Knowledge management and expression method

Abstract

The invention provides a knowledge management and expression method, which relates to the technical field of data information management, and comprises the steps of judging whether a mode layer is matched with knowledge data, generating a graph structure framework, obtaining core characteristics of the mode layer, generating a knowledge graph containing knowledge point nodes, association relationship edges and visual elements, and the like. Aiming at the problem that an existing knowledge graph is weak in interpretability, abstract knowledge features are converted into intuitive visual elements through a multi-dimensional visualization rule, specific data are recorded in cooperation with a node feature form, and meanwhile logic compliance is ensured through verification of the node text matching degree and the knowledge relevance index; the problem of noise sensitivity is solved, invalid correlation is filtered through matching degree and correlation index thresholds, isolated nodes are eliminated, noises such as entity connection errors and relation ambiguity are effectively reduced, and the accuracy and reliability of the atlas are guaranteed; the graph is split into the connected sub-graphs to realize block management, so that the query delay is greatly reduced, and the use efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of data information management technology, specifically to a knowledge management and representation method. Background Technology

[0002] In the field of knowledge management, the core need lies in transforming scattered, fragmented information into a knowledge network with structured organization, relational mapping, and reasoning capabilities, thereby supporting the efficient implementation of business decisions, information retrieval, and intelligent applications. Knowledge graphs, as a key technology for representing and reasoning about structured knowledge, have been widely applied in various fields such as search engines, intelligent question answering, and recommendation systems. Essentially, it is a structured semantic knowledge base that clearly describes the relationships and attribute information between entities in the real world through a graph structure, providing a basic framework for the systematic storage, transmission, and application of knowledge.

[0003] Current knowledge graph applications suffer from the following problems: weak interpretability, with simple vectorized representations failing to intuitively express business needs; sensitivity to noise, leading to incorrect entity connections and ambiguous relation extraction; and low data query efficiency, with high latency for ultra-large-scale graph queries. To address these issues, a knowledge management and representation method is proposed. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a knowledge management and representation method that can make business knowledge explicit through different forms of expression, making the relationships between business knowledge and the reasoning process more transparent. This is especially true in highly specialized fields or business scenarios, where interpretability and traceability are enhanced. The value of business knowledge is reflected by the strength of management relationships, the frequency of use of knowledge points, and the importance of the use scenario, thereby promoting the efficient utilization of business knowledge in actual business scenarios.

[0005] The technical solution adopted in this invention is as follows: A knowledge management and representation method, comprising the following steps: S1. Obtain knowledge data entities, construct a pattern layer of a knowledge graph based on the knowledge data entities, and calculate the matching degree between the knowledge data entity node text and the pattern layer node text. When the matching degree between the knowledge data entity node text and the pattern layer node text is greater than a set matching threshold, it is determined that the pattern layer is compatible with the knowledge data entity. S2. Perform position mapping on the pattern layer to generate a graph structure framework, calculate the knowledge relevance index of the pattern layer, and determine that the graph structure framework is qualified for association when the knowledge relevance index is greater than the relevance threshold; at the same time, obtain the core features of the pattern layer, including usage frequency and scene importance. S3. Generate a knowledge graph containing knowledge point nodes, relational edges, and visualization elements based on the graph structure framework, pattern layer knowledge relevance index, and core features; preset multi-dimensional visualization rules and determine the correspondence between the multi-dimensional visualization rules and the visualization element knowledge graph; S4. Exclude isolated nodes in the knowledge graph of visualized elements in S3, randomly select a non-isolated node, search for its connected nodes by traversal, and mark it as a connected subgraph; repeat the traversal process until all non-isolated nodes are marked, and obtain multiple connected subgraphs. S5. Calculate the spatial attributes of the knowledge data entity graph in each connected subgraph. The spatial attributes of the knowledge data entity graph include the number of nodes associated with each node, the average length of the shortest path from a node to other nodes, and the number of times a node acts as a bridge for the shortest path from other nodes. Input the calculated number of nodes associated with each node, the average length of the shortest path from a node to other nodes, the number of times a node acts as a bridge for the shortest path from other nodes, the usage frequency obtained in step S2, and the scene importance into the LR model to calculate the knowledge data entity score. Compare the calculated knowledge data entity score with the set knowledge data entity threshold, mark important knowledge data entities, and divide them into hierarchical entities based on important knowledge data entities.

[0006] Further, in step S1, obtaining knowledge data entities involves extracting knowledge data entities from the knowledge data using an entity recognition algorithm and extracting the relationships between the knowledge data entities using a relation extraction algorithm; the entity recognition algorithm is the BERT-CRF algorithm, and the relation extraction algorithm is the PCNN algorithm.

[0007] Further, in step S1, the formula for calculating the node text matching degree is: In the formula, The number of the preset time period. For the first The text matching degree of nodes within a preset time period. It is a natural constant. For the first The string matching degree for a preset time period. For string reference matching degree, For the first Semantic similarity over a preset time period For semantic reference similarity, For the first Attribute validation values ​​for a preset time period. This is a reference validation value for the attribute. This is a correction factor for attribute validation.

[0008] Further, in step S2, the formula for calculating the knowledge relevance index is: In the formula, The number of the preset time period. For the first A knowledge relevance index for a preset time period. It is a natural constant. For the first Entity matching degree over a preset time period For entity reference matching degree, For the first Consistency of relationships within a preset time period For the sake of relational reference consistency, For the first The concept of coverage for a preset time period.

[0009] Furthermore, in step S2, the multi-dimensional visualization rules are as follows: Node size mapping: Five fixed node sizes are preset. The maximum usage frequency of all knowledge data entities is obtained. Frequency intervals are divided by the maximum usage frequency divided by 5. Knowledge point nodes whose usage frequency falls into different intervals are matched with the corresponding sizes. If the maximum usage frequency is <10, all knowledge point nodes adopt the size corresponding to the maximum interval. Node color mapping: Five gradient colors are preset. The maximum scene importance index of all knowledge data entities is obtained. The importance interval is divided by the maximum index by 5. Knowledge point nodes whose scene importance falls into different intervals are matched with the corresponding colors. If the maximum index is <10, all knowledge point nodes adopt the color corresponding to the maximum interval. Node line number mapping: Three types of node description lines are preset: thin line, medium thick line and thick line, which correspond to the low, medium and important levels of the scene importance of knowledge data entities, respectively. After determining the level, the corresponding line is matched. Relationship line type mapping: Four preset relationship line types are used, and the corresponding line type is matched after the type is determined according to the degree of business process dependency. Feature form configuration: Configure a small form for each knowledge point node to record specific values ​​or field descriptions of usage frequency and scenario importance; In the acquired knowledge graph, the visualization elements correspond to the core features of the knowledge points.

[0010] Further, in step S4, obtaining the connected subgraph specifically includes the following steps: S41: Traverse the knowledge graph of visualized elements in S3, exclude isolated nodes with no relation to each other, and mark the remaining non-isolated nodes to form a set of non-isolated nodes; S42: Randomly select an initial node from the set of non-isolated nodes, and mark the initial node and its connected nodes as a connected subgraph by traversing and searching all nodes that are directly or indirectly connected to the initial node; S43: Repeat the operation of S42 until all nodes in the set of non-isolated nodes are marked, and obtain multiple connected subgraphs.

[0011] Further, in step S5, the method for dividing the hierarchical entities is as follows: taking the important knowledge data entity as the core, a hierarchical entity set is divided according to the relational distance between the node and the important knowledge data entity. If there is no knowledge data entity with an entity score ≥ 0.5 in a certain connected subgraph, then the knowledge data entity with the highest entity score in the connected subgraph is set as the important knowledge data entity; taking the important knowledge data entity as the core level and setting it as level 0, traversing the first-degree relation nodes of the important knowledge data entity and setting them as level 1, traversing the first-degree relation nodes of the level 1 node and setting them as level 2, until there are no new relation nodes to traverse, and obtaining the hierarchical entities containing different levels.

[0012] Furthermore, after obtaining the hierarchical entities containing different levels, breadcrumb navigation is used during the drilling process to maintain knowledge path coordination.

[0013] The beneficial effects of this invention are: This invention addresses the problem of weak interpretability in existing knowledge graphs by transforming abstract knowledge features into intuitive visual elements through multi-dimensional visualization rules, and recording specific data in conjunction with node feature forms. At the same time, it ensures logical compliance through verification of node text matching degree and knowledge relevance index. To address noise sensitivity, invalid associations are filtered out and isolated nodes are excluded by using matching degree and relevance index thresholds, effectively reducing noise such as entity connection errors and relationship ambiguities, and ensuring the accuracy and reliability of the graph. To address the pain point of high query latency, the graph is split into connected subgraphs for block management. An LR model is used to calculate entity scores to mark important knowledge data entities. During queries, important knowledge data entities and related nodes are loaded first, significantly reducing query latency and improving efficiency. Attached Figure Description

[0014] Figure 1 This is a flowchart of the knowledge management and representation method according to an embodiment of the present invention; Figure 2 This is a flowchart of a connected subgraph according to an embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] like Figure 1 and Figure 2 As shown in the figure, a knowledge management and representation method according to an embodiment of the present invention includes the following steps: S1. Obtain knowledge data entities, construct a pattern layer of the knowledge graph based on the knowledge data entities, and calculate the matching degree between the knowledge data entity node text and the pattern layer node text. When the matching degree between the knowledge data entity node text and the pattern layer node text is greater than a set matching threshold, it is determined that the pattern layer is compatible with the knowledge data entity.

[0017] Knowledge data is acquired by extracting knowledge data entities through entity recognition algorithms and extracting the relationships between knowledge data entities through relation extraction algorithms; the entity recognition algorithm is BERT-CRF algorithm and the relation extraction algorithm is PCNN algorithm.

[0018] The BERT-CRF algorithm uses a pre-trained BERT model to semantically encode knowledge data text and combines it with a CRF layer to annotate entity boundaries; the PCNN relation extraction algorithm uses a segmented convolutional neural network to extract local features from text fragments containing entity pairs.

[0019] The node text matching score is calculated as follows: First, the node text in the knowledge data and the node text in the pattern layer are segmented to obtain the string matching score. Then, the entity semantic information in the two types of node text is identified, and the semantic similarity is obtained based on the cosine similarity of the vector space. Next, the attribute values ​​corresponding to the target node text attributes are extracted and attribute verification is performed to obtain the attribute verification value. Finally, the string matching score, semantic similarity, and attribute verification value are fused using a preset formula to obtain the node text matching score. The calculation formula is as follows: In the formula, The number of the preset time period. For the first The text matching degree of nodes within a preset time period. It is a natural constant. For the first The string matching degree for a preset time period. For string reference matching degree, For the first Semantic similarity over a preset time period For semantic reference similarity, For the first Attribute validation values ​​for a preset time period. This is a reference validation value for the attribute. The attribute validation correction factor is set to a matching threshold of 0.8.

[0020] S2. Perform position mapping on the pattern layer to generate a graph structure framework, calculate the knowledge correlation index of the pattern layer, and determine that the graph structure framework is qualified for association when the knowledge correlation index is greater than the correlation threshold; at the same time, obtain the core features of the pattern layer, including usage frequency and scene importance.

[0021] The core features, including frequency of use and importance of scenarios, were obtained by extracting them from the operation logs of the knowledge management platform. Specifically: Usage frequency includes the call records of knowledge data entities, including the number of times the entity is queried, viewed, and referenced by users; and the number of times the entity is automatically called when embedded in business processes. The importance of a scenario includes the number of users and the number of times it is shared.

[0022] The knowledge relevance index is calculated as follows: First, concept coverage, entity matching degree, and relationship consistency are calculated; then, the knowledge relevance index is calculated using a preset formula, which is: In the formula, The number of the preset time period. For the first A knowledge relevance index for a preset time period. It is a natural constant. For the first Entity matching degree over a preset time period For entity reference matching degree, For the first Consistency of relationships within a preset time period For the sake of relational reference consistency, For the first The concept coverage rate for a preset time period is set with an association threshold of 0.8.

[0023] S3. Generate a knowledge graph containing knowledge point nodes, relational edges, and visualization elements based on the graph structure framework, pattern layer knowledge relevance index, and core features; preset multi-dimensional visualization rules and determine the correspondence between the multi-dimensional visualization rules and the visualization element knowledge graph.

[0024] The correspondence between multi-dimensional visualization rules and the knowledge graph of visualization elements is that the core features of the pattern layer are transformed into visualization elements of the knowledge graph through pre-defined quantified multi-dimensional visualization rules. The specific calculation method for the correspondence of multi-dimensional visualization rules is as follows: Node size mapping calculation: 5 preset fixed sizes ), obtain the maximum usage frequency of all knowledge data entities, and match them according to the usage frequency range. The frequency range is calculated as follows: interval in, The frequency of use of a single knowledge data entity (number of times it is called per unit of time). The maximum usage frequency for all entities. · It is the floor function; if All nodes adopt the same method. To avoid the problem of the size being too small to be recognized.

[0025] Node color mapping calculation: 5 preset gradient colors ( Matching is performed based on the importance interval of the scenario. The importance interval is calculated as follows: interval in, This refers to the contextual importance indicators for individual knowledge data entities. Contextual importance indicators include the number of users and the number of times they are shared. The most important indicator for all entities; if All nodes adopt the same method. This ensures that importance is visually distinguished.

[0026] Node line number mapping: 3 preset line number options Matching based on scenario importance indicators; scenario importance indicators Scenario Importance Indicators Scenario Importance Indicators .

[0027] Relationship line type mapping: The relationship line type mapping presets four line types: solid line, dashed line, dotted line, and double solid line. The relationship line types include four types of relationships: process to process, process to data, process to message, and process to AI.

[0028] When the data types of the two related knowledge data entities are both processes, the process-to-process relationship is triggered, and the line mapping of the relationship is set to a solid line. When the data type of one knowledge data entity is process and the data type of another knowledge data entity is data, the linear mapping of the relationship is set to a dashed line; When the data type of one knowledge data entity is process and the data type of another knowledge data entity is message, the linear mapping of the relationship is set to dotted line; When one knowledge data entity has a data type of process and another knowledge data entity has a data type of AI model, the linear mapping of the relationship is set to double solid lines.

[0029] Feature form configuration: Configure a small form for each knowledge point node to record specific values ​​or field descriptions of usage frequency and scenario importance; In the acquired knowledge graph, the visual elements correspond to the core features of the knowledge points.

[0030] S4. Exclude isolated nodes in the knowledge graph of visualized elements in S3, randomly select a non-isolated node, and search for its connected nodes by traversing it, marking it as a connected subgraph; repeat the traversal process until all non-isolated nodes are marked, resulting in multiple connected subgraphs.

[0031] S5. Calculate the spatial attributes of the knowledge data entity graph in each connected subgraph. The spatial attributes of the knowledge data entity graph include the number of nodes associated with each node, the average length of the shortest path from a node to other nodes, and the number of times a node acts as a bridge for the shortest path from other nodes. Input the calculated number of nodes associated with each node, the average length of the shortest path from a node to other nodes, the number of times a node acts as a bridge for the shortest path from other nodes, the usage frequency obtained in step S2, and the scene importance into the LR model to calculate the knowledge data entity score. Compare the calculated knowledge data entity score with the set knowledge data entity threshold, mark important knowledge data entities, and divide them into hierarchical entities based on important knowledge data entities.

[0032] The method for quantifying the spatial attributes of a knowledge data entity graph is as follows: Number of edges associated with a node (denoted as ) The calculation logic involves traversing all associated edges within the connected subgraph and counting the edges related to the first edge. The number of edges directly connected to each node is calculated using the following formula: In the formula, A set of nodes for a single connected subgraph; For indicator functions: if node With nodes There are directly related edges. ;otherwise .

[0033] The average length of the shortest path from node to other nodes (denoted as ) The Dijkstra algorithm is used to calculate the shortest path length from node i to every other node j, and then the arithmetic mean is calculated. The formula is as follows: In the formula, For connected subgraphs The total number of nodes; Let be the shortest path length from node i to node j; if i and j have no connected path. .

[0034] The number of times a node acts as a bridge for the shortest path between other nodes (denoted as ) The calculation logic is to traverse all node pairs within the subgraph. Calculate the nodes included in its shortest path. The number of times; the calculation formula is: In the formula, For indicator functions: if node To the node The shortest path passes through the nodes , ;otherwise ; To avoid node pairing and Duplicate statistics.

[0035] In addition, because , , Frequency of use Importance of the scene Large dimensional differences necessitate standardization using standard scores to eliminate the influence of dimensions. The characteristic standardization formula is as follows (denoted by the standardized characteristic as...). ): In the formula, Indexing for feature dimensions ( correspond , correspond , correspond , correspond , correspond ); For the first The node of the first Dimensional original eigenvalues; For the first in the historical knowledge graph The mean of the dimensional features; For the first in the historical knowledge graph Standard deviation of dimensional features.

[0036] The linear prediction function of the LR model (denoted as ) The 5-dimensional standardized features are weighted and summed with the model parameters to obtain the linear prediction value. The calculation formula is as follows: In the formula, This represents the bias term in the LR model; For the first Weight parameters for dimensional features; For the first The node of the first Dimensional standardization features.

[0037] linear prediction value Mapping to the [0, 1] interval yields the knowledge data entity score: In the formula, For the first The knowledge data entity score of each node (value range [0, 1]); It is a natural constant; The above is the linear prediction value.

[0038] Specifically, step S4, obtaining the connected subgraph, includes the following steps: S41: Traverse the knowledge graph of visualized elements in S3, exclude isolated nodes with no relation to each other, and mark the remaining non-isolated nodes to form a set of non-isolated nodes; S42: Randomly select an initial node from the set of non-isolated nodes, and mark the initial node and its connected nodes as a connected subgraph by traversing and searching all nodes that are directly or indirectly connected to the initial node. S43: Repeat the operation of S42 until all nodes in the non-isolated node set are marked, and obtain multiple connected subgraphs.

[0039] Specifically, in step S5, the method for dividing the hierarchical entities is as follows: taking the important knowledge data entity as the core, a hierarchical entity set is divided according to the relational distance between the node and the important knowledge data entity. If there is no knowledge data entity with an entity score ≥ 0.5 in a certain connected subgraph, then the knowledge data entity with the highest entity score in the connected subgraph is set as the important knowledge data entity; taking the important knowledge data entity as the core level and setting it as level 0, traversing the 1-degree relation nodes of the important knowledge data entity and setting them as level 1, traversing the 1-degree relation nodes of the level 1 node and setting them as level 2, until there are no new relation nodes to traverse, and obtaining the hierarchical entities containing different levels.

[0040] To ensure smooth drill-down, when a user clicks on a node at a certain level to trigger drill-down, the system prioritizes loading the connected subgraph containing that node. During the drill-down process, breadcrumb navigation maintains knowledge path coordination, node visualization elements maintain feature coordination, and relational lines maintain association coordination.

[0041] This invention obtains knowledge data entities with a score greater than 0.5 through step S5, marks them as important knowledge data entities, and applies them to smooth drill-down in conjunction with hierarchical loading and dynamic display.

[0042] Upon initial loading, the current node and its first-degree related nodes are displayed, with a height of h between levels. Nodes within a level automatically wrap to a new line, with a height of h / 3. During subsequent loading, when the user clicks on a first-degree node, its second-degree related nodes and their corresponding relationships are loaded. If the number of nodes in the connected subgraph exceeds 5000, important knowledge data entities and first- and second-degree nodes are displayed first, while the loading of other nodes is controlled by buttons to avoid lag. Important knowledge data entity nodes are loaded first, and less important knowledge data entity nodes are loaded later.

[0043] The real-time update of the knowledge graph is triggered by a change in business logic. A change of ≥10% in business logic will trigger a real-time update. Business changes include changes in the number of nodes and fluctuations in the values ​​of core features. After triggering, the data of the changed knowledge point nodes or related edges will be updated in real time, and the knowledge relevance index will be recalculated to ensure the accuracy of the graph association after the update. The intermediate processing of the scheduled update is to perform a full update of the knowledge graph at a preset time period every day, including the import of new knowledge data entities and related edges, the deletion of invalid knowledge data entities and related edges, and the re-statistical and annotation of the core features of all knowledge data entities.

[0044] During the drill-down process, the rules for visualization elements should remain consistent. When the knowledge graph is dynamically updated, changes in core features require synchronous updates to the corresponding visualization elements. After the update, the knowledge relevance index should be re-verified to ensure that the corresponding relationships are compatible with the data.

[0045] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0046] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0047] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0048] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0049] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0050] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0051] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A knowledge management and representation method, characterized in that, Includes the following steps: S1. Obtain knowledge data entities, construct a pattern layer of a knowledge graph based on the knowledge data entities, and calculate the matching degree between the knowledge data entity node text and the pattern layer node text. When the matching degree between the knowledge data entity node text and the pattern layer node text is greater than a set matching threshold, it is determined that the pattern layer is compatible with the knowledge data entity. S2. Perform position mapping on the pattern layer to generate a graph structure framework, calculate the knowledge relevance index of the pattern layer, and determine that the graph structure framework is qualified for association when the knowledge relevance index is greater than the relevance threshold; at the same time, obtain the core features of the pattern layer, including usage frequency and scene importance. S3. Generate a knowledge graph containing knowledge point nodes, relational edges, and visual elements based on the graph structure framework, the knowledge relevance index of the pattern layer, and core features; Pre-set multi-dimensional visualization rules and determine the correspondence between the multi-dimensional visualization rules and the knowledge graph of visualization elements; S4. Exclude isolated nodes in the knowledge graph of visualized elements in S3, randomly select a non-isolated node, search for its connected nodes by traversal, and mark it as a connected subgraph; repeat the traversal process until all non-isolated nodes are marked, and obtain multiple connected subgraphs. S5. Calculate the spatial attributes of the knowledge data entity graph in each connected subgraph. The spatial attributes of the knowledge data entity graph include the number of nodes associated with each node, the average length of the shortest path from a node to other nodes, and the number of times a node acts as a bridge for the shortest path from other nodes. Input the calculated number of nodes associated with each node, the average length of the shortest path from a node to other nodes, the number of times a node acts as a bridge for the shortest path from other nodes, the usage frequency obtained in step S2, and the scene importance into the LR model to calculate the knowledge data entity score. Compare the calculated knowledge data entity score with the set knowledge data entity threshold, mark important knowledge data entities, and divide them into hierarchical entities based on important knowledge data entities.

2. The knowledge management and representation method according to claim 1, characterized in that: In step S1, obtaining knowledge data entities involves extracting knowledge data entities from the knowledge data using an entity recognition algorithm and extracting the relationships between the knowledge data entities using a relation extraction algorithm; the entity recognition algorithm is the BERT-CRF algorithm, and the relation extraction algorithm is the PCNN algorithm.

3. The knowledge management and representation method according to claim 1, characterized in that, In step S1, the formula for calculating the node text matching degree is: In the formula, The number of the preset time period. For the first The text matching degree of nodes within a preset time period. It is a natural constant. For the first The string matching degree for a preset time period. For string reference matching degree, For the first Semantic similarity over a preset time period For semantic reference similarity, For the first Attribute validation values ​​for a preset time period. This is a reference validation value for the attribute. This is a correction factor for attribute validation.

4. The knowledge management and representation method according to claim 1, characterized in that, In step S2, the formula for calculating the knowledge relevance index is: In the formula, The number of the preset time period. For the first A knowledge relevance index for a preset time period. It is a natural constant. For the first Entity matching degree over a preset time period For entity reference matching degree, For the first Consistency of relationships across preset time periods For the sake of relational reference consistency, For the first The concept of coverage for a preset time period.

5. The knowledge management and representation method according to claim 1, characterized in that, In step S2, the multi-dimensional visualization rules are as follows: Node size mapping: Five fixed node sizes are preset. The maximum usage frequency of all knowledge data entities is obtained. Frequency intervals are divided by the maximum usage frequency divided by 5. Knowledge point nodes whose usage frequency falls into different intervals are matched with the corresponding sizes. If the maximum usage frequency is <10, all knowledge point nodes adopt the size corresponding to the maximum interval. Node color mapping: Five gradient colors are preset. The maximum scene importance index of all knowledge data entities is obtained. The importance interval is divided by the maximum index by 5. Knowledge point nodes that fall into different scene importance intervals are matched with the corresponding colors. If the maximum index is less than 10, then all knowledge point nodes will use the color corresponding to the maximum range. Node line number mapping: Three types of node description lines are preset: thin line, medium thick line and thick line, which correspond to the low, medium and important levels of the scene importance of knowledge data entities, respectively. After determining the level, the corresponding line is matched. Relationship line type mapping: Four preset relationship line types are used, and the corresponding line type is matched after the type is determined according to the degree of business process dependency. Feature form configuration: Configure a small form for each knowledge point node to record specific values ​​or field descriptions of usage frequency and scenario importance; In the acquired knowledge graph, the visualization elements correspond to the core features of the knowledge points.

6. The knowledge management and representation method according to claim 1, characterized in that, Step S4, obtaining the connected subgraph specifically includes the following steps: S41: Traverse the knowledge graph of visualized elements in S3, exclude isolated nodes with no relation to each other, and mark the remaining non-isolated nodes to form a set of non-isolated nodes; S42: Randomly select an initial node from the set of non-isolated nodes, and mark the initial node and its connected nodes as a connected subgraph by traversing and searching all nodes that are directly or indirectly connected to the initial node; S43: Repeat the operation of S42 until all nodes in the set of non-isolated nodes are marked, and obtain multiple connected subgraphs.

7. The knowledge management and representation method according to claim 1, characterized in that, In step S5, the method for dividing the hierarchical entities is as follows: taking important knowledge data entities as the core, the hierarchical entity set is divided according to the relationship distance between the node and the important knowledge data entity. If there is no knowledge data entity with an entity score ≥ 0.5 in a certain connected subgraph, then the knowledge data entity with the highest entity score in the connected subgraph is set as the important knowledge data entity. Taking the important knowledge data entity as the core level and setting it as level 0, traverse the first-degree relation nodes of the important knowledge data entity and set them as level 1, traverse the first-degree relation nodes of the level 1 node and set them as level 2, until there are no new relation nodes to traverse, and obtain the hierarchical entities containing different levels.

8. The knowledge management and representation method according to claim 7, characterized in that: After obtaining the hierarchical entities containing different levels, breadcrumb navigation is used to maintain knowledge path coordination during the drill-down process.

Images