Dependent gap identification and topology self-healing method and system for knowledge topology structure

Abstract

The embodiment of the application relates to the technical field of knowledge topology processing, and discloses a dependency gap identification and topology self-healing method and system for a knowledge topology structure. The application identifies a suspended derivative knowledge node by acquiring an in-degree feature and an ontology attribute identifier of a target knowledge node in a pre-dependence directed acyclic graph. The application calculates a structural under-coverage gap according to a representation load value and a pre-dependence complexity value of the target knowledge node, limits the target knowledge node from entering a downstream consumable state when the structural under-coverage gap exceeds a threshold value, and generates a pre-dependence compensation task. When a candidate pre-dependence edge pointing to the target knowledge node is detected, edge legality verification and loop verification are performed, the pre-dependence directed acyclic graph is updated after the verification is passed, and the target knowledge node is released to enter the downstream consumable state. Therefore, the application can prevent derivative knowledge nodes in the knowledge topology structure from being misjudged as legal root nodes, and improve the topology integrity and operation stability.

Description

Technical Field

[0001] This application relates to the fields of knowledge engineering and knowledge topology processing technology, and in particular to a method and system for dependency gap identification and topology self-healing oriented towards knowledge topology structures. It can be applied to scenarios such as knowledge base construction, knowledge mother tree maintenance, skill graph governance, question bank knowledge point management, agent evaluation routing, and knowledge reasoning. Background Technology

[0002] In the process of constructing knowledge bases, question banks, competency graphs, or knowledge mother trees, it is usually necessary to maintain two types of relationships simultaneously: 1. Classification and attribution relationship, which is used to indicate which knowledge level, subject domain, or coverage matrix position a knowledge node belongs to; 2. Pre-dependency relationship, which is used to indicate whether a knowledge node depends on another knowledge node in the process of definition, derivation, application, or verification.

[0003] Existing knowledge graph or graph database systems often handle these relationships in a mixed manner, or only verify whether nodes and edges satisfy the static schema, failing to determine whether a knowledge node with an in-degree of 0 is a legitimate basic node or an isolated derived node due to missing prerequisite edges. Especially in automated knowledge base construction, LLM-assisted knowledge point extraction, cross-industry knowledge tree expansion, and automated question generation systems, if derived knowledge nodes are misjudged as basic root nodes, it may lead to: underestimating the knowledge difficulty; incorrect question generation order or progressive verification path; agents skipping necessary prerequisite knowledge during routing; unstable results in downstream knowledge reasoning; and semantic confusion between the coverage matrix and the dependency graph.

[0004] Therefore, an algorithm is needed that can distinguish between hierarchical knowledge skeletons and pre-dependency graphs, and identify the risk of missing edges in knowledge nodes based on local topological states. Summary of the Invention

[0005] One objective of this application is to provide a method and system for dependency gap identification and topology self-healing oriented towards knowledge topology structures, so as to improve the situation in related technologies where it is difficult to distinguish between hierarchical knowledge skeletons and preceding dependency graphs, and the risk of missing edges of knowledge nodes cannot be identified based on local topological states.

[0006] In a first aspect, embodiments of this application provide a dependency gap identification and topology self-healing method for knowledge topology structures. The knowledge topology structure includes a hierarchical knowledge skeleton and a directed acyclic graph of prerequisite dependencies maintained separately from the hierarchical knowledge skeleton. The hierarchical knowledge skeleton is used to represent the hierarchical affiliation of knowledge nodes, and the directed acyclic graph of prerequisite dependencies is used to represent the prerequisite dependencies between knowledge nodes. The method includes: obtaining the in-degree feature of the target knowledge node in the directed acyclic graph of prerequisite dependencies and the ontology attribute identifier of the target knowledge node; when the in-degree feature satisfies the isolation condition and the ontology attribute identifier indicates that the target knowledge node is a derived knowledge node, the target knowledge node is identified as an isolated ecological node. The process involves: obtaining the representational load value and the pre-dependency complexity value of the target knowledge node, and calculating the structural undercoverage gap based on these values; restricting the target knowledge node from entering the downstream consumable state when the structural undercoverage gap exceeds a preset threshold, and generating a pre-dependency compensation task associated with the target knowledge node; detecting candidate pre-dependency edges pointing to the target knowledge node during the effective period of the pre-dependency compensation task; performing edge validity and acyclicity checks on the candidate pre-dependency edges; and writing the candidate pre-dependency edges into the directed acyclic graph of pre-dependencies when both checks pass, and releasing the target knowledge node to enter the downstream consumable state.

[0007] In some embodiments, the hierarchical knowledge skeleton includes multiple layers of knowledge nodes and classification association edges between adjacent layers, and the pre-dependency directed acyclic graph includes pre-dependency edges between knowledge nodes in the same layer or across layers; the classification association edges are used to calculate knowledge coverage relationships, and the pre-dependency edges are used to calculate prerequisite relationships, difficulty propagation relationships or progressive verification order between knowledge nodes, and the classification association edges and the pre-dependency edges are stored separately, verified separately, and calculated separately.

[0008] In some embodiments, the coverage matrix of the hierarchical knowledge skeleton is generated based on the classification association edges of adjacent layers, and the preceding dependency edges in the directed acyclic graph of preceding dependencies do not participate in the chain multiplication calculation of the coverage matrix.

[0009] In some embodiments, the ontology attribute identifier is determined based on the specification definition text of the target knowledge node; when the specification definition text depends on other knowledge nodes, mathematical objects, logical operators, procedural steps, or registered constructs, the ontology attribute identifier is determined to be a derived type; when the specification definition text is introduced as a base object without depending on other knowledge nodes, the ontology attribute identifier is determined to be a primitive type.

[0010] In some embodiments, the structural undercoverage gap is determined according to the following formula: ,in, This indicates a gap in the structure's coverage. This represents the representational load value of the target knowledge node. This represents the complexity value of the prerequisite dependencies of the target knowledge node.

[0011] In some embodiments, the characterization load value is determined based on the abstraction level parameter and the formalization requirement parameter of the target knowledge node; the abstraction level parameter is used to represent the degree of abstraction of the knowledge object represented by the target knowledge node, and the formalization requirement parameter is used to represent the strictness of the expression form required to fully define the knowledge object.

[0012] In some embodiments, the prerequisite dependency complexity value is determined based on at least one topological feature of the target knowledge node in the directed acyclic graph of prerequisite dependencies. The topological features include the number of direct prerequisite dependencies, the number of defining prerequisite dependencies, the number of procedural prerequisite dependencies, the depth of the longest forced path from the core root node to the target knowledge node, the number of reachable ancestor nodes, the number of ancestors across knowledge domains, the confidence of prerequisite dependency edges, or the structural morphology type.

[0013] In some embodiments, edge validity verification includes: verifying whether the candidate preceding dependency edge has at least one of the following: target task family, dependency type, required mastery level, scope, blocking task identifier, test condition identifier, expected observable output, and failure rule.

[0014] In some embodiments, only pre-dependent edges that are in an active state and have a confidence level greater than or equal to a preset confidence threshold are included in the core pre-dependent graph; pre-dependent edges that do not reach the preset confidence threshold are marked as pending review and do not participate in the calculation of the pre-dependent complexity value.

[0015] In some embodiments, edge validity verification includes at least one of self-loop check, duplicate edge check, scope overlap check, blocking task overlap check, and prerequisite dependency loop check.

[0016] In some embodiments, acyclic verification includes: performing a topology sorting algorithm or a strongly connected component detection algorithm on the core prerequisite graph after writing candidate prerequisite dependencies; when a directed loop is detected in the core prerequisite graph, rejecting the writing of candidate prerequisite dependencies and maintaining the downstream consumable state limit of the target knowledge node.

[0017] In some embodiments, knowledge nodes and preceding dependent edges in the knowledge topology use a structured registry as the single source of data authenticity. Before a knowledge node or preceding dependent edge enters the downstream consumable state, field pattern verification, source reference verification, graph integrity verification, maturity status verification, or downstream access qualification verification are performed on the structured registry.

[0018] In some embodiments, the downstream consumable state is used to control whether the target knowledge node can be invoked by a question-generating system, an evaluation system, a recommendation system, an agent routing system, a knowledge reasoning system, or a knowledge base retrieval enhancement system.

[0019] Secondly, embodiments of this application provide a dependency gap identification and topology self-healing system for knowledge topology structures, comprising: a knowledge topology management module for maintaining a hierarchical knowledge skeleton and a pre-dependency directed acyclic graph maintained separately from the hierarchical knowledge skeleton; a node semantic recognition module for determining the ontology attribute identifier of the target knowledge node based on the canonical definition text of the target knowledge node; a local topological state recognition module for obtaining the in-degree feature of the target knowledge node in the pre-dependency directed acyclic graph, and identifying the target knowledge node as an isolated ecological node when the in-degree feature satisfies the isolation condition and the ontology attribute identifier is derived; and a gap quantification module. The system is divided into four modules: a structure undercoverage module, a compensation task generation module, and a topology self-healing update module. The former is used to calculate the structural undercoverage gap based on the representation load value and the complexity value of the pre-dependent dependencies of the target knowledge node. The latter is used to restrict the target knowledge node from entering the downstream consumable state and generate a pre-dependent compensation task when the structural undercoverage gap is greater than a preset threshold. The former is used to detect candidate pre-dependent edges pointing to the target knowledge node and perform edge legality verification and acyclicity verification on the candidate pre-dependent edges. The latter is used to write the candidate pre-dependent edges into the pre-dependent directed acyclic graph and release the target knowledge node to enter the downstream consumable state when both edge legality verification and acyclicity verification pass.

[0020] The embodiments of this application have the following beneficial effects: Unlike related technologies, the embodiments of this application identify isolated ecological nodes that are derived in the ontology but lack preceding dependency edges in the topology by obtaining the in-degree feature of the target knowledge node in the preceding dependency directed acyclic graph and the ontology attribute identifier of the target knowledge node; further, it calculates the structural undercoverage gap based on the representational load value and preceding dependency complexity value of the target knowledge node, and restricts the target knowledge node from entering the downstream consumable state when the structural undercoverage gap exceeds a threshold, generating a preceding dependency compensation task. After detecting a candidate preceding dependency edge pointing to the target knowledge node, the system performs edge legality verification and acyclicity verification, and updates the preceding dependency directed acyclic graph after the verification passes, releasing the target knowledge node to enter the downstream consumable state. Thus, this application can prevent derived knowledge nodes in the knowledge topology from being misjudged as legitimate root nodes, improving the topological integrity and operational stability in knowledge base construction, knowledge reasoning, agent routing, and automated question generation scenarios. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the related technologies or embodiments will be briefly introduced below. Obviously, the drawings described below only show some embodiments of this application and should not be considered as limiting the scope of protection. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 These are schematic diagrams illustrating application scenarios of dependency gap identification and topology self-healing for knowledge topology structures in some embodiments of this application; Figure 2 This is a schematic diagram of the structure of a directed acyclic graph with pre-dependencies provided in some embodiments of this application; Figure 3A This is a flowchart illustrating the dependency gap identification and topology self-healing method for knowledge topology provided in some embodiments of this application; Figure 3B This is a schematic diagram of the structure of a directed acyclic graph with pre-dependencies provided in some other embodiments of this application; Figure 4 This is a schematic diagram of the structure of a dependency gap identification and topology self-healing system for knowledge topology provided in some embodiments of this application. Detailed Implementation

[0023] To make the objectives and advantages of the embodiments of this application more readily understood, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. The detailed description of the embodiments of this application in the accompanying drawings is not intended to limit the scope of protection claimed by this application, but only represents selected embodiments of this application. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] It should be noted that, unless there is a conflict, the various technical features involved in the embodiments of this application described below can be combined with each other, and all are within the protection scope of this application. Furthermore, although functional modules are divided in the device or structural schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," "third," and other similar expressions used herein do not limit the data or execution order, but are only for illustrative purposes and to distinguish identical or similar items with substantially the same function and effect, and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features.

[0025] Unless otherwise defined, the technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. It should be understood that the term "and / or" as used in this specification includes any and all combinations of one or more of the listed items.

[0026] In the process of constructing knowledge bases, question banks, competency graphs, or knowledge mother trees, it is usually necessary to maintain two types of relationships simultaneously: 1. Classification and attribution relationship, which is used to indicate which knowledge level, subject domain, or coverage matrix position a knowledge node belongs to; 2. Pre-dependency relationship, which is used to indicate whether a knowledge node depends on another knowledge node in the process of definition, derivation, application, or verification.

[0027] Existing knowledge graph or graph database systems often handle these relationships in a mixed manner, or only verify whether nodes and edges satisfy the static schema, failing to determine whether a knowledge node with an in-degree of 0 is a legitimate basic node or an isolated derived node due to missing prerequisite edges. Especially in automated knowledge base construction, LLM-assisted knowledge point extraction, cross-industry knowledge tree expansion, and automated question generation systems, if derived knowledge nodes are misjudged as basic root nodes, it may lead to: underestimating the knowledge difficulty; incorrect question generation order or progressive verification path; agents skipping necessary prerequisite knowledge during routing; unstable results in downstream knowledge reasoning; and semantic confusion between the coverage matrix and the dependency graph.

[0028] In view of this, embodiments of this application provide a dependency gap identification and topology self-healing method for knowledge topology structures. By acquiring the in-degree characteristics and ontology attribute identifiers of target knowledge nodes in the directed acyclic graph of prior dependencies, isolated ecological nodes that are derived in terms of ontology but lack prior dependency edges in the topology are identified. Furthermore, structural under-coverage gaps are calculated based on the representational load value and prior dependency complexity value of the target knowledge node. When the structural under-coverage gap exceeds a threshold, the target knowledge node is restricted from entering the downstream consumable state, generating a prior dependency compensation task. After detecting a candidate prior dependency edge pointing to the target knowledge node, the system performs edge validity verification and acyclicity verification. After the verification passes, the directed acyclic graph of prior dependencies is updated, releasing the target knowledge node to enter the downstream consumable state. Thus, this application can prevent derived knowledge nodes in the knowledge topology structure from being misjudged as legitimate root nodes, improving the topological integrity and operational stability in knowledge base construction, knowledge reasoning, agent routing, and automated question generation scenarios.

[0029] Please refer to the following: Figure 1 as well as Figure 2 , Figure 1 The illustrations depict application scenarios of dependency gap identification and topology self-healing for knowledge topology structures provided by some embodiments of this application. Figure 2 The schematic diagram illustrates the structure of a directed acyclic graph with pre-dependencies provided in some embodiments of this application.

[0030] See Figure 1As shown, this application scenario includes a dependency gap identification and topology self-healing system 100 for knowledge topology structures. The dependency gap identification and topology self-healing system 100 for knowledge topology structures includes a knowledge topology structure (e.g., Figure 2 The preceding dependency directed acyclic graph 10 is shown. It should be understood that the dependency gap identification and topology self-healing system 100 for knowledge topology structures can be any suitable type of device or component, such as a tablet computer, desktop computer, laptop computer, server, smartphone, FPGA chip, microcontroller, or single-chip microcomputer. In the embodiments of this application, node / knowledge node has the same meaning, referring to a node in the knowledge topology structure.

[0031] Understandably, a Directed Acyclic Graph (DAG) with pre-dependencies is a directed graph structure that does not form closed cycles. A DAG with pre-dependencies consists of at least two nodes, and the edges connecting any two nodes are directed, with each directed edge (hereinafter referred to as a "directed edge") pointing from one of the two nodes to the other. A DAG with pre-dependencies does not contain any directed cycles; a directed cycle is a path that starts from a node in the graph, follows the direction of an edge, and finally returns to that node.

[0032] Generally, each node in a pre-dependency directed acyclic graph represents a task or event, and the directed edge connecting two nodes represents a "causal relationship", "dependency relationship", or "execution order", etc. DAG in blockchain, deep learning computation graph, and compilation-optimized data flow graph are all typical applications of pre-dependency directed acyclic graphs.

[0033] See Figure 2 As shown, the directed acyclic graph 10 with pre-dependencies includes nodes A, B, C, D, E, F, and G. Directed edge 11 connects node A and node B and points from node A to node B. Directed edge 12 connects node B and node D and points from node B to node D, and so on, ... Directed edge 16 connects node C and node G and points from node C to node G.

[0034] For example, in this embodiment of the application, any node in the knowledge topology is selected as the target knowledge node, and the in-degree feature of the target knowledge node in the preceding dependent directed acyclic graph and the ontology attribute identifier of the target knowledge node are obtained. For example, node A in the preceding dependent directed acyclic graph 10 is selected as the target knowledge node, and the in-degree feature of node A in the preceding dependent directed acyclic graph 10 and the ontology attribute identifier of node A are obtained.

[0035] In this context, in-degree refers to the number of edges pointing to a given node. For example, in the directed acyclic graph 10 with prior dependencies, there are no edges pointing to node A; that is, the number of edges pointing to node A is 0, and the in-degree of node A is zero. Ontology attribute identifiers are obtained by labeling / marking nodes according to standard node definitions during the node creation and registration phase using semantic classification tools (such as Large Language Models (LLMs) / classifiers). These ontology attribute identifiers, as read-only (or immutable) fields, are fixed / fixed within the node's metadata. Throughout the subsequent computational flow of the entire knowledge topology, these ontology attribute identifiers, as the underlying "genes," do not change with the addition or deletion of edges.

[0036] For example, when the in-degree feature of the target knowledge node is zero (i.e., the isolation condition is met) and the ontology attribute identifier indicates that the target knowledge node is a derived knowledge node, this embodiment of the application obtains the representational load value and the prerequisite dependency complexity value of the target knowledge node, and calculates the structural undercoverage gap based on the representational load value and the prerequisite dependency complexity value. It is understood that the representational load value is determined based on the abstraction level parameter and formal requirement parameter of the target knowledge node. The prerequisite dependency complexity value is determined based on at least one topological feature of the target knowledge node in the directed acyclic graph of prerequisite dependencies.

[0037] For example, in this embodiment of the application, the structural undercoverage gap is compared with a preset threshold. When the structural undercoverage gap is greater than the preset threshold, the target knowledge node is restricted from entering the downstream consumable state, and a pre-dependency compensation task associated with the target knowledge node is generated. The pre-dependency compensation task has a corresponding timeliness parameter, which is used to characterize the effective period of the pre-dependency compensation task.

[0038] During the effective period of the pre-dependency compensation task, this embodiment continuously detects candidate pre-dependency edges pointing to the target knowledge node and performs edge validity verification and acyclicity verification on the candidate pre-dependency edges. When both edge validity verification and acyclicity verification pass, the candidate pre-dependency edge is written into the pre-dependency directed acyclic graph, and the target knowledge node is released to enter the downstream consumable state.

[0039] It should be understood that, Figure 1The illustrated embodiments are merely illustrative of some embodiments of this application, demonstrating the use of a knowledge-based topology-oriented dependency gap identification and topology self-healing system 100 to perform dependency gap identification and topology self-healing operations. This knowledge-based topology-oriented dependency gap identification and topology self-healing system 100 is a laptop computer, and does not limit the structure, type, or number of such systems in other embodiments. For example, in other embodiments, the knowledge-based topology-oriented dependency gap identification and topology self-healing system 100 may also be a tablet computer, desktop computer, microcontroller, microcontroller, or other suitable type of device or apparatus. Furthermore, in other embodiments, the preceding dependency directed acyclic graph may also include... Figure 2 The preceding dependencies shown in the directed acyclic graph 10 have more or fewer nodes, or have more or fewer nodes than the preceding dependencies. Figure 2 The preceding dependency directed acyclic graph 10 shows different configurations.

[0040] The following will describe in detail the dependency gap identification and topology self-healing method for knowledge topology provided in this application, with reference to exemplary applications and implementations of the dependency gap identification and topology self-healing system for knowledge topology provided in the embodiments of this application.

[0041] Please see Figure 3A The dependency gap identification and topology self-healing method for knowledge topology provided in this application includes steps S31 to S37 to achieve dependency gap identification and topology self-healing of knowledge topology.

[0042] Step S31: Obtain the in-degree feature of the target knowledge node in the directed acyclic graph of the preceding dependencies and the ontology attribute identifier of the target knowledge node.

[0043] In this embodiment, the target knowledge node is any node in the knowledge topology. The knowledge topology includes a hierarchical knowledge skeleton and a directed acyclic graph of prerequisite dependencies, which is maintained separately from the hierarchical knowledge skeleton. The hierarchical knowledge skeleton represents the hierarchical relationship of knowledge nodes, and the directed acyclic graph of prerequisite dependencies represents the prerequisite dependencies between knowledge nodes. The hierarchical knowledge skeleton and the directed acyclic graph of prerequisite dependencies share knowledge node objects, but their edge types, computational responsibilities, and verification rules are separate.

[0044] In some embodiments, the knowledge topology is a knowledge mother tree system, which includes a hierarchical knowledge skeleton and a directed acyclic graph of pre-dependencies.

[0045] A hierarchical knowledge skeleton comprises multiple layers of knowledge nodes and categorical association edges between adjacent layers. For example, a hierarchical knowledge skeleton might include multiple layers of knowledge nodes from L0 to L3 or L0 to L4, along with categorical association edges between adjacent layers. Here, L0 represents the top-level knowledge domain, L1 represents a knowledge branch, L2 represents a topic domain, and L3 or L4 represents a knowledge node that can be computed, mapped, or used to generate questions. The categorical association edges between adjacent layers represent the hierarchical relationship of knowledge nodes and can be used to generate a knowledge coverage matrix / calculate knowledge coverage relationships.

[0046] The coverage matrix of the hierarchical knowledge skeleton is generated based on the classification association edges between adjacent layers. Pre-dependency edges in the directed acyclic graph of prerequisites do not participate in the chain multiplication calculation of the coverage matrix. The hierarchical knowledge skeleton corresponds to the knowledge mother tree and is used to represent the classification, hierarchical position, and coverage matrix of knowledge nodes.

[0047] The pre-dependency directed acyclic graph (DAG) is maintained independently of the classification association edges and is used to represent prerequisite dependencies, definition dependencies, program dependencies, difficulty propagation relationships, or progressive verification orders between knowledge nodes. Edges in the pre-dependency DAG do not participate in the chain multiplication calculation of the knowledge coverage matrix. The pre-dependency DAG includes pre-dependency edges between knowledge nodes at the same or different layers. These pre-dependency edges are used to calculate prerequisite relationships, difficulty propagation relationships, or progressive verification orders between knowledge nodes, and the classification association edges are stored, verified, and calculated separately from the pre-dependency edges. The pre-dependency DAG corresponds to the Prerequisite DAG and is used to represent prerequisite relationships, difficulty propagation, and progressive verification orders between knowledge nodes.

[0048] For example, in this embodiment of the application, any node in the knowledge topology is selected as the target knowledge node, and the in-degree feature of the target knowledge node in the preceding dependent directed acyclic graph and the ontology attribute identifier of the target knowledge node are obtained. For example, see Figure 2 As shown, node A in the pre-dependent directed acyclic graph 10 is selected as the target knowledge node, and the in-degree feature of node A in the pre-dependent directed acyclic graph 10 and the ontology attribute identifier of node A are obtained.

[0049] Step S32: When the in-degree feature satisfies the isolation condition and the ontology attribute identifier indicates that the target knowledge node is a derived knowledge node, the target knowledge node is identified as an isolated ecological node.

[0050] Understandably, the isolation condition is satisfied when the in-degree feature is 0, that is, when the in-degree feature of the target knowledge node is 0, the in-degree feature of the target knowledge node satisfies the isolation condition.

[0051] It should be understood that the ontology attribute identifier indicating that the target knowledge node is a derived knowledge node means that the ontology attribute identifier of the target knowledge node is derived. That is, when the ontology attribute identifier of the target knowledge node is derived, the target knowledge node is a derived knowledge node.

[0052] The ontology attribute identifier is determined based on the specification definition text of the target knowledge node. When the specification definition text depends on other knowledge nodes, mathematical objects, logical operators, process steps, or registered concepts, the ontology attribute identifier is determined to be of the derived type. When the specification definition text does not depend on other knowledge nodes but is introduced as a basic object, the ontology attribute identifier is determined to be of the primitive type.

[0053] For example, when the in-degree feature satisfies the isolation condition and the ontology attribute identifier indicates that the target knowledge node is a derived knowledge node, the target knowledge node is identified as an isolated ecosystem node. In this embodiment, if the canonical definition text of a certain L3 knowledge node indicates that it belongs to a derived knowledge node, but the in-degree feature of the L3 knowledge node in the preceding dependent directed acyclic graph is 0, then the L3 knowledge node is identified as an isolated ecosystem node.

[0054] Step S33: Obtain the representation load value and the prerequisite dependency complexity value of the target knowledge node, and calculate the structural undercover gap based on the representation load value and the prerequisite dependency complexity value.

[0055] For example, the structural undercover gap is determined according to the following formula: ,in, This indicates a gap in the structure's coverage. This represents the representational load value of the target knowledge node. This represents the complexity value of the prerequisite dependencies of the target knowledge node. This refers to the maximum value function; in this example, it refers to taking... (i.e., the difference between the characterization load value and the preceding dependency complexity value) and The maximum value in.

[0056] It should be understood that the representational loading value can be determined by at least one of the following factors: the abstraction level parameter of the knowledge node, the formal requirement parameter of the knowledge node, the expressive rigor of the knowledge node, the definitional dependency strength of the knowledge node, and the conceptual complexity of the knowledge node in the knowledge system. For example, the representational loading value can be determined based on the abstraction level parameter and the formal requirement parameter of the target knowledge node. The abstraction level parameter is used to represent the degree of abstraction of the knowledge object represented by the target knowledge node, and the formal requirement parameter is used to represent the rigor of the expressive form required to fully define the knowledge object.

[0057] The complexity value of prerequisite dependencies is determined based on at least one topological feature of the target knowledge node in the directed acyclic graph of prerequisite dependencies. Topological features include the number of direct prerequisite dependencies, the number of defining prerequisite dependencies, the number of procedural prerequisite dependencies, the depth of the longest forced path from the core root node to the target knowledge node, the number of reachable ancestor nodes, the number of ancestors across knowledge domains, and the confidence or structural morphology type of prerequisite dependency edges (e.g., chain-dominated, fan-in dominated, or hybrid branching).

[0058] It is understood that the representational load value is determined based on the abstraction level parameter and the formalization requirement parameter of the target knowledge node. For example, the representational load value is obtained by averaging the abstraction level parameter and the formalization requirement parameter. It should be understood that the abstraction level parameter refers to a quantitative score of the degree of abstraction of the concept (object) represented by the target knowledge node. In this embodiment, the abstraction level parameter is taken as any integer value in the range [1, 5]. The abstraction level parameter is used to characterize the degree of abstraction of the concept (or object) represented by the target knowledge node. An abstraction level parameter of 1 indicates that the concept (object) represented by the target knowledge node is a concrete fact / specific instance; an abstraction level parameter of 5 indicates that the concept (object) represented by the target knowledge node is a meta-theory / axiom.

[0059] The formal requirement parameter refers to a quantitative score of the rigor of the formal machine required to fully define the concept (object) represented by the target knowledge node. In this embodiment, the formal requirement parameter is set to any integer value in the range [1, 5]. The formal requirement parameter characterizes the rigor of the expression of the concept (object) represented by the target knowledge node. A formal requirement parameter of 1 indicates that the concept (object) represented by the target knowledge node can be expressed using natural language; a formal requirement parameter of 3 indicates that the concept (object) represented by the target knowledge node is expressed using a semi-formal canonical formula; and a formal requirement parameter of 5 indicates that the concept (object) represented by the target knowledge node is expressed using a fully axiomatic and verifiable formula.

[0060] It's worth noting that the abstract level parameter is a score obtained by quantifying the level of abstraction of the concept (object) represented by the node during the node creation and registration phase using semantic classification tools (such as Large Language Models (LLM) / classifiers). The formal requirement parameter is a score obtained by quantifying the rigor of the expression of the concept (object) represented by the node during the node creation and registration phase using semantic classification tools (such as Large Language Models (LLM) / classifiers). Both the abstract level parameter and the formal requirement parameter are fixed as read-only fields (or immutable fields) in the node's metadata. In the subsequent flow and computation of the entire knowledge topology, these underlying "genes" do not change with the addition or deletion of edges.

[0061] Here, the reachable ancestor of the target knowledge node refers to the node that can be reached by tracing back along the directed edge from the target knowledge node. For example, please refer to [link to relevant documentation]. Figure 2 The preceding dependent directed acyclic graph 10 shows that when node G is the target knowledge node, tracing back from node G along the directed edges (including directed edge 16, directed edge 14 and directed edge 11) in the opposite direction, the reachable nodes include node C, node B and node A. Therefore, node C, node B and node A are all reachable ancestors of node G, so the number of reachable ancestor nodes of node G is 3.

[0062] The longest forced path depth from the core root node to the target knowledge node refers to the number of directed edges in the longest path from the core root node to the target knowledge node. (See, for example...) Figure 3B The preceding dependency directed acyclic graph 20 shows node A as the core root node. When node E is the target knowledge node, there are two paths from the core root node (node ​​A) to the target knowledge node (node ​​E). The first path is A→B→C→E, and the second path is A→H→I→G→J→E. Clearly, the second path is the longest path. The second path A→H→I→G→J→E includes five directed edges: directed edge 207, directed edge 208, directed edge 209, directed edge 210, and directed edge 211. Therefore, the depth of the longest forced path is 5.

[0063] For example, in this application embodiment, a graph traversal algorithm (such as depth-first traversal algorithm, breadth-first traversal algorithm, etc.) is used to traverse the knowledge topology to obtain the number of reachable ancestor nodes of the target knowledge node and the longest forced path depth from the core root node to the target knowledge node. Based on the number of reachable ancestor nodes and the longest forced path depth from the core root node to the target knowledge node, the pre-dependency complexity value of the target knowledge node is calculated.

[0064] This application's embodiments calculate structural undercoverage gaps based on the representational load value and the preceding dependency complexity value of the target knowledge node. For example, in some embodiments, the gap is obtained by subtracting the preceding dependency complexity value from the representational load value and then taking the maximum value of 0; another example is that the representational load value and the preceding dependency complexity value are first mapped to a unified numerical range according to a preset mapping rule to obtain a first mapping value and a second mapping value, and then the gap is obtained by subtracting the second mapping value from the first mapping value and taking the maximum value of 0; yet another example is that the gap is obtained by weighting the representational load value and the preceding dependency complexity value and then taking the maximum value of 0. This application's embodiments do not impose any limitations on these methods.

[0065] Of course, other suitable methods can also be used to calculate the structural undercoverage gap based on the representation load value and the pre-dependency complexity value of the target knowledge node, and this application embodiment does not limit this in any way.

[0066] Step S34: When the structural undercoverage gap is greater than a preset threshold, restrict the target knowledge node from entering the downstream consumable state and generate a pre-dependency compensation task associated with the target knowledge node.

[0067] In this embodiment, the pre-dependency compensation task has a corresponding timeliness parameter, which is used to characterize the effective period of the pre-dependency compensation task. It is understood that engineers can customize and set preset thresholds based on experimental and empirical data; this embodiment does not impose any specific limitations on this.

[0068] In this embodiment, the timeliness parameter includes the starting version and the ending version. The validity period is the time from the starting version of the knowledge topology structure to the time when the version of the knowledge topology structure is updated to the ending version, that is, the time from the starting version of the knowledge topology structure to the time when it is updated to the ending version. Candidate edges are the preceding edges to be written to the knowledge topology structure and pointing to the target knowledge node.

[0069] The starting and ending versions of the dependency compensation task share the same versioning system as the knowledge topology. The starting version is a "snapshot copy" of the current version of the knowledge topology when the dependency compensation task is created; that is, the starting version of the dependency compensation task is the same as the current version of the knowledge topology. The ending version is obtained by adding the allowed iteration version window to the starting version. For example, if a knowledge topology allows 5 iterations, then the iteration version window is 5. In summary, the dependency compensation task itself does not have an independent version space. It marks two anchor points (the starting version and the ending version) on the version axis of the knowledge topology. The time between these two anchor points is the valid period, which is the time from the starting version to the ending version of the knowledge topology. Understandably, when the version of the continuously advancing knowledge topology surpasses the ending version (i.e., the version of the knowledge topology is updated to the next version after the ending version), it indicates that the valid period of the dependency compensation task has been exceeded, triggering a timeout isolation operation. For example, if the expiration version is V1.13 and the next version after that is V1.14, then when the knowledge topology is updated to version V1.14, it is determined that the validity period of the preceding dependency compensation task has expired.

[0070] For example, in this embodiment of the application, the structural undercoverage gap is compared with a preset threshold. When the structural undercoverage gap is greater than the preset threshold, the target knowledge node's permission to enter the downstream consumable state is restricted, and the target knowledge node is not allowed to enter the downstream consumable state. A pre-dependency compensation task associated with the target knowledge node is then generated. The pre-dependency compensation task is used to characterize the need to process the target knowledge node within the effective period characterized by the timeliness parameter.

[0071] The prerequisite dependency compensation task is used to complete the missing prerequisite dependencies of the target knowledge node. The effective period of the task includes the start time, end time and detection trigger frequency. The detection trigger frequency is used to control the scanning period of candidate prerequisite dependency edges.

[0072] It is worth noting that the downstream consumable state refers to the state in which a knowledge node can be called, traversed, calculated, or used to generate tasks by a question-generating system, evaluation system, recommendation system, intelligent agent routing system, knowledge reasoning system, or other downstream modules. Understandably, there are various ways to restrict the access of a target knowledge node to the downstream consumable state, and this application embodiment does not impose any limitations on this. For example, closing the access point of the target knowledge node, placing the target knowledge node in a locked or suspended state, temporarily restricts the target knowledge node's access to the downstream consumable state.

[0073] In this embodiment, the downstream consumable state is used to control whether the target knowledge node can be invoked by the question-generating system, evaluation system, recommendation system, intelligent agent routing system, knowledge reasoning system, or knowledge base retrieval enhancement system. Knowledge nodes and preceding dependencies in the knowledge topology use a structured registry as the single source of data authenticity. Before a knowledge node or preceding dependency enters the downstream consumable state, field pattern verification, source citation verification, graph integrity verification, maturity status verification, or downstream access qualification verification are performed on the structured registry.

[0074] As in the foregoing embodiments, this application further determines the representational load value based on the abstraction level parameters and formal requirement parameters of the L3 knowledge node, and determines the pre-dependency complexity value based on the number of its direct predecessor dependencies, the number of reachable ancestor nodes, the depth of the longest dependency path, and the number of ancestors across knowledge domains. If the representational load value is significantly higher than the pre-dependency complexity value, a structural undercover gap is generated, and the L3 knowledge node is restricted from entering the question generation system, evaluation system, or agent routing system (i.e., the L3 knowledge node is restricted from entering the downstream consumable state).

[0075] Step S35: During the effective period of the pre-dependency compensation task, detect candidate pre-dependency edges pointing to the target knowledge node.

[0076] For example, in response to a target knowledge node being restricted from entering a downstream consumable state, a pre-dependency compensation task corresponding to the target knowledge node is created, and a valid period is assigned to the pre-dependency compensation task. During the valid period of the pre-dependency compensation task, a set of candidate reference nodes associated with the target knowledge node is obtained. The knowledge nodes in the candidate reference node set are determined through at least one of the following methods: based on knowledge semantic vector similarity, based on knowledge topic tag overlap, based on knowledge concept reference relationships, based on historical learning path co-occurrence relationships, based on knowledge inference chain relationships, based on pre-knowledge recommendation results generated by a large language model, or based on the frequency of prior accesses in user learning behavior.

[0077] Obtain the node feature information of the target knowledge node and the reference feature information of each candidate reference node in the candidate reference node set. The node feature information includes at least one of the following: concept type, topic tag, number of formulas, number of parameters, derivation level, knowledge difficulty level, set of referenced entities, input / output semantics, and domain identifier. The reference feature information includes at least some of the above features corresponding to the candidate reference nodes.

[0078] Based on node feature information and reference feature information, a prior association score is calculated between each candidate reference node and the target knowledge node. The prior association score is used to characterize the degree of matching between the candidate reference node and the target knowledge node as prior knowledge. The prior association score is calculated based on at least one of the following: semantic dependency probability, inference citation ratio, concept coverage, probability of historical learning order, vector space proximity distance, graph path association strength, and user behavior statistical association degree.

[0079] In response to a prerequisite association score between a candidate reference node and a target knowledge node being greater than or equal to a preset association threshold, candidate prerequisite dependency edges are generated from the candidate reference node to the target knowledge node. These candidate prerequisite dependency edges characterize the candidate dependency relationship between the candidate reference node and the target knowledge node as prerequisite knowledge.

[0080] Candidate preceding dependency edges are written to the candidate edge cache queue, and the edge status corresponding to the candidate preceding dependency edge is marked as pending verification. The pending verification status is used to indicate that the candidate preceding dependency edge has not yet been written to the preceding dependency directed acyclic graph. The candidate edge cache queue is used to cache candidate preceding dependency edges for edge validity verification and acyclicity verification.

[0081] Step S36: Perform edge validity verification and acyclicity verification on the candidate prerequisite edges.

[0082] Among them, edge validity verification includes verifying whether the candidate preceding dependency edge has at least one of the following: target task family, dependency type, required mastery level, scope, blocking task identifier, test condition identifier, expected observable output, and failure rule.

[0083] Only pre-dependent edges that are active and have a confidence level greater than or equal to a pre-set confidence threshold are included in the core pre-dependent graph; pre-dependent edges that do not reach the pre-set confidence threshold are marked as pending review and are not included in the calculation of the pre-dependent complexity value.

[0084] In some embodiments, edge validity verification includes at least one of self-loop check, duplicate edge check, scope overlap check, blocking task overlap check, and prerequisite dependency loop check.

[0085] Cycle-free verification includes: performing a topology sorting algorithm or a strongly connected component detection algorithm on the core prerequisite graph after writing candidate prerequisite edges; when a directed cycle is detected in the core prerequisite graph, the writing of candidate prerequisite edges is rejected, and the downstream consumable state limit of the target knowledge node is maintained.

[0086] Retrieve candidate prerequisite dependencies that are in a pending verification state from the candidate edge cache queue, and parse the source knowledge node and target knowledge node corresponding to the candidate prerequisite dependency. The source knowledge node is the starting node of the candidate prerequisite dependency, and the target knowledge node is the ending node of the candidate prerequisite dependency. The candidate prerequisite dependency is used to represent the candidate dependency relationship where the source knowledge node serves as the prerequisite knowledge of the target knowledge node.

[0087] Next, obtain the source node attribute information of the source knowledge node and the target node attribute information of the target knowledge node. The source node attribute information and the target node attribute information include at least one of the following: knowledge type, difficulty level, derivation level, domain identifier, knowledge topic tag, input and output semantics, concept entity set, formula features, parameter features, and knowledge maturity status.

[0088] Based on the attribute information of the source node and the attribute information of the target node, edge semantic validity verification is performed. This edge semantic validity verification includes at least one of the following: verifying whether the source knowledge node and the target knowledge node belong to a compatible knowledge domain; verifying whether the difficulty level of the source knowledge node is lower than or equal to that of the target knowledge node; verifying whether the derivation level of the source knowledge node meets the prerequisite derivation requirements of the target knowledge node; verifying whether the conceptual entities in the source knowledge node cover the target conceptual entities in the target knowledge node; verifying whether the input semantics of the target knowledge node match the output semantics of the source knowledge node; and verifying whether the source knowledge node belongs to an allowed prerequisite type for the corresponding knowledge type of the target knowledge node.

[0089] In response to the successful verification of edge semantic validity, the graph structure index information of the preceding dependent directed acyclic graph is obtained. The graph structure index information includes at least one of the following: node adjacency list, node reachable path set, node topology level, node in-degree information, node out-degree information, and historical dependency path cache.

[0090] Based on graph structure index information, it detects whether a cycle is formed after writing candidate predecessor dependency edges into the directed acyclic graph of predecessor dependencies. For example, the acyclicity check includes: detecting whether the target knowledge node can be traced back to the source knowledge node through existing dependency paths; if the target knowledge node can be traced back to the source knowledge node, the writing result corresponding to the candidate predecessor dependency edge is determined to be a cycle; if the target knowledge node cannot be traced back to the source knowledge node, the writing result corresponding to the candidate predecessor dependency edge is determined to be acyclic.

[0091] If the write result corresponding to the candidate preceding dependency edge is acyclic, the candidate preceding dependency edge is determined to have passed the acyclicity check. If the edge semantic validity check or the acyclicity check fails, the edge state corresponding to the candidate preceding dependency edge is updated to the check failure state, and the corresponding failure reason identifier is recorded. The failure reason identifier includes at least one of the following: domain incompatibility, difficulty inversion, semantic mismatch, illegal inference relationship, cyclic conflict, duplicate dependency, or invalid path.

[0092] Step S37: When both edge validity and acyclicity checks pass, write the candidate predecessor dependency edge into the predecessor dependency directed acyclic graph and release the target knowledge node to enter the downstream consumable state.

[0093] In response to the candidate preceding dependency edges passing edge validity and acyclicity checks, the candidate preceding dependency edges are determined to be valid dependencies. Valid dependencies are used to represent preceding dependencies that satisfy semantic constraints and directed acyclic constraints, and include source knowledge node identifiers, target knowledge node identifiers, edge type identifiers, and edge confidence information.

[0094] Valid dependency edges are written into the edge storage structure corresponding to the directed acyclic graph of preceding dependencies, and the in-degree information of the target knowledge node is updated. The edge storage structure includes at least one of the following: an adjacency list, an edge index table, and a path cache. Updating the node in-degree information includes: increasing the number of incoming edges corresponding to the target knowledge node and updating the preceding dependency statistics of the target knowledge node.

[0095] Based on valid dependency edges, update the dependency coverage state corresponding to the target knowledge node. The dependency coverage state characterizes the completeness of the target knowledge node's preceding dependencies. The dependency coverage state includes at least one of the following: uncovered, partially covered, or fully covered.

[0096] Next, based on the updated dependency coverage state, the structural undercoverage gap corresponding to the target knowledge node is recalculated. The structural undercoverage gap characterizes the difference between the complexity of the knowledge representation of the target knowledge node and the degree of support from its preceding dependencies. The degree of support from preceding dependencies is determined based on at least one of the following: the number of effective preceding dependency edges, the coverage rate of preceding paths, the depth of preceding dependencies, and the completeness of preceding dependencies.

[0097] In response to the updated structural undercoverage gap being less than or equal to a preset gap threshold, the target knowledge node is determined to meet the downstream admission criteria. These downstream admission criteria characterize the target knowledge node's dependency integrity for entering downstream systems, which include at least one of the following: a question-generating system, an agent routing system, a knowledge reasoning system, a course generation system, and a resource scheduling system.

[0098] In response to the target knowledge node meeting the downstream access conditions, the node status corresponding to the target knowledge node is updated. Specifically, the node status is updated from restricted consumption state to consumable state.

[0099] The target knowledge node is released into a downstream consumable state, and a node status change notification is sent to the downstream system. This node status change notification indicates that the target knowledge node is now eligible for consumption. In response to the node status change notification, the downstream system adds the target knowledge node to its callable knowledge set, reasonable knowledge set, or learnable knowledge set.

[0100] Close the prerequisite dependency compensation task corresponding to the target knowledge node and record the topology self-healing result. The topology self-healing result includes at least one of the following: the number of newly added effective dependency edges, the degree of gap convergence, the compensation time, the self-healing success indicator, and the state release time.

[0101] As described in the previous embodiment, when a candidate prerequisite dependency edge pointing to the L3 knowledge node is written in the subsequent knowledge base construction process, it is checked whether the candidate prerequisite dependency edge contains metadata such as target task family, dependency type, blocking task identifier, test condition identifier, expected observable output, failure rules, and edge confidence. An acyclic verification is then performed on the directed acyclic graph of the prerequisite dependencies after the writing. If the verification passes, the candidate prerequisite dependency edge is added to the core prerequisite dependency graph, the prerequisite dependency complexity value of the L3 knowledge node is updated, and the state of the L3 knowledge node is updated from an isolated ecosystem to a normal derived node.

[0102] This application identifies isolated ecosystem nodes that are derived in the ontology but lack preceding dependency edges in the topology by obtaining the in-degree feature and ontology attribute identifier of the target knowledge node in the preceding dependency directed acyclic graph. Further, it calculates the structural undercoverage gap based on the representational load value and preceding dependency complexity value of the target knowledge node, and restricts the target knowledge node from entering the downstream consumable state when the structural undercoverage gap exceeds a threshold, generating a preceding dependency compensation task. After detecting a candidate preceding dependency edge pointing to the target knowledge node, the system performs edge validity verification and acyclicity verification, and updates the preceding dependency directed acyclic graph after the verification passes, releasing the target knowledge node to enter the downstream consumable state. Therefore, this application can prevent derived knowledge nodes in the knowledge topology from being misjudged as legitimate root nodes, improving the topological integrity and operational stability in knowledge base construction, knowledge reasoning, agent routing, and automated question generation scenarios.

[0103] In some implementations, the embodiments of this application achieve the acquisition of ontology attribute identifiers of target knowledge nodes through steps S311 to S313.

[0104] Step S311: Parse the specification definition text corresponding to the target knowledge node.

[0105] Step S312: The response is to set the standard definition text corresponding to the target knowledge node as the first standard text, and assign the ontology attribute identifier of the target knowledge node to the first identifier.

[0106] Step S313: Respond to the second specification text corresponding to the target knowledge node, and assign the second identifier to the ontology attribute identifier of the target knowledge node.

[0107] In this context, a reference node is any node in the knowledge topology other than the target knowledge node. The first specification text is text that depends on the specification definition text corresponding to the reference node, and the second specification text is text that does not depend on the specification definition text corresponding to the reference node. The specification definition text corresponding to the target knowledge node refers to the text defining the semantics of the target knowledge node.

[0108] Understandably, the first specification text, besides depending on the specification definition text corresponding to the reference node, can also depend on other mathematical objects or logical operators, etc. That is, text that depends on the specification definition text corresponding to the reference node or other mathematical objects or logical operators, etc., is the first specification text. Similarly, text that does not depend on the specification definition text corresponding to the reference node, and text that does not depend on other mathematical objects or logical operators, etc., is the second specification text.

[0109] For example, in this embodiment of the application, the specification definition text corresponding to the target knowledge node is obtained and parsed. If the specification definition text corresponding to the target knowledge node is a first specification text, the ontology attribute identifier of the target knowledge node is assigned the first identifier. The first identifier is used to indicate that the target knowledge node is a node of a derived type (Derived Object) and depends on the specification definition text of other nodes. If the specification definition text corresponding to the target knowledge node is a second specification text, the ontology attribute identifier of the target knowledge node is assigned the second identifier. The second identifier is used to indicate that the target knowledge node is a node of a primitive type (Primitive Root) and does not depend on the specification definition text of other nodes.

[0110] In some implementations, the embodiments of this application realize the calculation of structural undercoverage gaps based on the characterization load value and the preceding dependency complexity value through steps S331 to S332.

[0111] Step S331: Subtract the preceding dependency complexity value from the characterization load value to obtain the first difference.

[0112] Step S332: Determine the maximum value between the first difference and the preset value as the structural undercover gap.

[0113] In this step, the preset value is 0. For example, in this embodiment, the representational load value is subtracted from the preceding dependency complexity value to obtain a first difference, which is the difference between the representational load value and the preceding dependency complexity value. The first difference is compared with a preset value, and the maximum value between the first difference and the preset value is determined to be the structural undercoverage gap of the target knowledge node.

[0114] The following examples illustrate the dependency gap identification and topology self-healing method for knowledge topology provided in this application, which involves intercepting high-risk isolated nodes and allowing low-risk under-coverage gap nodes through alarms.

[0115] First, let's take the interception of the high-risk isolated knowledge node with ID Node_Q21 as an example.

[0116] Among them, the in-degree feature of node Node_Q21 is 0, and the ontology attribute identifier is the first identifier (indicating that node Node_Q21 is a derived knowledge node). The representation load value and the pre-dependency complexity value of node Node_Q21 are 4.50 and 1.00, respectively. The preset threshold is set to 1.5, and the preset value is 0.

[0117] The first difference is obtained by calculating the difference between the characterization load value and the pre-dependency complexity value. The difference between the preset value 0 and the first value The maximum value is 3.50, therefore the structural undercoverage gap of node Node_Q21 is 3.50.

[0118] If the structural undercoverage gap of node_Q21 is greater than the preset threshold of 1.5, then node_Q21 is restricted from entering the downstream consumable state, and a pre-dependency compensation task associated with node_Q21 is generated.

[0119] During the effective period of the pre-dependency compensation task, if a candidate pre-dependency edge pointing from node Node_D21 to node Node_Q21 is detected, edge validity and acyclicity checks are performed on the candidate pre-dependency edge. When both edge validity and acyclicity checks of the candidate pre-dependency edge pass, the candidate pre-dependency edge is written into the pre-dependency directed acyclic graph, and node Node_Q21 is released to enter the downstream consumable state.

[0120] Furthermore, let's take the low-risk structural undercoverage gap knowledge node with alarm release ID Node_Q22 as an example for illustration.

[0121] Among them, the in-degree feature of node Node_Q22 is zero, and the ontology attribute identifier is the first identifier (indicating that node Node_Q22 is a derived knowledge node). The representation load value and the pre-dependency complexity value of node Node_Q22 are 2.20 and 1.00, respectively. The preset threshold is set to 1.5, and the preset value is 0.

[0122] The first difference is obtained by calculating the difference between the characterization load value and the pre-dependency complexity value. The difference between the preset value 0 and the first value The maximum value is 1.20, therefore the structural undercoverage gap of node Node_Q22 is 1.20.

[0123] The comparison shows that the structural undercoverage gap of node Node_Q22 is less than the preset threshold of 1.5, which does not exceed the preset threshold. Therefore, it enters the low-risk alarm path instead of the strong blocking path, that is, it performs the low-risk alarm operation and allows node Node_Q22 to write to the knowledge topology structure.

[0124] For low-risk structural coverage gap nodes, the system allows such nodes to remain in a readable or limitedly available state, but retains a low-risk alarm mark in the node metadata for unified review by subsequent scanning tasks. This proves that the embodiments of this application are not only applicable to high-risk strong interception scenarios, but also to low-risk audit release scenarios.

[0125] In summary, this application identifies isolated ecosystem nodes that are derived in the ontology but lack preceding dependency edges in the topology by obtaining the in-degree feature and ontology attribute identifier of the target knowledge node in the preceding dependency directed acyclic graph. Furthermore, it calculates the structural undercoverage gap based on the representational load value and preceding dependency complexity value of the target knowledge node, and restricts the target knowledge node from entering the downstream consumable state when the structural undercoverage gap exceeds a threshold, generating a preceding dependency compensation task. After detecting a candidate preceding dependency edge pointing to the target knowledge node, the system performs edge validity verification and acyclicity verification, and updates the preceding dependency directed acyclic graph after the verification passes, releasing the target knowledge node to enter the downstream consumable state. Therefore, this application can prevent derived knowledge nodes in the knowledge topology from being misjudged as legitimate root nodes, improving the topological integrity and operational stability in knowledge base construction, knowledge reasoning, agent routing, and automated question generation scenarios.

[0126] As another aspect of the embodiments of this application, this application provides a corresponding dependency gap identification and topology self-healing system for knowledge topology structures. The dependency gap identification and topology self-healing system for knowledge topology structures can be a software module. This software module includes several instructions stored in a memory. A processor can access the memory and execute the instructions to complete the dependency gap identification and topology self-healing methods for knowledge topology structures described in the various embodiments above.

[0127] In some feasible implementations, the dependency gap identification and topology self-healing system for knowledge topology structures can also be constructed using hardware devices. For example, the system can be constructed using one or more chips, which can coordinate with each other to implement the dependency gap identification and topology self-healing methods for knowledge topology structures described in the various implementations above. In some embodiments, the system can also be constructed using various logic devices, such as general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), microcontrollers, field-programmable gate arrays (FPGAs), ARM (Acorn RISC Machine) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or any combination of these devices or components.

[0128] Please see Figure 4 , Figure 4 The schematic diagram illustrates the structure of a knowledge-based topology dependency gap identification and topology self-healing system provided in some embodiments of this application.

[0129] Please see Figure 4As shown, the dependency gap identification and topology self-healing system 200 for knowledge topology includes a knowledge topology management module 210, a node semantic recognition module 220, a local topology state recognition module 230, a gap quantization module 240, a compensation task generation module 250, an edge monitoring and verification module 260, and a topology self-healing update module 270.

[0130] Specifically, the knowledge topology management module 210 is used to maintain the hierarchical knowledge skeleton and the directed acyclic graph of pre-dependencies maintained separately from the hierarchical knowledge skeleton. The node semantic recognition module 220 is used to determine the ontology attribute identifier of the target knowledge node based on the canonical definition text of the target knowledge node. The local topology state recognition module 230 is used to obtain the in-degree feature of the target knowledge node in the directed acyclic graph of pre-dependencies, and when the in-degree feature satisfies the isolation condition and the ontology attribute identifier is derived, the target knowledge node is identified as an isolated ecological node. The gap quantification module 240 is used to calculate the structural undercover gap based on the representation load value and the pre-dependency complexity value of the target knowledge node. The compensation task generation module 250 is used to restrict the target knowledge node from entering the downstream consumable state and generate a pre-dependency compensation task when the structural undercover gap is greater than a preset threshold. The edge monitoring and verification module 260 is used to detect candidate pre-dependency edges pointing to the target knowledge node and perform edge legality verification and acyclicity verification on the candidate pre-dependency edges. The topology self-healing update module 270 is used to write the candidate predecessor dependency edge into the predecessor dependency directed acyclic graph and release the target knowledge node into the downstream consumable state when both the edge validity check and the acyclic check pass.

[0131] It should be noted that, for the sake of simplicity and brevity, the aforementioned dependency gap identification and topology self-healing system for knowledge topology structures can execute the corresponding functional modules and achieve the corresponding beneficial effects of the dependency gap identification and topology self-healing method for knowledge topology structures provided in the embodiments of this application. Technical details not described in detail in the embodiments of the dependency gap identification and topology self-healing system for knowledge topology structures can be found in the dependency gap identification and topology self-healing method for knowledge topology structures provided in the embodiments of this application. The specific working process of the aforementioned dependency gap identification and topology self-healing system for knowledge topology structures can also be found in the specific execution process of the dependency gap identification and topology self-healing method for knowledge topology structures provided in the foregoing embodiments of this application, and will not be elaborated upon here.

[0132] This application provides a computer-readable storage medium storing processor-executable computer program instructions. When executed by a processor, the computer program instructions cause the processor to perform the dependency gap identification and topology self-healing method for knowledge topology provided in this application, or to perform the steps in any possible implementation of the dependency gap identification and topology self-healing method for knowledge topology provided in this application.

[0133] Those skilled in the art will understand that the embodiments provided in this application are merely illustrative. The order in which the steps in the methods of the embodiments are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The order can be adjusted, merged, and deleted according to actual needs. Modules or sub-modules, units or sub-units in the apparatus or system of the embodiments can be merged, divided, and deleted according to actual needs. For example, the division of units is only a logical functional division, and there may be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed.

[0134] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, and of course, it can also be implemented using hardware. Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. It should be understood that the storage medium can be flash memory, hard disk, optical disk, register, magnetic surface memory, removable disk, CD-ROM, random access memory (RAM), read-only memory (ROM), electrically programmable ROM, and electrically erasable programmable ROM, etc.

[0135] It should be noted that the above embodiments are for illustrating the technical concept and features of this application, and are intended to enable those skilled in the art to understand the content of this application and implement it accordingly. They should not be construed as limiting the scope of protection of this application. Those skilled in the art can understand that all or part of the processes of the above embodiments can be implemented, modified according to the technical solutions described in the embodiments of this application, or equivalent substitutions can be made to some of the technical features. It is understood that these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should be considered as equivalent changes and modifications made based on the embodiments of this application, all of which should fall within the scope of the claims of this application.

Claims

1. A method for dependency gap identification and topology self-healing oriented towards knowledge topology structures, characterized in that, The knowledge topology includes a hierarchical knowledge skeleton and a directed acyclic graph of prerequisite dependencies maintained separately from the hierarchical knowledge skeleton. The hierarchical knowledge skeleton is used to represent the hierarchical affiliation of knowledge nodes, and the directed acyclic graph of prerequisite dependencies is used to represent the prerequisite dependencies between knowledge nodes. The method includes: Obtain the in-degree feature of the target knowledge node in the preceding dependent directed acyclic graph and the ontology attribute identifier of the target knowledge node; When the in-degree feature satisfies the isolation condition and the ontology attribute identifier indicates that the target knowledge node is a derived knowledge node, the target knowledge node is identified as an isolated ecological node. Obtain the representation load value and the prerequisite dependency complexity value of the target knowledge node, and calculate the structural undercover gap based on the representation load value and the prerequisite dependency complexity value; When the structural undercoverage gap is greater than a preset threshold, the target knowledge node is restricted from entering the downstream consumable state, and a pre-dependency compensation task associated with the target knowledge node is generated. During the effective period of the aforementioned dependency compensation task, candidate preceding dependency edges pointing to the target knowledge node are detected. Perform edge validity verification and acyclicity verification on the candidate pre-dependencies; When the edge validity check and acyclicity check both pass, the candidate prerequisite dependency edge is written into the prerequisite dependency directed acyclic graph, and the target knowledge node is released to enter the downstream consumable state.

2. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The hierarchical knowledge skeleton includes multiple layers of knowledge nodes and classification association edges between adjacent layers. The pre-dependency directed acyclic graph includes pre-dependency edges between knowledge nodes in the same layer or across layers. The classification association edges are used to calculate knowledge coverage relationships, and the pre-dependency edges are used to calculate prerequisite relationships, difficulty propagation relationships, or progressive verification order between knowledge nodes. Furthermore, the classification association edges and the pre-dependency edges are stored, verified, and calculated separately.

3. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The coverage matrix of the hierarchical knowledge skeleton is generated based on the classification association edges of adjacent layers. The preceding dependency edges in the directed acyclic graph of the preceding dependencies do not participate in the chain multiplication calculation of the coverage matrix.

4. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The ontology attribute identifier is determined based on the specification definition text of the target knowledge node; when the specification definition text depends on other knowledge nodes, mathematical objects, logical operators, process steps, or registered constructs, the ontology attribute identifier is determined to be a derived type; when the specification definition text does not depend on other knowledge nodes but is introduced as a basic object, the ontology attribute identifier is determined to be a primitive type.

5. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The structural under-coverage gap is determined according to the following formula: in, This indicates that the structure has an under-coverage gap. This represents the representational load value of the target knowledge node. This represents the complexity value of the prerequisite dependencies of the target knowledge node.

6. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 5, characterized in that, The representational load value is determined based on the abstraction level parameter and formalization requirement parameter of the target knowledge node; the abstraction level parameter is used to represent the degree of abstraction of the knowledge object represented by the target knowledge node, and the formalization requirement parameter is used to represent the strictness of the expression form required to fully define the knowledge object.

7. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 5, characterized in that, The complexity value of the prerequisite dependency is determined based on at least one topological feature of the target knowledge node in the directed acyclic graph of the prerequisite dependencies. The topological features include the number of direct prerequisite dependencies, the number of defining prerequisite dependencies, the number of procedural prerequisite dependencies, the depth of the longest forced path from the core root node to the target knowledge node, the number of reachable ancestor nodes, the number of ancestors across knowledge domains, the confidence of the prerequisite dependency edges, or the structural morphology type.

8. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The edge validity verification includes: verifying whether the candidate preceding dependency edge has at least one of the following: target task family, dependency type, required mastery level, scope, blocking task identifier, test condition identifier, expected observable output, and failure rule.

9. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, Only pre-dependent edges that are active and have a confidence level greater than or equal to a preset confidence threshold are included in the core pre-dependent graph; pre-dependent edges that do not reach the preset confidence threshold are marked as pending review and are not included in the calculation of the pre-dependent complexity value.

10. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The edge validity check includes at least one of the following: self-loop check, duplicate edge check, scope overlap check, blocking task overlap check, and prerequisite dependency loop check.

11. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The acyclic verification includes: performing a topological sorting algorithm or a strongly connected component detection algorithm on the core prerequisite dependency graph after writing candidate prerequisite dependency edges; when a directed loop is detected in the core prerequisite dependency graph, rejecting the writing of the candidate prerequisite dependency edge and maintaining the downstream consumable state limit of the target knowledge node.

12. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The knowledge nodes and preceding dependent edges in the knowledge topology use a structured registry as the single source of data authenticity. Before a knowledge node or preceding dependent edge enters the downstream consumable state, the structured registry is subjected to field pattern verification, source reference verification, graph integrity verification, maturity status verification, or downstream access qualification verification.

13. The dependency gap identification and topology self-healing method for knowledge topology structures according to claim 1, characterized in that, The downstream consumable status is used to control whether the target knowledge node can be invoked by the question-generating system, evaluation system, recommendation system, intelligent agent routing system, knowledge reasoning system, or knowledge base retrieval enhancement system.

14. A dependency gap identification and topology self-healing system for knowledge topology structures, characterized in that, include: The knowledge topology management module is used to maintain the hierarchical knowledge skeleton and the directed acyclic graph of pre-dependencies that is maintained separately from the hierarchical knowledge skeleton. The node semantic recognition module is used to determine the ontology attribute identifier of the target knowledge node based on the standardized definition text of the target knowledge node; The local topology identification module is used to obtain the in-degree feature of the target knowledge node in the preceding dependent directed acyclic graph, and when the in-degree feature satisfies the isolation condition and the ontology attribute is a derived type, the target knowledge node is identified as an isolated ecological node. The gap quantification module is used to calculate the structural undercover gap based on the representation load value and the pre-dependency complexity value of the target knowledge node; The compensation task generation module is used to restrict the target knowledge node from entering the downstream consumable state and generate a pre-dependent compensation task when the structural undercoverage gap is greater than a preset threshold. The edge monitoring and verification module is used to detect candidate preceding dependency edges pointing to the target knowledge node, and to perform edge validity verification and acyclicity verification on the candidate preceding dependency edges. The topology self-healing update module is used to write the candidate predecessor dependency edge into the predecessor dependency directed acyclic graph and release the target knowledge node into the downstream consumable state when both the edge validity check and the acyclic check pass.

Images