Construction method and system of intellectual property mapping knowledge domain

Abstract

The invention provides a construction method and system of an intellectual property knowledge graph, and relates to the technical field of data processing, knowledge cells of an intellectual property document are constructed, the knowledge cells are analyzed by using a natural language processing technology, entities in the intellectual property field, attributes of the entities and relationships among the entities are identified and abstracted, and the intellectual property knowledge graph is constructed. Obtaining entity information; establishing a first intellectual property knowledge graph based on the knowledge cells and the entity information; optimizing the structure of the first intellectual property knowledge graph by using a graph neural network technology to obtain a second intellectual property knowledge graph; and a user behavior analysis module is integrated to the second intellectual property knowledge graph, so that a final intellectual property knowledge graph is obtained by tracking a user query mode and preference information, identifying user demand hotspots and adjusting the content of the second intellectual property knowledge graph. According to the invention, the intellectual property information management intelligence level is improved, and the accuracy and practicability of the intellectual property mapping knowledge domain are enhanced.

Description

Technical Field

[0001] Embodiments of the present invention relate to the field of data processing technology, and in particular, to a method and system for constructing an intellectual property knowledge graph. Background Art

[0002] With the rapid development of the field of intellectual property, the number of various intellectual property documents, such as patent documents, trademark records, copyright works, etc., has increased dramatically. These documents contain rich information, but for enterprises and researchers, how to effectively manage and use this information has become an important challenge.

[0003] The existing intellectual property management system relies on manual data entry and uses text processing tools with limited functions, which has many problems. The intellectual property documents of the existing solution are stored in different databases, which makes it difficult to manage and analyze them in a unified manner; the traditional information extraction technology is used, which makes information extraction difficult; the user interface is static and simple, resulting in inaccurate matching of user needs. Summary of the invention

[0004] The embodiments of the present invention provide a method and system for constructing an intellectual property knowledge graph, so as to solve the problems in the prior art of difficulty in unified management and comprehensive analysis, difficulty in information extraction, and inaccurate matching of user needs.

[0005] In a first aspect, an embodiment of the present invention provides a method for constructing an intellectual property knowledge graph, comprising: Receiving real-time data streams from different sources, the real-time data streams containing structured data and unstructured data; Construct knowledge cells of intellectual property documents, and parse the knowledge cells using natural language processing technology to identify and abstract entities, attributes of entities, and relationships between entities in the field of intellectual property to obtain entity information; Based on the knowledge cells and the entity information, establish a first intellectual property knowledge graph; Optimizing the structure of the first intellectual property knowledge graph using graph neural network technology to obtain a second intellectual property knowledge graph, wherein the graph neural network technology includes a graph neural network model and a graph convolution operation; Integrate the user behavior analysis module into the second intellectual property knowledge graph to identify user demand hotspots by tracking user query patterns and preference information, adjust the content of the second intellectual property knowledge graph, and obtain the final intellectual property knowledge graph.

[0006] Optionally, the structure of the first intellectual property knowledge graph is optimized by using a graph neural network technology to obtain a second intellectual property knowledge graph, and the graph neural network technology includes a graph neural network model and a graph convolution operation, including: Obtaining node information and edge information in the first intellectual property knowledge graph, and performing feature encoding on the node information and edge information respectively to obtain a node feature vector and an edge feature vector; Based on the node feature vector and the edge feature vector, iteratively updating the node representation through the graph neural network model, and extracting the multi-level association pattern between the nodes; Based on the association pattern, the combined features of the node and its neighboring nodes are extracted using graph convolution operations, and the mutual influence between the node attributes and the graph structure is analyzed through the combined features to obtain an enhanced node representation; Based on the enhanced node representation, the edge information in the first intellectual property knowledge graph is adjusted through an optimization algorithm to determine the optimal structural representation of the node, and a second intellectual property knowledge graph is generated based on the optimal structural representation.

[0007] Optionally, the iterative updating of node representations through a graph neural network model based on the node feature vector and the edge feature vector to extract multi-level association patterns between nodes includes: Initializing the node feature vector and the edge feature vector, and using the initialized node feature vector and edge feature vector as input of a graph neural network model; In each iteration step, the message passing mechanism of the graph neural network model is used to pass messages, so that the node receives information from the neighboring nodes, and fuses the received information with the node's own characteristics to obtain node fusion information; the node fusion information is processed using the graph convolution layer to update the node representation; it is determined whether the change amplitude of the updated node representation is less than the preset threshold; If not, the steps of message transmission, information fusion and threshold judgment are repeated until the predetermined convergence standard is reached and the iterative update is completed; the predetermined convergence standard is that the change amplitude of the updated node representation is less than the preset threshold, or the preset number of iterations is reached; wherein, the updated node representation in each iterative step is used to reflect the multi-level association pattern between nodes.

[0008] Optionally, based on the association pattern, the combined features of the node and its neighboring nodes are extracted by graph convolution operation, and the mutual influence between the node attributes and the graph structure is analyzed by the combined features to obtain the enhanced node representation, including: Identify the position of the node and the position of the neighboring nodes in the graph structure, and extract the feature information of the node and the neighboring nodes through graph convolution operation; the graph structure is a structure composed of updated node representations; Combining the feature information to obtain local graph features of the node, wherein the local graph features include neighbor node information and node attributes; Based on the local graph features, analyzing node attributes and the positions of nodes in the graph structure to determine the influence of the node attributes on the connection mode of the nodes and the roles of the nodes, and obtaining enhanced node representations, wherein the positions include the degree and centrality of the nodes; Based on the enhanced node representation, the identification step, the extraction step, the combination step and the analysis step are repeatedly performed until a predetermined convergence condition is reached, wherein the convergence condition is that the change amplitude of the enhanced node representation is less than a preset threshold value, or a preset number of iterations is reached; through multiple iterative processes, the enhanced node representation is optimized so that each enhanced node representation contains its own attribute information and the influence of neighboring nodes, thereby obtaining an enhanced node representation.

[0009] Optionally, based on the local graph features, analyzing node attributes and the position of the node in the graph structure to determine the influence of the node attributes on the connection mode of the node and the role of the node, and obtaining an enhanced node representation, the position includes the degree and centrality of the node, including: Based on the local graph features, the nodes are preliminarily encoded using graph embedding technology, and the preliminarily encoded node attributes are converted into numerical feature vectors to obtain a preliminarily encoded result; The preliminary encoding result is inputted using the graph neural network architecture, the node attributes are dynamically analyzed in combination with the attention mechanism, and different weights are set corresponding to different node attributes to obtain the node representation to be updated; Based on the node representation to be updated, a structured deep learning framework is introduced to evaluate the position characteristics of the nodes in the graph structure, and the dependencies between the nodes are captured through a high-order neighborhood aggregation algorithm to quantify the relative importance of the nodes in the graph structure, thereby obtaining a relative importance quantification result; Based on the relative importance quantification results, a community detection algorithm is used to determine the influence of the node attributes on the connection mode of the nodes and the role of the nodes, and the information between the nodes is transmitted through the graph convolution layer and the node representation to be updated is updated to obtain the updated node representation; Based on the updated node representation, a feature enhancement strategy is implemented, the intrinsic attributes and external structure of the node are analyzed through ensemble learning, and the quality of the node representation is improved through a self-supervised learning mechanism to obtain an enhanced node representation.

[0010] Optionally, the user behavior analysis module is integrated into the second intellectual property knowledge graph to track user query patterns and preference information, identify user demand hotspots, and adjust the content of the second intellectual property knowledge graph to obtain a final intellectual property knowledge graph, including: Integrate a user behavior analysis module into the second intellectual property knowledge graph to collect user query patterns and preference information by recording user interaction behaviors to obtain query data, wherein the user interaction behaviors include query history and click habits; Analyze the query data, identify key areas and hot topics of user concern, and obtain user demand hotspots; Based on the user demand hotspots, the content of the second intellectual property knowledge graph is adjusted to generate a final intellectual property knowledge graph.

[0011] Optionally, adjusting the content of the second intellectual property knowledge graph based on the user demand hotspot to generate a final intellectual property knowledge graph includes: Analyze the user demand hot spots by using natural language processing technology, extract key words, and quantify the importance of the key words by using a semantic similarity algorithm to obtain quantitative results reflecting the hot areas; According to the quantification result, the weights of entities and relationships between entities in the second intellectual property knowledge graph are adjusted to obtain a third intellectual property knowledge graph that highlights hot areas; Based on the third intellectual property knowledge graph, an active learning algorithm is used to dynamically adjust the graph content and optimize the display of nodes and edges according to user feedback information to obtain an optimized fourth intellectual property knowledge graph; Based on the fourth intellectual property knowledge graph, a personalized recommendation mechanism is introduced to generate the final intellectual property knowledge graph.

[0012] In a second aspect, an embodiment of the present invention provides a system for constructing an intellectual property knowledge graph, including: A parsing module is used to construct knowledge cells of intellectual property documents and parse the knowledge cells using natural language processing technology to identify and abstract entities, attributes of entities, and relationships between entities in the field of intellectual property to obtain entity information; An establishment module, used to establish a first intellectual property knowledge graph based on the knowledge cells and the entity information; A determination module, configured to optimize the structure of the first intellectual property knowledge graph using a graph neural network technology to obtain a second intellectual property knowledge graph, wherein the graph neural network technology includes a graph neural network model and a graph convolution operation; A generation module is used to integrate the user behavior analysis module into the second intellectual property knowledge graph, so as to identify user demand hotspots by tracking user query patterns and preference information, adjust the content of the second intellectual property knowledge graph, and obtain the final intellectual property knowledge graph.

[0013] In a third aspect, an embodiment of the present invention provides a computing device, comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the method for constructing an intellectual property knowledge graph as described in any one of the first aspects.

[0014] In a fourth aspect, an embodiment of the present invention provides a computer storage medium having computer program instructions stored thereon, wherein the computer program instructions, when executed by a processor, implement the method for constructing an intellectual property knowledge graph as described in any one of the first aspects.

[0015] In an embodiment of the present invention, knowledge cells of an intellectual property document are constructed, and natural language processing technology is used to parse the knowledge cells to identify and abstract entities, attributes of entities, and relationships between entities in the field of intellectual property to obtain entity information; a first intellectual property knowledge graph is established based on the knowledge cells and the entity information; the structure of the first intellectual property knowledge graph is optimized using graph neural network technology to obtain a second intellectual property knowledge graph, and the graph neural network technology includes a graph neural network model and a graph convolution operation; a user behavior analysis module is integrated into the second intellectual property knowledge graph to identify user demand hotspots by tracking user query patterns and preference information, and the content of the second intellectual property knowledge graph is adjusted to obtain a final intellectual property knowledge graph. The technical solution provided by the present invention improves the intelligence level of intellectual property information management, and through multi-step processing and optimization, it is closer to the actual needs of users, and enhances the accuracy and practicality of the intellectual property knowledge graph; the node attributes are converted into numerical feature vectors through graph embedding technology, thereby improving the richness and accuracy of node representation; the node attributes are dynamically analyzed in combination with the attention mechanism, and different weights corresponding to different node attributes are set, thereby enhancing the comprehensiveness and expression ability of the node representation; the position characteristics of the nodes in the graph structure are evaluated through a structured deep learning framework, and the dependencies between nodes are captured through a high-order neighborhood aggregation algorithm, and the relative importance of the nodes in the graph structure is quantified; the community detection algorithm is used to determine the influence of node attributes on the connection mode of the nodes and the role of the nodes, and information is transmitted and the node representation is updated through the graph convolution layer, thereby further optimizing the node representation; the quality of the node representation is improved through integrated learning and self-supervised learning mechanisms, and finally an enhanced node representation is obtained, thereby improving the overall quality and practicality of the knowledge graph.

[0016] These and other aspects of the present invention will become more apparent from the following description of the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1A flowchart of a method for constructing an intellectual property knowledge graph provided by an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a system for constructing an intellectual property knowledge graph provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0020] In some of the processes described in the specification and claims of the present invention and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0022] Figure 1 A flowchart of a method for constructing an intellectual property knowledge graph is provided for an embodiment of the present invention, such as Figure 1 As shown, the method includes: In view of the shortcomings of the existing technology in processing complex and large amounts of intellectual property information, such as information islands and weak knowledge relevance, a systematic method is urgently needed to integrate and optimize intellectual property resources. Based on this, the present invention provides a method for constructing an intellectual property knowledge graph, such as Figure 1 ,include: Step 101: construct knowledge cells of the intellectual property document, and parse the knowledge cells using natural language processing technology to identify and abstract entities, attributes of entities, and relationships between entities in the field of intellectual property to obtain entity information; In this step, knowledge cells refer to the smallest information units in intellectual property documents, including patent documents, trademark records, copyrighted works, etc.; entities refer to specific objects in intellectual property documents, such as inventors, patent numbers, company names, etc.; entity attributes refer to information about entity characteristics, such as the inventor's position, the application date of the patent, etc.; relationships between entities refer to connections between entities, such as the relationship between an inventor and his or her company, patent citation relationships, etc.

[0023] This step converts the intellectual property document into a computer-readable format, and performs cleaning and standardization processing; uses NLP technology to identify key entities in the document, such as inventors, patent numbers, etc.; extracts the relationships between entities from the document, such as citation relationships, cooperative relationships, etc.; extracts attribute information describing the entity, such as the inventor's position, the patent application date, etc.; and integrates the above extracted information into structured entity information.

[0024] Step 102: Establishing a first intellectual property knowledge graph based on the knowledge cells and the entity information; In this step, graph nodes are created based on the entity information obtained in step 101, and the relationships between entities are represented as edges; the extracted entity information and relationship information are integrated into the graph to form the basic framework of the first knowledge graph; corresponding attributes are assigned to the nodes and edges in the graph, such as the inventor's position, the application date of the patent, etc.

[0025] Step 103: Optimize the structure of the first intellectual property knowledge graph using graph neural network technology to obtain a second intellectual property knowledge graph, wherein the graph neural network technology includes a graph neural network model and a graph convolution operation; In this step, graph neural network (GNN) refers to a deep learning model specially designed to process graph structured data, which can capture the complex relationships between nodes; graph convolution operation refers to an operation performed on the graph structure, which is used to update the node representation so that it can reflect the information of the node and its neighbors.

[0026] This step first initializes the representation vector for each node in the graph; it iteratively updates the node representation through the graph convolution layer to extract the multi-level association pattern between nodes; uses the graph neural network model to update the node representation so that the node representation not only reflects the properties of the node itself, but also integrates the influence of neighboring nodes; optimizes the graph structure based on the updated node representation to obtain a second intellectual property knowledge graph with a more reasonable structure.

[0027] Step 104: Integrate a user behavior analysis module into the second intellectual property knowledge graph to track user query patterns and preference information, identify user demand hotspots, and adjust the content of the second intellectual property knowledge graph to obtain a final intellectual property knowledge graph; This step integrates the user behavior analysis module into the second intellectual property knowledge graph; records the user's query behavior, click habits and other information; identifies the hot areas of user concern by analyzing the user's query patterns and preference information; adjusts the content of the knowledge graph according to user demand hotspots to ensure that the graph can better meet user needs; after completing the content adjustment, generates the final intellectual property knowledge graph, which can dynamically respond to changes in user needs.

[0028] The embodiments of the present invention can achieve the following beneficial effects through the above steps: By constructing knowledge cells and parsing the cell contents using natural language processing technology, the systematic integration of intellectual property information is achieved; The graph neural network technology is used to optimize the knowledge graph structure, which improves the expressiveness and accuracy of the graph; The integrated user behavior analysis module dynamically adjusts the graph content according to user needs, enhancing the practicality and user experience of the graph.

[0029] In view of the structural limitations of the first intellectual property knowledge graph, in order to improve its expression ability and accuracy, based on this, the present invention provides a specific embodiment, in which the step 103 uses the graph neural network technology to optimize the structure of the first intellectual property knowledge graph to obtain the second intellectual property knowledge graph, and the graph neural network technology includes a graph neural network model and a graph convolution operation, and specifically includes the following steps: Step 310: Obtain node information and edge information in the first intellectual property knowledge graph, and perform feature encoding on the node information and edge information respectively to obtain a node feature vector and an edge feature vector; In this step, node information refers to the information representing entities in the knowledge graph, such as inventors, patent numbers, etc.; edge information refers to the information representing the relationships between entities in the knowledge graph, such as citation relationships, cooperation relationships, etc.; feature encoding refers to the process of converting non-numeric information into numerical vectors for computer processing.

[0030] This step encodes the attribute information of each node to generate a node feature vector. For example, an inventor may have attributes such as name, position, and company, which are converted into a numerical vector through encoding. The type and strength of each edge are encoded to generate an edge feature vector. For example, a citation relationship can have different weights, which are converted into a numerical vector through encoding.

[0031] Step 311: Based on the node feature vector and the edge feature vector, iteratively update the node representation through the graph neural network model to extract the multi-level association pattern between nodes; This step initializes the representation vector of each node based on the node feature vector obtained in the first step; in each iteration, the node receives information from neighboring nodes through the message passing mechanism of the graph neural network, and fuses the received information with its own feature information; using the graph neural network model, the node representation vector is updated according to the fused information to extract the multi-level association pattern between nodes; the above process is repeated until the predetermined convergence criterion is reached or the specified number of iterations is completed.

[0032] Step 312: Based on the association pattern, a graph convolution operation is used to extract the combined features of the node and its neighboring nodes, and the mutual influence between the node attributes and the graph structure is analyzed through the combined features to obtain an enhanced node representation; This step aggregates the feature information of the node and its neighboring nodes through the graph convolution layer to extract the combined features; fuses the node's own feature information with the aggregated neighboring node feature information to obtain an enhanced node representation; analyzes the mutual influence between node attributes, such as the inventor's position, patent application date, etc. and the graph structure, further optimizes the node representation, and obtains an enhanced node representation.

[0033] Step 313: Based on the enhanced node representation, the edge information in the first intellectual property knowledge graph is adjusted by an optimization algorithm to determine the optimal structural representation of the node, and a second intellectual property knowledge graph is generated according to the optimal structural representation; In this step, according to the enhanced node representation, the edge information in the first intellectual property knowledge graph is adjusted through an optimization algorithm to optimize the connection relationship between nodes; the optimal structural representation of the nodes is determined through the optimization algorithm so that the graph can better reflect the real relationship between entities; based on the optimal structural representation, a second intellectual property knowledge graph with a more optimized structure is generated, which can more accurately reflect the entities and their relationships in the field of intellectual property.

[0034] The embodiments of the present invention can achieve the following beneficial effects through the above steps: Through feature coding and graph neural network technology, the feature information of nodes and edges is effectively extracted, improving the expressiveness of the graph; The combined features of nodes and neighboring nodes are extracted using graph convolution operations, which enhances the node representation so that the node representation not only reflects its own attribute information but also integrates the influence of neighboring nodes. By adjusting the graph structure through optimization algorithms, the final generated second intellectual property knowledge graph can more accurately reflect the actual relationship between entities, thereby improving the practicality and accuracy of the graph.

[0035] In order to capture the complex relationship between nodes and lay a foundation for the subsequent generation of a more optimized second intellectual property knowledge graph, the present invention provides a specific embodiment. The step 311, based on the node feature vector and the edge feature vector, iteratively updates the node representation through the graph neural network model to extract the multi-level association pattern between nodes, specifically includes the following steps: Step 321: Initialize the node feature vector and the edge feature vector, and use the initialized node feature vector and edge feature vector as input of the graph neural network model; In this step, the graph neural network model refers to a deep learning model specifically used to process graph structured data, which can capture the complex relationships between nodes.

[0036] This step initializes a feature vector for each node based on the original information of the node, such as the inventor's position, patent application date, etc.; initializes a feature vector for each edge based on the type and strength of the edge, such as the strength of the citation relationship, the closeness of the cooperative relationship, etc.; and sends the initialized node feature vector and edge feature vector as input to the graph neural network model for processing.

[0037] Step 322: In each iteration step, message passing is performed through the message passing mechanism of the graph neural network model, so that the node receives information from the neighboring node, and fuses the received information with the node's own characteristics to obtain node fusion information; the node fusion information is processed using the graph convolution layer to update the node representation; and it is determined whether the change amplitude of the updated node representation is less than a preset threshold; In this step, the message passing mechanism refers to a mechanism in the graph neural network, through which a node can receive information from its neighboring nodes; information fusion refers to combining the information received by a node from its neighboring nodes with its own feature information to obtain node fusion information; the graph convolution layer refers to a layer used to process graph structured data, through which the node representation can be updated; the node representation refers to a vector representing the node features, reflecting the attribute information of the node and its relationship with its neighboring nodes.

[0038] In each iterative step, each node receives information from its neighboring nodes through the message passing mechanism of the graph neural network; the received information is fused with the node's own feature information to obtain node fusion information; the node fusion information is processed using the graph convolution layer to update the node representation; and it is determined whether the change in the updated node representation is less than a preset threshold.

[0039] Step 323: If not, the message transmission, information fusion and threshold judgment steps are repeatedly executed until a predetermined convergence standard is reached and the iterative update is completed; the predetermined convergence standard is that the change amplitude of the updated node representation is less than a preset threshold, or a preset number of iterations is reached; wherein the updated node representation in each iteration step is used to reflect the multi-level association mode between nodes; In this step, if the change in the updated node representation is greater than the preset threshold, the message passing, information fusion and threshold judgment steps are repeated; the iterative process will continue until the change in the updated node representation is less than the preset threshold, or the preset number of iterations is reached; the updated node representation in each iterative step is used to reflect the multi-level association pattern between nodes, making the node representation more accurate and rich.

[0040] The embodiments of the present invention can achieve the following beneficial effects through the above steps: The initialization of node feature vectors and edge feature vectors provides a basis for subsequent feature extraction; By using the message passing mechanism and graph convolutional layer of the graph neural network, the multi-level association patterns between nodes are effectively extracted, which enhances the accuracy of node representation. By iteratively updating the node representation, it is ensured that the change amplitude of the node representation converges to the preset threshold, thereby improving the structural rationality of the knowledge graph.

[0041] In order to enhance the quality of node representation so that the node representation can better reflect the multi-level association pattern between nodes, the present invention provides a specific embodiment, in which step 312, based on the association pattern, uses graph convolution operation to extract the combined features of the node and neighboring nodes, and analyzes the mutual influence between the node attributes and the graph structure through the combined features to obtain the enhanced node representation, specifically including the following steps: Step 331: Identify the position of the node and the position of the neighboring nodes in the graph structure, and extract the feature information of the node and the neighboring nodes through graph convolution operation; the graph structure is a structure composed of the updated node representation; In this step, graph convolution operation refers to an operation performed on graph structured data, which is used to extract feature information of nodes and their neighboring nodes; node position refers to the position of the node in the graph, including indicators such as the node's degree (DegreeCentrality), closeness centrality (Closeness Centrality) and betweenness centrality (BetweennessCentrality).

[0042] This step determines the position of each node in the graph structure, including the node's degree, closeness centrality, betweenness centrality, etc.; determines the neighbor nodes of each node and their position in the graph; and extracts the feature information of the node and its neighbor nodes through graph convolution operations, including the node's own attribute information and the information of its neighbor nodes.

[0043] Step 332: combining the feature information to obtain local graph features of the node, wherein the local graph features include neighbor node information and node attributes; This step combines the feature information of the node itself with the feature information of its neighboring nodes to obtain the local graph features of the node. The local graph features include the attribute information of the node itself, such as the inventor's position, the application date of the patent, etc., as well as the information of its neighboring nodes, such as the strength of the citation relationship, the closeness of the cooperation relationship, etc.

[0044] Step 333: Based on the local graph features, analyze the node attributes and the position of the node in the graph structure to determine the influence of the node attributes on the connection mode of the node and the role of the node, and obtain an enhanced node representation, wherein the position includes the degree and centrality of the node; This step is based on the local graph features of the nodes, analyzing the mutual influence between node attributes, such as the inventor's position, patent application date, etc., and the node's position in the graph structure, such as degree, centrality, etc.; determining the influence of node attributes on node connection patterns, such as citation relationships, cooperative relationships, etc., and the role of nodes, such as importance in the graph; based on the above analysis results, the enhanced node representation is obtained Step 334: Based on the enhanced node representation, repeatedly perform the identification step, the extraction step, the combination step and the analysis step until a predetermined convergence condition is reached, wherein the convergence condition is that the change amplitude of the enhanced node representation is less than a preset threshold value, or a preset number of iterations is reached; through multiple iterations, the enhanced node representation is optimized so that each enhanced node representation contains its own attribute information and the influence of neighboring nodes, thereby obtaining an enhanced node representation; This step is based on the enhanced node representation, and repeatedly executes the steps of identifying node positions, extracting feature information through graph convolution operations, combining feature information, and analyzing node attributes and positions; the iterative process continues until the change in the enhanced node representation is less than the preset threshold, or the preset number of iterations is reached; through multiple iterative processes, the enhanced node representation is continuously optimized so that each enhanced node representation not only contains its own attribute information, but also the influence of neighboring nodes, and finally an enhanced node representation is obtained.

[0045] The embodiments of the present invention can achieve the following beneficial effects through the above steps: The feature information of nodes and neighboring nodes is extracted through graph convolution operations, which improves the richness and accuracy of node representation; By combining node feature information and analyzing the mutual influence between node attributes and graph structure, the comprehensiveness and expressiveness of node representation are enhanced; By optimizing and enhancing node representation through multiple iterations, it is ensured that each node representation can reflect the node's own attribute information and the influence of its neighboring nodes, thus improving the overall quality and practicality of the knowledge graph.

[0046] In order to further enhance the node representation, make it richer and more distinguishable, and provide a solid foundation for subsequent knowledge discovery, predictive modeling and other work. Based on this, the present invention provides a specific embodiment, the step 333, based on the local graph features, analyzes the node attributes and the position of the node in the graph structure to determine the influence of the node attributes on the connection mode of the node and the role of the node, and obtains an enhanced node representation, the position includes the degree and centrality of the node, specifically including the following steps: Step 341: Based on the local graph features, the nodes are preliminarily encoded using graph embedding technology, and the preliminarily encoded node attributes are converted into numerical feature vectors to obtain a preliminarily encoded result; In this step, graph embedding technology refers to a technology that converts node attributes into numerical vectors, which is used to represent the characteristics of nodes in the graph; preliminary encoding refers to the process of converting node attributes into numerical feature vectors, which provides a basis for subsequent processing.

[0047] This step extracts information about nodes and their neighboring nodes based on local graph features; uses graph embedding technology to preliminarily encode nodes and convert node attributes into numerical feature vectors; generates a preliminary encoding vector for each node as input for subsequent steps.

[0048] Step 342: Use the graph neural network architecture to input the preliminary encoding result, combine the attention mechanism to dynamically analyze the node attributes, and set different weights corresponding to different node attributes to obtain the node representation to be updated; In this step, the graph neural network architecture refers to a deep learning model used to process graph structured data; the attention mechanism refers to a technology used to dynamically assign weights in deep learning to emphasize important information; the node representation to be updated refers to the preliminary encoding result obtained through the graph neural network and the attention mechanism, which prepares for the next step of node representation update.

[0049] This step inputs the preliminary encoding results into the graph neural network architecture; dynamically analyzes node attributes in combination with the attention mechanism, and sets different weights for different node attributes; obtains the node representation to be updated through the graph neural network and attention mechanism.

[0050] Step 343: Based on the node representation to be updated, a structured deep learning framework is introduced to evaluate the position characteristics of the nodes in the graph structure, and the dependencies between the nodes are captured by a high-order neighborhood aggregation algorithm to quantify the relative importance of the nodes in the graph structure, thereby obtaining a relative importance quantification result; In this step, the structured deep learning framework refers to a deep learning framework used to process structured data; the high-order neighborhood aggregation algorithm refers to an algorithm used to capture high-order relationships between nodes.

[0051] This step uses a structured deep learning framework to evaluate the positional characteristics of nodes in the graph structure; captures the dependencies between nodes through a high-order neighborhood aggregation algorithm; evaluates the relative importance of nodes in the graph structure and obtains quantitative results.

[0052] Step 344: Based on the relative importance quantification result, a community detection algorithm is used to determine the influence of the node attribute on the connection mode of the node and the role of the node, and the information between the nodes is transmitted through the graph convolution layer and the node representation to be updated is updated to obtain the updated node representation; In this step, the community detection algorithm refers to a technique for identifying groups of nodes in a graph; the graph convolution layer refers to a layer for performing convolution operations on graph structured data; and the updated node representation refers to the node representation that is updated after information is passed through the graph convolution layer.

[0053] More specifically, the present invention also provides a node representation formula, through which the node representation of each node is iteratively updated in the graph convolution layer through the information of neighboring nodes, thereby capturing the multi-level association pattern between nodes.

[0054] Among them, the calculation formula of the node representation formula is as follows: ; Is a node In the Node representation of the layer; Is a node In the Node representation of the layer; is the activation function used for nonlinear transformation; It is The weight matrix of the layer, which is used to transform the node representation of the node; It is The bias term of the layer is used to adjust the node representation of the node; Is a node The set of neighbor nodes of Is a node The number of neighbor nodes; Neighbor node The weight is used to weight the information of neighbor nodes; Neighbor node In the Node representation of a layer.

[0055] Step 345: Based on the updated node representation, a feature enhancement strategy is implemented to analyze the intrinsic attributes and external structure of the node through ensemble learning, and the quality of the node representation is improved through a self-supervised learning mechanism to obtain an enhanced node representation; In this step, feature enhancement strategy refers to a method used to improve the quality of node representation; ensemble learning refers to a technology that improves prediction performance by integrating multiple models; self-supervised learning mechanism refers to a technology that can learn without labels and improves model performance by constructing self-supervised tasks.

[0056] The embodiments of the present invention can achieve the following beneficial effects through the above steps: The node attributes are converted into numerical feature vectors through graph embedding technology, which improves the richness and accuracy of node representation; The attention mechanism is used to dynamically analyze node attributes and set different weights for different node attributes, which enhances the comprehensiveness and expressiveness of node representation. The structured deep learning framework is used to evaluate the position characteristics of nodes in the graph structure, and the high-order neighborhood aggregation algorithm is used to capture the dependencies between nodes and quantify the relative importance of nodes in the graph structure. The community detection algorithm is used to determine the impact of node attributes on the connection pattern of nodes and the role of nodes, and the information is passed through the graph convolution layer to update the node representation and further optimize the node representation; The quality of node representation is improved through integrated learning and self-supervised learning mechanisms, and ultimately an enhanced node representation is obtained, which improves the overall quality and practicality of the knowledge graph.

[0057] In order to further improve the practicality and user experience of the second intellectual property knowledge graph, based on this, the present invention provides a specific embodiment, wherein the step 104 integrates a user behavior analysis module into the second intellectual property knowledge graph to track user query patterns and preference information, identify user demand hotspots, and adjust the content of the second intellectual property knowledge graph to obtain a final intellectual property knowledge graph, specifically comprising the following steps: Step 401: Integrate a user behavior analysis module into the second intellectual property knowledge graph to collect user query patterns and preference information by recording user interaction behaviors to obtain query data, wherein the user interaction behaviors include query history and click habits; In this step, the user behavior analysis module refers to a software module used to collect and analyze user behavior data, which can track users' query patterns and preference information; user interaction behavior refers to the user's behavior records when using the knowledge graph, including query history and click habits.

[0058] This step integrates the user behavior analysis module into the second intellectual property knowledge graph, so that it can monitor user behavior in real time; record the user's query history and click habits, such as search keywords, browsed nodes, etc., and click habits such as which nodes to spend more time on, click frequency, etc.; collect the user's query pattern and preference information through the user behavior analysis module to obtain detailed query data.

[0059] Step 402: Analyze the query data, identify key areas and hot topics that users are concerned about, and obtain user demand hotspots; This step pre-processes the collected query data to remove noise and redundant information; uses natural language processing techniques, such as text mining and keyword extraction, to identify user query patterns; uses data analysis techniques, such as cluster analysis and association rule mining, to identify key areas and hot topics of user concern; combines the above analysis results to determine the key areas and hot topics of user concern and obtain user demand hotspots.

[0060] Step 403: Based on the user demand hot spots, adjust the content of the second intellectual property knowledge graph to generate a final intellectual property knowledge graph; This step adjusts the content of the second intellectual property knowledge graph according to user demand hotspots, including but not limited to adding nodes and relationships in hot areas, optimizing the display order of hot topics, etc.; after adjusting the content, the final intellectual property knowledge graph is generated to ensure that the graph content is closer to user needs and improve the practicality and user experience of the graph.

[0061] The embodiments of the present invention can achieve the following beneficial effects through the above steps: By recording and analyzing user interaction behaviors, we can accurately identify key areas and hot topics that users are concerned about, making the graph content more in line with user needs; Adjust the graph content based on user demand hotspots, add nodes and relationships in hot areas, optimize the display order of hot topics, and improve the practicality and user experience of the graph; By integrating the user behavior analysis module, dynamic updating of the graph content is achieved, so that the final intellectual property knowledge graph can continuously adapt to changes in user needs and maintain its timeliness and effectiveness.

[0062] In view of the importance of user demand to the optimization of knowledge graph content, it is necessary to make targeted adjustments to the content of the second intellectual property knowledge graph. Based on this, the present invention provides a specific embodiment. The step 403, based on the user demand hotspots, adjusts the content of the second intellectual property knowledge graph to generate the final intellectual property knowledge graph, which specifically includes the following steps: Step 411: Analyze the user demand hot spots by using natural language processing technology, extract key words, and quantify the importance of the key words by using a semantic similarity algorithm to obtain a quantitative result reflecting the hot spot area; In this step, the semantic similarity algorithm refers to an algorithm for quantifying the semantic similarity between words, thereby evaluating the importance of the words.

[0063] This step uses natural language processing technology to analyze users' query patterns and preference information, identify key areas of concern and hot topics of users, extract important words that reflect users' concerns from user demand hotspots, and quantify the importance of key words through semantic similarity algorithms to obtain quantitative results that reflect hot areas.

[0064] Step 412: According to the quantification result, the weights of the entities and the relationships between the entities in the second intellectual property knowledge graph are adjusted to obtain a third intellectual property knowledge graph that highlights hot areas; This step adjusts the weights of entities related to hot areas in the second intellectual property knowledge graph based on the quantification results to make these entities more prominent in the graph; adjusts the relationship weights between entities to make the relationships between entities related to hot areas more obvious; through the above adjustments, a third intellectual property knowledge graph that highlights hot areas is generated.

[0065] Step 413: Based on the third intellectual property knowledge graph, an active learning algorithm is used to dynamically adjust the graph content according to user feedback information, optimize the display of nodes and edges, and obtain an optimized fourth intellectual property knowledge graph; In this step, the active learning algorithm refers to a machine learning algorithm that improves the performance of the model by selecting the most valuable data samples for labeling; feedback information refers to user feedback on recommended content, including but not limited to click-through rate, dwell time, etc.

[0066] This step uses an active learning algorithm based on the third intellectual property knowledge graph to select the most valuable data samples for optimization based on user feedback information; dynamically adjusts the graph content based on user feedback information to make the content in hot areas more prominent; optimizes the display of nodes and edges in the graph to make the content in hot areas more intuitive and easy to understand; through the above adjustments, an optimized fourth intellectual property knowledge graph is generated.

[0067] Step 414: Based on the fourth intellectual property knowledge graph, a personalized recommendation mechanism is introduced to generate a final intellectual property knowledge graph; In this step, the personalized recommendation mechanism refers to a recommendation system that provides personalized recommendation content to users based on their historical behaviors and preferences.

[0068] This step introduces a personalized recommendation mechanism based on the fourth intellectual property knowledge graph, and recommends highly relevant knowledge graph content fragments based on the user's historical query records and preferences; through the personalized recommendation mechanism, the final intellectual property knowledge graph is generated to make the graph content closer to the user's actual needs.

[0069] The embodiments of the present invention can achieve the following beneficial effects through the above steps: By analyzing and quantifying the hot spots of user needs, we adjust the weights of entities and relationships in the graph to make the content in hot areas more prominent. Active learning algorithms are used to dynamically adjust the graph content based on user feedback, optimize the display of nodes and edges, and make the graph content more intuitive and easy to understand; Introduce a personalized recommendation mechanism to provide users with more personalized recommendation content based on their historical behavior and preferences, thereby improving user experience and satisfaction.

[0070] Figure 2 A schematic diagram of a big data processing system for implementing hybrid data analysis is provided for an embodiment of the present invention. Figure 2 As shown, the system includes: The parsing module 21 is used to construct knowledge cells of the intellectual property document and parse the knowledge cells using natural language processing technology to identify and abstract entities, attributes of entities and relationships between entities in the field of intellectual property to obtain entity information; An establishing module 22, used to establish a first intellectual property knowledge graph based on the knowledge cells and the entity information; A determination module 23 is used to optimize the structure of the first intellectual property knowledge graph using a graph neural network technology to obtain a second intellectual property knowledge graph, wherein the graph neural network technology includes a graph neural network model and a graph convolution operation; The generation module 24 is used to integrate the user behavior analysis module into the second intellectual property knowledge graph, so as to identify user demand hotspots by tracking user query patterns and preference information, adjust the content of the second intellectual property knowledge graph, and obtain the final intellectual property knowledge graph.

[0071] Figure 2 The system for constructing the intellectual property knowledge graph can execute Figure 1The implementation principle and technical effect of the method for constructing an intellectual property knowledge graph described in the embodiment are not described in detail. The specific way in which each module and unit performs operations in the system for constructing an intellectual property knowledge graph in the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated here.

[0072] In one possible design, Figure 2 The system for constructing an intellectual property knowledge graph of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0073] The processing component 32 is used to: construct knowledge cells of the intellectual property document, and use natural language processing technology to parse the knowledge cells, identify and abstract entities, attributes of entities and relationships between entities in the field of intellectual property, and obtain entity information; establish a first intellectual property knowledge graph based on the knowledge cells and the entity information; use graph neural network technology to optimize the structure of the first intellectual property knowledge graph to obtain a second intellectual property knowledge graph, and the graph neural network technology includes a graph neural network model and a graph convolution operation; integrate a user behavior analysis module into the second intellectual property knowledge graph to identify user demand hotspots by tracking user query patterns and preference information, and adjust the content of the second intellectual property knowledge graph to obtain a final intellectual property knowledge graph.

[0074] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0075] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0076] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0077] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0078] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0079] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0080] The embodiment of the present invention further provides a computer storage medium storing a computer program, which can achieve the above-mentioned Figure 1 A method for constructing an intellectual property knowledge graph in the illustrated embodiment.

[0081] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0082] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0083] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0084] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing an intellectual property knowledge graph, characterized in that: include: Construct knowledge cells of intellectual property documents, and parse the knowledge cells using natural language processing technology to identify and abstract entities, attributes of entities, and relationships between entities in the field of intellectual property to obtain entity information; Based on the knowledge cells and the entity information, establish a first intellectual property knowledge graph; Optimizing the structure of the first intellectual property knowledge graph using graph neural network technology to obtain a second intellectual property knowledge graph, wherein the graph neural network technology includes a graph neural network model and a graph convolution operation; Integrate the user behavior analysis module into the second intellectual property knowledge graph to identify user demand hotspots by tracking user query patterns and preference information, adjust the content of the second intellectual property knowledge graph, and obtain the final intellectual property knowledge graph.

2. The method according to claim 1, characterized in that The method of using the graph neural network technology to optimize the structure of the first intellectual property knowledge graph to obtain the second intellectual property knowledge graph, wherein the graph neural network technology includes a graph neural network model and a graph convolution operation, including: Obtaining node information and edge information in the first intellectual property knowledge graph, and performing feature encoding on the node information and edge information respectively to obtain a node feature vector and an edge feature vector; Based on the node feature vector and the edge feature vector, iteratively updating the node representation through the graph neural network model, and extracting the multi-level association pattern between the nodes; Based on the association pattern, the combined features of the node and its neighboring nodes are extracted using graph convolution operations, and the mutual influence between the node attributes and the graph structure is analyzed through the combined features to obtain an enhanced node representation; Based on the enhanced node representation, the edge information in the first intellectual property knowledge graph is adjusted through an optimization algorithm to determine the optimal structural representation of the node, and a second intellectual property knowledge graph is generated based on the optimal structural representation.

3. The method according to claim 2, characterized in that The node representation is iteratively updated through a graph neural network model based on the node feature vector and the edge feature vector to extract a multi-level association pattern between nodes, including: Initializing the node feature vector and the edge feature vector, and using the initialized node feature vector and edge feature vector as input of a graph neural network model; In each iteration step, the message passing mechanism of the graph neural network model is used to pass messages, so that the node receives information from the neighboring nodes, and fuses the received information with the node's own characteristics to obtain node fusion information; the node fusion information is processed using the graph convolution layer to update the node representation; it is determined whether the change amplitude of the updated node representation is less than the preset threshold; If not, the steps of message transmission, information fusion and threshold judgment are repeated until the predetermined convergence standard is reached and the iterative update is completed; the predetermined convergence standard is that the change amplitude of the updated node representation is less than the preset threshold, or the preset number of iterations is reached; wherein, the updated node representation in each iterative step is used to reflect the multi-level association pattern between nodes.

4. The method according to claim 3, characterized in that Based on the association pattern, the combined features of the node and the neighboring nodes are extracted by using graph convolution operation, and the mutual influence between the node attributes and the graph structure is analyzed by using the combined features to obtain the enhanced node representation, including: Identify the position of the node and the position of the neighboring nodes in the graph structure, and extract the feature information of the node and the neighboring nodes through graph convolution operation; the graph structure is a structure composed of updated node representations; Combining the feature information to obtain local graph features of the node, wherein the local graph features include neighbor node information and node attributes; Based on the local graph features, analyzing node attributes and the positions of nodes in the graph structure to determine the influence of the node attributes on the connection mode of the nodes and the roles of the nodes, and obtaining enhanced node representations, wherein the positions include the degree and centrality of the nodes; Based on the enhanced node representation, the identification step, the extraction step, the combination step and the analysis step are repeatedly performed until a predetermined convergence condition is reached, wherein the convergence condition is that the change amplitude of the enhanced node representation is less than a preset threshold value, or a preset number of iterations is reached; through multiple iterative processes, the enhanced node representation is optimized so that each enhanced node representation contains its own attribute information and the influence of neighboring nodes, thereby obtaining an enhanced node representation.

5. The method according to claim 4, characterized in that Based on the local graph features, the node attributes and the position of the node in the graph structure are analyzed to determine the influence of the node attributes on the connection mode of the node and the role of the node, and an enhanced node representation is obtained, wherein the position includes the degree and centrality of the node, including: Based on the local graph features, the nodes are preliminarily encoded using graph embedding technology, and the preliminarily encoded node attributes are converted into numerical feature vectors to obtain a preliminarily encoded result; The preliminary encoding result is inputted using the graph neural network architecture, the node attributes are dynamically analyzed in combination with the attention mechanism, and different weights are set corresponding to different node attributes to obtain the node representation to be updated; Based on the node representation to be updated, a structured deep learning framework is introduced to evaluate the position characteristics of the nodes in the graph structure, and the dependencies between the nodes are captured through a high-order neighborhood aggregation algorithm to quantify the relative importance of the nodes in the graph structure, thereby obtaining a relative importance quantification result; Based on the relative importance quantification results, a community detection algorithm is used to determine the influence of the node attributes on the connection mode of the nodes and the role of the nodes, and the information between the nodes is transmitted through the graph convolution layer and the node representation to be updated is updated to obtain the updated node representation; Based on the updated node representation, a feature enhancement strategy is implemented, the intrinsic attributes and external structure of the node are analyzed through ensemble learning, and the quality of the node representation is improved through a self-supervised learning mechanism to obtain an enhanced node representation.

6. The method according to claim 1, characterized in that The user behavior analysis module is integrated into the second intellectual property knowledge graph to track user query patterns and preference information, identify user demand hotspots, and adjust the content of the second intellectual property knowledge graph to obtain a final intellectual property knowledge graph, including: Integrate a user behavior analysis module into the second intellectual property knowledge graph to collect user query patterns and preference information by recording user interaction behaviors to obtain query data, wherein the user interaction behaviors include query history and click habits; Analyze the query data, identify key areas and hot topics of user concern, and obtain user demand hotspots; Based on the user demand hotspots, the content of the second intellectual property knowledge graph is adjusted to generate a final intellectual property knowledge graph.

7. The method according to claim 6, characterized in that The step of adjusting the content of the second intellectual property knowledge graph based on the user demand hotspots to generate a final intellectual property knowledge graph includes: Analyze the user demand hot spots by using natural language processing technology, extract key words, and quantify the importance of the key words by using a semantic similarity algorithm to obtain quantitative results reflecting the hot areas; According to the quantification result, the weights of entities and relationships between entities in the second intellectual property knowledge graph are adjusted to obtain a third intellectual property knowledge graph that highlights hot areas; Based on the third intellectual property knowledge graph, an active learning algorithm is used to dynamically adjust the graph content and optimize the display of nodes and edges according to user feedback information to obtain an optimized fourth intellectual property knowledge graph; Based on the fourth intellectual property knowledge graph, a personalized recommendation mechanism is introduced to generate the final intellectual property knowledge graph.

8. A big data processing system for implementing hybrid data analysis, characterized in that: include: A parsing module is used to construct knowledge cells of intellectual property documents and parse the knowledge cells using natural language processing technology to identify and abstract entities, attributes of entities, and relationships between entities in the field of intellectual property to obtain entity information; An establishment module, used to establish a first intellectual property knowledge graph based on the knowledge cells and the entity information; A determination module, configured to optimize the structure of the first intellectual property knowledge graph using a graph neural network technology to obtain a second intellectual property knowledge graph, wherein the graph neural network technology includes a graph neural network model and a graph convolution operation; A generation module is used to integrate the user behavior analysis module into the second intellectual property knowledge graph, so as to identify user demand hotspots by tracking user query patterns and preference information, adjust the content of the second intellectual property knowledge graph, and obtain the final intellectual property knowledge graph.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a method for constructing an intellectual property knowledge graph as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, it implements a method for constructing an intellectual property knowledge graph as described in any one of claims 1 to 7.

Images