, significantly improving the trust alignment efficiency and accuracy of decision solutions and greatly reducing decision correction costs.

[0053] In this optional embodiment, optionally, for any intermediate decision path node, a decision confidence level for the intermediate decision path node is generated based on the data decision source and decision parameters of the intermediate decision path node, including: Based on the data decision source of the intermediate decision path node, a corresponding source confidence score is generated. The higher the source confidence score, the stronger the authority and reliability of the structured decision case and / or decision document corresponding to the data decision source. Analyze the parameter confidence between the decision parameters of the intermediate decision path node and the current decision scenario. For example, if the semantic matching degree between the four decision documents and the current return management scenario is 99%, 97%, 92%, and 95%, respectively, then the parameter confidence is an average matching degree of 95.75%. The higher the parameter confidence, the better the structured decision case and / or decision document matches the current decision scenario, and the stronger the reliability of the structured decision case and / or decision document. Based on the source confidence and parameter confidence corresponding to the intermediate decision path node, the decision confidence of the intermediate decision path node is generated, such as by weighting the source confidence and parameter confidence to obtain the decision confidence.

[0054] As can be seen, implementing this optional embodiment also allows for the sequential analysis of the source confidence and parameter confidence of any intermediate decision-making path node, followed by an analysis of both confidence levels, ultimately yielding the decision confidence. This improves the accuracy and reliability of the decision confidence analysis for each intermediate decision-making path node, thereby enhancing the reasoning decision-making path diagram for multi-round dialogues addressing cross-departmental decision-making needs. It provides a quantifiable credibility reference for each decision-making path link, reducing the difficulty in measuring credibility and the lack of evidence for disputes in cross-departmental decision-making processes. Furthermore, by analyzing source confidence, it improves the official nature and timeliness of decision-making basis, reducing the misuse of non-authoritative and outdated rules from the source. Additionally, by analyzing parameter confidence, it quantifies the fit between cases, documents, and the current scenario, reducing the deviation in solution adaptation caused by simply copying experience.

[0055] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating another intelligent processing method for cross-departmental decision-making disclosed in an embodiment of the present invention. Figure 2 The described method can be applied to any scenario requiring cross-departmental decision-making, such as product return decision analysis, new user acquisition campaign decision-making, and raw material price increase scenarios. Figure 2 As shown, the method may include the following operations: 201. Detect the target user's cross-departmental decision-making needs triggered by natural language expression in the current decision-making scenario.

[0056] 202. Based on the preset decision analysis model, perform intent parsing and problem decomposition operations on cross-departmental decision requirements to obtain the first decision task and the second decision task corresponding to the cross-departmental decision requirements; wherein, the first decision task is used to perform structured query operations, and the second decision task is used to perform semantic retrieval operations.

[0057] 203. Based on the cross-departmental decision-making needs, obtain the short-term memory cache database and long-term memory database corresponding to the current session of the cross-departmental decision-making needs.

[0058] In this embodiment of the invention, the short-term memory cache database stores decision parameters corresponding to multiple rounds of dialogue that already exist for cross-departmental decision-making needs, such as confirmed decision parameters mentioned in the first three rounds of dialogue of this consultation, such as 60% of the 3C returns being due to damaged packaging and 30% of after-sales disputes; the long-term memory database stores a set of historical structured decision cases and the business parameters corresponding to each historical structured decision case in the set of historical structured decision cases, such as storing all historical cases of 3C return management, corresponding reasons for returns, rectification plans, implementation effects, and other data.

[0059] 204. Determine whether there is scenario decision data matching the target decision task in the short-term memory cache database and the long-term memory database; wherein, the target decision task includes the first decision task and the second decision task; if it is determined that scenario decision data exists, proceed to step 205; if it is determined that there is no scenario decision data, proceed to step 206.

[0060] In this embodiment of the invention, optionally, if it is determined that there are similar decision requirements in the short-term memory cache database and the long-term memory database, and the first decision task and the second decision task corresponding to the decision requirement are consistent with the first decision task and the second decision task mentioned in step 204, it indicates that there is scenario decision data; otherwise, it does not exist.

[0061] 205. Determine scenario decision data from the short-term memory cache database and the long-term memory database, and based on the first and second decision tasks, determine a structured decision information package from the scenario decision data that matches the current dialogue for cross-departmental decision-making needs.

[0062] In this embodiment of the invention, the aforementioned example is used for illustration. Based directly on the first decision task and the second decision task, the decision parameters for this return are obtained from the scenario decision data, and the corresponding structured decision cases are obtained from the long-term memory database, which are used as the corresponding structured decision information package.

[0063] 206. Perform a graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task to obtain a structured decision case that matches the first decision task, and perform a vector semantic retrieval operation on the knowledge decision graph according to the second decision task to obtain a decision document that matches the second decision task.

[0064] 207. Perform data fusion operations on structured decision cases and decision documents to obtain a structured decision information package that matches the current dialogue of cross-departmental decision-making needs. The structured decision information package is used as the basis for the target user's decision-making in the current decision-making scenario.

[0065] In this embodiment of the invention, steps 201-202 and 206-207 are described in detail. Please refer to the other descriptions of steps 101-104 in Embodiment 1. This embodiment of the invention will not repeat them.

[0066] It is evident that implementation Figure 2 The described method triggers cross-departmental decision-making needs through natural language, accurately captures and efficiently decomposes these needs, resulting in two types of decision tasks: structured query and semantic retrieval. Based on these structured query and semantic retrieval tasks, the method performs structured decision case retrieval and unstructured decision document retrieval on the knowledge decision graph. This reduces information fragmentation and cumbersome retrieval in cross-departmental decision-making, improving the retrieval efficiency and accuracy of both structured and unstructured decision documents. Furthermore, the method integrates structured decision cases and decision documents to obtain structured decision information packages, consolidating fragmented information and improving the generation efficiency and accuracy of these packages. This, in turn, enhances the efficiency and accuracy of cross-departmental decision-making, improves inter-departmental collaboration, and ultimately enhances enterprise management efficiency, security, and accuracy. Furthermore, after decomposing structured and unstructured decision-making tasks, the system first determines whether corresponding scenario-based decision data exists in the short-term memory cache database and the long-term memory database. If so, the data is directly reused, ensuring that the obtained structured decision information package is accurate with the current decision scenario while improving acquisition efficiency. Additionally, by reusing confirmed information from the current session through the short-term memory, the system reduces the repeated asking of the same questions in multiple rounds of dialogue, improving the decision-making interaction experience. Moreover, by reusing similar historical structured decision-making cases in the long-term memory, the system eliminates the need to search the knowledge decision graph, improving the decision response speed. This, in turn, helps to further improve the efficiency and accuracy of cross-departmental decision-making and resource utilization.

[0067] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an intelligent processing device for cross-departmental decision-making disclosed in an embodiment of the present invention. Figure 3The described device can be applied in any scenario requiring cross-departmental decision-making, such as product return decision analysis, new user acquisition campaign decision-making, and raw material price increase scenarios. Figure 3 As shown, the device may include: The detection module 301 is used to detect cross-departmental decision-making needs triggered by the target user through natural language expression for the current decision-making scenario; The analysis module 302 is used to perform intent parsing and problem decomposition operations on cross-departmental decision-making needs according to a preset decision analysis model, and to obtain a first decision task and a second decision task corresponding to the cross-departmental decision-making needs; wherein, the first decision task is used to perform structured query operations, and the second decision task is used to perform semantic retrieval operations. The retrieval module 303 is used to perform a graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task, so as to obtain a structured decision case that matches the first decision task. The retrieval module 303 is also used to perform vector semantic retrieval operations on the knowledge decision graph according to the second decision task to obtain the decision document that matches the second decision task; The analysis module 302 is also used to perform data fusion operations on structured decision cases and decision documents to obtain a structured decision information package that matches the current dialogue of cross-departmental decision-making needs; wherein, the structured decision information package is used as the basis for the target user's decision for the current decision scenario.

[0068] It is evident that implementation Figure 3 The described device triggers cross-departmental decision-making needs through natural language, accurately captures and efficiently decomposes these needs, resulting in two types of decision tasks: structured query and semantic retrieval. Based on these structured query and semantic retrieval tasks, it performs structured decision case retrieval and unstructured decision document retrieval on the knowledge decision graph. This reduces information fragmentation and cumbersome retrieval in cross-departmental decision-making, improves the retrieval efficiency and accuracy of both structured and unstructured decision documents, and integrates structured decision cases and documents to obtain structured decision information packages. This consolidates fragmented information and improves the generation efficiency and accuracy of structured decision information packages, thereby enhancing the efficiency and accuracy of cross-departmental decision-making, improving inter-departmental collaboration efficiency and accuracy, and ultimately enhancing enterprise management efficiency, security, and accuracy.

[0069] In this embodiment of the invention, optionally, the retrieval module 303 performs a graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task to obtain a structured decision case matching the first decision task, including: Perform graph node analysis on the first decision task to obtain all decision nodes of the first decision task, and analyze the node association relationship between all decision nodes based on the pre-constructed knowledge decision graph; Based on the node association relationships, the central decision node is determined from all decision nodes, and the graph retrieval path for the first decision task is generated based on the node association relationships, with the central decision node as the node center. Based on the graph retrieval path of the first decision task, a graph path retrieval operation is performed on the knowledge decision graph to obtain a structured decision case that matches the first decision task.

[0070] It is evident that implementation Figure 3 The described device can also construct a precise graph retrieval path by sequentially analyzing graph nodes, graph node relationships, and central graph nodes for structured decision-making tasks, and perform graph path retrieval on the knowledge decision graph based on the graph retrieval path. This improves the retrieval efficiency and accuracy of complete structured decision-making cases, thereby further improving the analysis efficiency and accuracy of structured decision-making information packages, and consequently improving the efficiency and accuracy of cross-departmental decision-making.

[0071] In this embodiment of the invention, optionally, the retrieval module 303 performs a vector semantic retrieval operation on the knowledge decision graph according to the second decision task to obtain the decision document matching the second decision task, including the following specific methods: The second decision task is vectorized to obtain the third decision task, and the decision filtering parameters for the third decision task are generated. Based on the third decision task, a vector semantic retrieval operation is performed on the knowledge decision graph to obtain the decision node parameters that match the third decision task. Based on the third decision task, the first decision document that matches the third decision task is selected from the pre-determined document decision database. By performing correlation analysis between the first decision document and the decision node parameters corresponding to the third decision task, a second decision document matching the decision node parameters is obtained. Based on the decision filtering parameters, a filtering operation is performed on the second decision document to obtain a decision document that matches the second decision task.

[0072] It is evident that implementation Figure 3The described device can also accurately capture users' deep implicit needs by converting unstructured decision-making tasks into semantic vectors, reducing information loss caused by keyword omissions, and analyzing corresponding decision filtering parameters. It performs semantic retrieval of the knowledge decision graph to obtain corresponding decision node parameters and analyzes decision documents to initially screen matching decision documents. Then, combined with the decision node parameters obtained from the knowledge decision graph, the initially screened decision documents are further refined to ensure semantic relevance. Finally, based on the decision filtering parameters, the refined decision documents are filtered to exclude irrelevant ones, ensuring that the final selected decision documents are highly matched to the current decision-making scenario. This greatly improves the accuracy and reliability of decision document retrieval, thereby further improving the efficiency and accuracy of structured decision information package analysis, and ultimately improving the efficiency and accuracy of cross-departmental decision-making.

[0073] In this embodiment of the invention, optionally, the analysis module 302 performs a data fusion operation on the structured decision-making cases and decision documents to obtain a structured decision-making information package that matches the current dialogue regarding cross-departmental decision-making needs. The specific methods include: For any target decision node in a structured decision case, filter the decision documents associated with the target decision node in the decision documents; where all decision nodes include all target decision nodes; Analyze the importance of each target decision node to the current decision scenario and the strength of the correlation between all target decision nodes; Based on the decision importance of each target decision node and the correlation strength between all target decision nodes, the decision documents corresponding to each target decision node are sorted to obtain document sorting parameters. Based on the document sorting parameters, a corresponding case-document relationship is established for each target decision node and its associated first decision document. Analyze decision documents to obtain multiple document decision parameters, and set decision basis parameters for structured decision cases based on all document decision parameters; obtain the data decision sources for structured decision cases and the data decision sources for each decision document; Based on case-document relationships, structured decision-making cases, all decision-making basis parameters, and all data decision sources, generate a structured decision information package that matches the current dialogue with cross-departmental decision-making needs.

[0074] It is evident that implementation Figure 3The described device can also systematically integrate discrete decision-making information by constructing an association mapping between case nodes and documents. Then, based on the importance and association strength of the decision nodes, it sorts the decision documents corresponding to the decision nodes, thereby prioritizing the decisions. At the same time, by combining the data decision sources of documents and cases, it generates a structured decision information package that matches the current decision-making scenario. This provides clear institutional basis for the cases, ensures the traceability of the decision-making process, and improves the accuracy and reliability of the structured decision information package generation, thereby helping to improve the efficiency and accuracy of cross-departmental decision-making.

[0075] In this embodiment of the invention, optionally, the specific method by which the analysis module 302 generates a structured decision information package matching the current dialogue for cross-departmental decision-making needs based on case-document relationships, structured decision cases, all decision basis parameters, and all data decision sources includes: Based on the case-document relationship, obtain all decision-making process objects (such as warehousing, merchandise, customer service, operations, etc.) involved in the structured decision-making case and decision-making document, as well as the execution order of each decision-making process object. The earlier the execution order, the higher the execution priority. Filter cases and documents related to each decision-making stage from structured decision-making cases and documents, and establish the relationship between the decision-making stage object, cases, and documents for each decision-making stage object; Based on all decision-making parameters and all data decision-making sources, the corresponding decision-making parameters and data decision-making sources are set in the documents and cases of the corresponding decision-making objects to obtain a structured decision information package for each decision-making object. Based on the execution order of objects in each decision-making stage and the structured decision information package of each decision-making stage object, a structured decision information package matching the current dialogue of cross-departmental decision-making needs is generated.

[0076] It is evident that implementation Figure 3 The described device can also break down decision-making processes based on responsibilities and authority, clarify execution priorities, bind corresponding decision cases and documents, set corresponding data decision sources, generate structured decision information packages for each decision-making process, and finally generate a final hierarchical structured decision information package based on the execution order of the decision-making processes. This eliminates the need for repeated cross-departmental verification, further improving the accuracy and reliability of the structured decision information package generation. It reduces the ambiguity of responsibility boundaries and the confusion of execution order in cross-departmental decision-making, as well as the conflict between cases and decisions and the lack of verifiable decision-making basis. This helps improve the efficiency and accuracy of cross-departmental decision execution, while also enhancing the traceability of the entire decision-making process.

[0077] In an optional embodiment, such as Figure 4As shown, the device may further include: The acquisition module 304 is used to acquire the short-term memory cache database and long-term memory database corresponding to the current session of the cross-department decision-making needs, based on the cross-department decision-making needs. The short-term memory cache database stores the decision parameters corresponding to the multi-turn dialogues that already exist for the cross-department decision-making needs. The long-term memory database stores the historical structured decision case set and the business parameters corresponding to each historical structured decision case in the historical structured decision case set. The judgment module 305 is used to determine whether there is scenario decision data matching the target decision task in the short-term memory cache database and the long-term memory database; wherein, the target decision task includes a first decision task and a second decision task; when it is determined that there is no scenario decision data, the retrieval module is triggered to perform a graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task to obtain a structured decision case matching the first decision task; The first determining module 306 is used to determine the scenario decision data from the short-term memory cache database and the long-term memory database when it is determined that scenario decision data exists. The analysis module 302 is also used to determine, based on the first and second decision tasks, a structured decision information package that matches the current dialogue with the cross-departmental decision-making needs from the scenario decision data.

[0078] It is evident that implementation Figure 4 The described device can also, after breaking down structured and unstructured decision-making tasks, first determine whether corresponding scenario decision data exists in the short-term memory cache database and the long-term memory database. If it does, it can directly reuse the data, ensuring that the obtained structured decision information package is accurate with the current decision scenario while improving acquisition efficiency. Furthermore, by reusing confirmed information from the current session through the short-term memory, it reduces the repeated asking of the same questions in multi-turn dialogues, improving the decision-making interaction experience. Additionally, by reusing similar historical structured decision cases in the long-term memory, it eliminates the need to search the knowledge decision graph, improving the decision response speed. This is conducive to further improving the efficiency and accuracy of cross-departmental decision-making and resource utilization.

[0079] In an optional embodiment, such as Figure 4 As shown, the device may further include: The second determining module 307 is used to determine all intermediate decision path nodes from the detection of cross-departmental decision needs to the generation of the structured decision information package based on the structured decision information package; wherein each intermediate decision path node has a corresponding decision result; wherein the decision result of the intermediate decision path node that occurs earlier in two related intermediate decision path nodes is the input parameter of the intermediate decision path node that occurs later. The second determining module 307 is also used to determine the decision confidence level of each intermediate decision path node; The generation module 308 is used to generate a decision reasoning path for a multi-turn dialogue for cross-departmental decision-making needs, based on the decision root path node of the decision reasoning path diagram required by the target user with the detection of cross-departmental decision-making needs and the decision final path node of the decision reasoning path diagram with the generation of structured decision information packages, according to the decision path generation parameters. The generation module 308 is also used to generate a reasoning decision path diagram for multi-round dialogues that meet cross-departmental decision-making needs, based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node.

[0080] It is evident that implementation Figure 4 The described device can also analyze the intermediate decision path nodes between detecting cross-departmental decision needs and generating the structured decision information package after obtaining the structured decision information package that matches the current decision scenario. It can then generate decision reasoning paths sequentially and combine the confidence level of each intermediate decision path node to generate a visual and traceable reasoning decision path diagram of the current decision scenario. This can restore the logic of intermediate decision nodes in the entire decision chain, quantify the confidence level of each stage, and make the entire decision path clear at a glance. This helps to improve the transparency of cross-departmental decision-making, reduce the trust costs caused by doubts about the decision logic among various participants, and quickly locate weak links in decision derivation through confidence level labeling. This is beneficial for subsequent backtracking to the corresponding decision path nodes for verification and optimization when deviations occur in the implementation of the decision. This greatly improves the trust alignment efficiency and accuracy of the decision scheme and significantly reduces the cost of decision correction.

[0081] In this optional embodiment, such as Figure 4 As shown, the second determining module 308 determines the specific method by which it determines the decision confidence level of each intermediate decision path node, including: For any intermediate decision path node, the data decision source of the intermediate decision path node is determined according to the decision parameters of the intermediate decision path node, wherein the decision parameters of each intermediate decision path node include structured decision cases and / or decision documents. Based on the data sources and decision parameters of the intermediate decision path nodes, generate the decision confidence level of the intermediate decision path nodes; Specifically, the generation module 309 generates a reasoning decision path diagram for multi-turn dialogues targeting cross-departmental decision-making needs based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node. The specific methods include: From multiple nodes in the knowledge decision graph, select the target knowledge decision node that matches each intermediate decision path node; Generate the decision-making relationships among the four elements: intermediate decision path nodes, target knowledge decision nodes corresponding to intermediate decision path nodes, data decision sources, and decision confidence levels. Based on the decision reasoning path and the decision association relationship corresponding to each intermediate decision path node, a reasoning decision path diagram for multi-round dialogues targeting cross-departmental decision-making needs is generated.

[0082] It is evident that implementation Figure 4 The described device can also obtain the decision confidence of intermediate decision path nodes by comprehensively analyzing the data decision sources and decision parameters of intermediate decision path nodes, thereby improving the accuracy and reliability of decision confidence generation. It then combines this confidence with the corresponding knowledge graph nodes, data decision sources, and decision reasoning paths of the intermediate decision path nodes in the knowledge decision graph to generate a reasoning decision path diagram for multi-round dialogues targeting cross-departmental decision needs. This improves the accuracy and reliability of the reasoning decision path diagram generation, making the entire reasoning path transparent. This, in turn, helps improve the transparency of cross-departmental decision-making, reduces the trust costs caused by doubts about the decision logic among participating parties, and allows for rapid identification of weak links in decision derivation through confidence level labeling. This facilitates direct backtracking to the corresponding decision path nodes for verification and optimization when deviations occur during decision implementation, significantly improving the trust alignment efficiency and accuracy of decision schemes and greatly reducing decision correction costs.

[0083] In this optional embodiment, optionally, the second determining module 308 generates the decision confidence level of any intermediate decision path node based on the data decision source and decision parameters of the intermediate decision path node in a specific way, including: Based on the data decision source of the intermediate decision path node, a corresponding source confidence score is generated. The higher the source confidence score, the stronger the authority and reliability of the structured decision case and / or decision document corresponding to the data decision source. Analyze the parameter confidence between the decision parameters of the intermediate decision path node and the current decision scenario. For example, if the semantic matching degree between the four decision documents and the current return management scenario is 99%, 97%, 92%, and 95%, respectively, then the parameter confidence is an average matching degree of 95.75%. The higher the parameter confidence, the better the structured decision case and / or decision document matches the current decision scenario, and the stronger the reliability of the structured decision case and / or decision document. Based on the source confidence and parameter confidence corresponding to the intermediate decision path node, the decision confidence of the intermediate decision path node is generated, such as by weighting the source confidence and parameter confidence to obtain the decision confidence.

[0084] It is evident that implementation Figure 4The described device can also analyze the source confidence and parameter confidence of any intermediate decision path node in sequence, and then analyze the confidence of both to finally obtain the decision confidence. This improves the accuracy and reliability of the decision confidence analysis for each intermediate decision path node, thereby helping to further improve the reasoning decision path diagram for multi-turn dialogues for cross-departmental decision-making needs. It provides a quantifiable credibility reference for each decision path link, reducing the difficulty in measuring credibility and the lack of basis for disputes in cross-departmental decision-making links. Furthermore, by analyzing the source confidence, it improves the officiality and timeliness of the decision basis, reducing the misuse of non-authoritative and outdated rules from the source. And by analyzing the parameter confidence, it quantifies the adaptability of cases, documents and current scenarios, reducing the deviation in solution adaptation caused by simply copying experience.

[0085] Example 4 Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a decision processing system disclosed in an embodiment of the present invention. Figure 5 The described device can be applied in any scenario requiring cross-departmental decision-making, such as product return decision analysis, new user acquisition campaign decision-making, and raw material price increase scenarios. Figure 5 As shown, the decision processing system may include: Memory 401 storing executable program code; Processor 402 coupled to memory 401; The processor 402 calls the executable program code stored in the memory 401 to execute some or all of the steps in the intelligent processing method for cross-departmental decision-making described in either Embodiment 1 or Embodiment 2 of the present invention.

[0086] Example 5 This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute some or all of the steps in any of the intelligent processing methods for cross-departmental decision-making disclosed in Embodiment 1 or Embodiment 2 of this invention.

[0087] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0088] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0089] Finally, it should be noted that the above embodiments are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent processing method for cross-departmental decision-making, characterized in that, The method includes: The system detects cross-departmental decision-making needs triggered by target users through natural language descriptions in the current decision-making scenario. Based on a preset decision analysis model, the cross-departmental decision requirements are subjected to intent parsing and problem decomposition operations to obtain a first decision task and a second decision task corresponding to the cross-departmental decision requirements; wherein, the first decision task is used to perform a structured query operation, and the second decision task is used to perform a semantic retrieval operation; Based on the first decision task, a graph path retrieval operation is performed on the pre-constructed knowledge decision graph to obtain a structured decision case that matches the first decision task; Based on the second decision task, a vector semantic retrieval operation is performed on the knowledge decision graph to obtain a decision document that matches the second decision task; A data fusion operation is performed on the structured decision case and the decision document to obtain a structured decision information package that matches the current dialogue of the cross-departmental decision-making needs. The structured decision information package is used as the basis for the target user's decision in the current decision scenario.

2. The method according to claim 1, characterized in that, The step of performing a graph path retrieval operation on a pre-constructed knowledge decision graph based on the first decision task to obtain a structured decision case matching the first decision task includes: Perform graph node analysis on the first decision task to obtain all decision nodes of the first decision task, and analyze the node association relationship between all decision nodes based on the pre-constructed knowledge decision graph. Based on the node association relationship, a central decision node is determined from all the decision nodes, and the central decision node is used as the node center. Based on the node association relationship, a graph retrieval path for the first decision task is generated. Based on the graph retrieval path of the first decision task, a graph path retrieval operation is performed on the knowledge decision graph to obtain a structured decision case that matches the first decision task.

3. The method according to claim 1, characterized in that, The step of performing vector semantic retrieval operations on the knowledge decision graph according to the second decision task to obtain a decision document matching the second decision task includes: The second decision task is vectorized to obtain the third decision task, and the decision filtering parameters of the third decision task are generated. According to the third decision task, a vector semantic retrieval operation is performed on the knowledge decision graph to obtain the decision node parameters that match the third decision task, and according to the third decision task, the first decision document that matches the third decision task is selected from the pre-determined document decision database. The first decision document is correlated with the decision node parameters corresponding to the third decision task to obtain a second decision document that matches the decision node parameters. Based on the decision filtering parameters, a filtering operation is performed on the second decision document to obtain a decision document that matches the second decision task.

4. The method according to any one of claims 1-3, characterized in that, The aforementioned data fusion operation performed on the structured decision case and the decision document to obtain a structured decision information package matching the current dialogue of the cross-departmental decision-making needs includes: For any target decision node in the structured decision case, in the decision documents, filter the decision documents associated with the target decision node; wherein, all decision nodes include all target decision nodes; Analyze the importance of each target decision node to the current decision scenario and the correlation strength among all target decision nodes; Based on the decision importance of each target decision node and the correlation strength between all target decision nodes, the decision documents corresponding to each target decision node are sorted to obtain document sorting parameters. Based on the document sorting parameters, a corresponding case-document association relationship is established for each target decision node and its associated first decision document. Analyze the decision documents to obtain multiple document decision parameters, and set decision basis parameters for the structured decision cases based on all the document decision parameters; obtain the data decision sources for the structured decision cases and the data decision sources for each decision document; Based on the case-document relationship, the structured decision cases, all the decision basis parameters, and all the data decision sources, a structured decision information package matching the current dialogue of the cross-departmental decision-making needs is generated.

5. The method according to any one of claims 1-3, characterized in that, The method further includes: Based on the cross-departmental decision-making requirements, a short-term memory cache database and a long-term memory database corresponding to the current session of the cross-departmental decision-making requirements are obtained; wherein, the short-term memory cache database stores decision parameters corresponding to multiple rounds of dialogue that already exist for the cross-departmental decision-making requirements; the long-term memory database stores a historical structured decision case set and business parameters corresponding to each historical structured decision case in the historical structured decision case set; Determine whether there is scenario decision data matching the target decision task in the short-term memory cache database and the long-term memory database; wherein, the target decision task includes the first decision task and the second decision task; When it is determined that the scenario decision data does not exist, the operation of performing a graph path retrieval operation on the pre-constructed knowledge decision graph based on the first decision task is executed to obtain a structured decision case that matches the first decision task. When the existence of the scenario decision data is determined, the scenario decision data is identified from the short-term memory cache database and the long-term memory database, and a structured decision information package matching the current dialogue of the cross-department decision requirement is determined from the scenario decision data based on the first decision task and the second decision task.

6. The method according to any one of claims 1-3, characterized in that, The method further includes: Based on the structured decision information package, all intermediate decision path nodes are determined from the detection of the cross-departmental decision requirement to the generation of the structured decision information package; wherein each intermediate decision path node has a corresponding decision result; wherein the decision result of the intermediate decision path node that occurs earlier in two related intermediate decision path nodes is the input parameter of the intermediate decision path node that occurs later. Determine the decision confidence level for each of the intermediate decision path nodes; The decision root path node is the decision reasoning path diagram required by the target user for detecting the cross-departmental decision-making needs, and the decision end path node is the decision reasoning path diagram generated by generating the structured decision information package. According to the decision path generation parameters, the decision root path node, all the intermediate decision path nodes and the decision end path node are used to generate a multi-turn dialogue decision reasoning path for the cross-departmental decision-making needs. Based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node, a reasoning decision path diagram for multi-round dialogues targeting the cross-departmental decision-making needs is generated.

7. The method according to claim 6, characterized in that, Determining the decision confidence level of each intermediate decision path node includes: For any of the intermediate decision path nodes, the data decision source of the intermediate decision path node is determined according to the decision parameters of the intermediate decision path node, wherein the decision parameters of each intermediate decision path node include the structured decision case and / or the decision document; Based on the data decision sources and decision parameters of the intermediate decision path nodes, the decision confidence level of the intermediate decision path nodes is generated; The step of generating a reasoning decision path diagram for multi-turn dialogues targeting cross-departmental decision-making needs based on the decision reasoning path and the decision confidence level corresponding to each intermediate decision path node includes: From the multiple nodes of the knowledge decision graph, select the target knowledge decision node that matches each of the intermediate decision path nodes; Generate the decision-making relationship among the intermediate decision-making path nodes, the target knowledge decision-making nodes corresponding to the intermediate decision-making path nodes, the data decision-making source, and the decision confidence level; Based on the decision reasoning path and the decision association relationship corresponding to each intermediate decision path node, a reasoning decision path diagram for multi-round dialogues targeting the cross-departmental decision-making needs is generated.

8. An intelligent processing device for cross-departmental decision-making, characterized in that, The device includes: The detection module is used to detect cross-departmental decision-making needs triggered by target users through natural language expressions in the current decision-making scenario; The analysis module is used to perform intent parsing and problem decomposition operations on the cross-departmental decision-making needs according to a preset decision analysis model, so as to obtain a first decision task and a second decision task corresponding to the cross-departmental decision-making needs; wherein, the first decision task is used to perform a structured query operation, and the second decision task is used to perform a semantic retrieval operation; The retrieval module is used to perform a graph path retrieval operation on the pre-constructed knowledge decision graph according to the first decision task, and obtain a structured decision case that matches the first decision task. The retrieval module is further configured to perform vector semantic retrieval operations on the knowledge decision graph according to the second decision task to obtain a decision document that matches the second decision task; The analysis module is further configured to perform data fusion operations on the structured decision case and the decision document to obtain a structured decision information package that matches the current dialogue of the cross-departmental decision-making needs; wherein, the structured decision information package is used as the decision basis for the target user in the current decision scenario.

9. A decision processing system, characterized in that, The decision processing system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the intelligent processing method for cross-departmental decision-making as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the intelligent processing method for cross-departmental decision-making as described in any one of claims 1-7.

Images