A knowledge graph processing method and system

Abstract

The embodiments of this specification provide a knowledge graph processing method and system, the method comprising: selecting a number of nodes and their edges from a shared knowledge graph based on one or more entity types involved in a target business domain to obtain a target sub-graph; the shared knowledge graph is obtained based on the fusion of knowledge graphs of one or more business domains; processing the target sub-graph to extract one or more graph features, the graph features including some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, graph rule features; providing the graph features to a target data processing task in a target business domain; wherein the graph features are used together with task customization features as input features of the target data processing task to implement the target data processing task.

Description

Technical Field

[0001] This specification relates to the field of data processing technology, and in particular to knowledge graph processing methods and systems. Background Art

[0002] In various business fields, there is corresponding data, and each business field needs to use data to achieve business tasks or goals. For example, the model can be trained through training samples to obtain a model with classification, relationship prediction and other capabilities. Knowledge graph data includes entities and relationships, which can cover more comprehensive or complete information. If rich graph features can be extracted from the knowledge graph and used for tasks or goals in various business fields, it will help greatly improve work efficiency.

[0003] Therefore, knowledge graph processing methods and systems are urgently needed to improve the effectiveness of business applications. Summary of the invention

[0004] In one aspect, the present specification provides a knowledge graph processing method, comprising: selecting a number of nodes and their edges from a shared knowledge graph based on one or more entity types involved in a target business domain to obtain a target sub-graph; the shared knowledge graph is obtained based on the fusion of knowledge graphs of one or more business domains; processing the target sub-graph to extract one or more graph features, the graph features including some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, graph rule features; providing the graph features to a target data processing task in a target business domain; wherein the graph features are used together with task customization features as input features of the target data processing task to implement the target data processing task.

[0005] Another aspect of the present specification provides a knowledge graph processing system, including: a sub-graph determination module, which is used to select a number of nodes and their edges from a shared knowledge graph based on one or more entity types involved in a target business domain to obtain a target sub-graph; the shared knowledge graph is obtained based on the fusion of knowledge graphs of one or more business domains; a graph feature acquisition module, which is used to process the target sub-graph to extract one or more graph features, and the graph features include some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, and graph rule features; a task processing module, which is used to provide the graph features to a target data processing task in a target business domain, wherein; the graph features are used together with task customization features as input features of the target data processing task to implement the target data processing task.

[0006] Another aspect of the present specification provides a knowledge graph processing device, including at least one storage medium and at least one processor, wherein the at least one storage medium is used to store computer instructions; the at least one processor is used to execute the computer instructions to implement the knowledge graph processing method. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0008] Figure 1 It is a schematic diagram of an application scenario of a knowledge graph processing system according to some embodiments of this specification;

[0009] Figure 2 is a block diagram of a knowledge graph processing system according to some embodiments of this specification;

[0010] Figure 3 is an exemplary flow chart of a knowledge graph processing method according to some embodiments of this specification;

[0011] Figure 4 is a schematic diagram of a method for processing a target sub-map according to some embodiments of this specification;

[0012] Figure 5 It is a schematic diagram of a shared knowledge graph according to some embodiments of this specification. DETAILED DESCRIPTION

[0013] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0014] It should be understood that the "system", "device", "unit" and / or "module" used in this specification is a method for distinguishing different components, elements, parts, parts or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0015] As shown in this specification and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0016] Flowcharts are used in this specification to illustrate the operations performed by the system according to the embodiments of this specification. It should be understood that the preceding or following operations are not necessarily performed precisely in order. Instead, the steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or one or more operations may be removed from these processes.

[0017] Figure 1 It is a schematic diagram of an application scenario of a knowledge graph processing system according to one or more embodiments of this specification.

[0018] The application scenario 100 may involve various business scenarios in various business fields, for example, business fields such as security, insurance, payment, and wealth.

[0019] There are data processing tasks for various business scenarios in different business fields. For example, in the payment business field, it is necessary to determine whether multiple payment accounts belong to the same merchant in order to clean up the ownership of a large number of payment accounts in the business data. For example, it can be determined whether multiple merchants belong to the same owner to clarify the business relationship between the merchants. In a specific data processing task, the business party may extract task-related features (or task-customized features) based on the existing data in this business domain. Exemplarily, for the data processing task of determining whether multiple payment accounts belong to the same merchant, the business party can extract the geographic location information of each payment account (which can be expressed in the form of a payment code), and then determine which payment accounts belong to the same merchant through the distance between the locations (such as payment accounts with a closer distance are more likely to belong to the same merchant). It can be considered that geographic location information is closely related to the task and has intuitive interpretability, and it can be considered that geographic location information belongs to task-customized features. However, in addition to such features that are closely related to the task, relatively broad and abstract auxiliary features can also be mined from a large amount of business data. For example, each payment account can have multiple attributes, such as registration time, location, registration device identification, etc., and payment accounts belonging to the same merchant may have similar attribute distribution and payment relationship distribution. Therefore, if a knowledge graph can be built based on business data and node representation vectors of each payment node can be obtained, then these node representation vectors can be used as auxiliary features together with task-customized features to complete data processing tasks (such as combining the distance between the geographical locations of payment accounts and the similarity of node representation vectors to determine multiple payment accounts belonging to the same merchant). Auxiliary features as a supplement to task-customized features may further improve prediction accuracy.

[0020] In view of this, some embodiments of this specification propose to construct a knowledge graph based on entities and / or relationships related to data processing tasks, and extract graph features for use in data processing tasks to improve the efficiency of business applications. For each business field, data processing tasks can be performed based on the knowledge graph constructed using data. In some embodiments, when performing data processing tasks, each business field can form a link from knowledge graph data processing (such as graph data processing, graph model training, etc.) to business applications based on business tasks or business goals. Different business applications need to establish their own graph data processing to business application links. Among them, the work involved in the graph data processing of different business applications may be repeated, resulting in a waste of invested resources and manpower. In addition, the business data coverage of a single business field is limited. If the data of multiple business fields can be shared, it will be more beneficial to the efficient and accurate implementation of data processing tasks.

[0021] In view of the above situation, some embodiments of this specification provide a knowledge graph processing method and system, which processes knowledge graph data based on a shared knowledge graph obtained by fusing knowledge graphs of one or more business domains, including selecting several nodes and their edges from the shared knowledge graph based on one or more entity types involved in the target business domain to obtain a target sub-graph, and processing the target sub-graph to extract one or more graph features, which may include: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, graph rule features, etc. The graph features are then provided to the target data processing task of the target business domain, so that the graph features are used together with the task customization features as the input features of the target data processing task to achieve the target data processing task. Through the knowledge graph processing method and system provided by some embodiments of this specification, the knowledge graph construction and graph feature parts of the data processing links in each business field are extracted, and are uniformly implemented by the knowledge graph platform, providing various graph features for each business field, effectively improving data processing efficiency, saving resources and labor costs. On the other hand, the method or system provided in this specification enables the target business party to introduce data from various business fields to use more complete knowledge graph data to perform data processing tasks, so as to greatly improve the implementation effect of business tasks or goals.

[0022] like Figure 1 As shown, the application scenario 100 of the knowledge graph processing system may include multiple servers such as servers 110 - 1, 110 - 2, 110 - 3, a processing device 120 and a network 130.

[0023] Multiple servers such as servers 110-1, 110-2, 110-3, etc. can correspond to multiple platforms or business areas respectively. Servers 110-1, 110-2, 110-3, ... can be used to manage resources and process data and / or information such as various graph features from at least one component of the system or an external data source (e.g., a cloud data center), and can implement various data processing tasks in the platform or business field. In some embodiments, each of servers 110-1, 110-2, 110-3, ... can be a single server or a server group. The server group can be centralized or distributed (for example, server 110-1 can be a distributed system), can be dedicated, or can be provided by other devices or systems at the same time. In some embodiments, servers 110-1, 110-2, 110-3, ... can be regional or remote. In some embodiments, servers 110-1, 110-2, 110-3, ... can be implemented on a cloud platform or provided in a virtual manner. By way of example only, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, the like, or any combination thereof.

[0024] In some embodiments, any one or more of the servers 110 - 1 , 110 - 2 , 110 - 3 , . . . may store data of the corresponding platform or business domain, such as data instances, knowledge graphs, and the like.

[0025] Any one or more of the servers 110-1, 110-2, 110-3, ... may include a processor 112. The processor 112 may process data and / or information obtained from other devices or system components, such as various graph features. The processor may execute program instructions based on these data, information and / or processing results to perform one or more functions described in the present application. In some embodiments, the processor 112 may include one or more sub-processing devices (e.g., a single-core processing device or a multi-core multi-core processing device). By way of example only, the processor 112 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), an application specific instruction processor (ASIP), a graphics processing unit (GPU), a physical processing unit (PPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a programmable logic circuit (PLD), a controller, a microcontroller unit, a reduced instruction set computer (RISC), a microprocessor, etc., or any combination thereof.

[0026] In some embodiments, the processing device 120 may correspond to a server of a knowledge graph platform. The processing device 120 may process data and / or information obtained from other devices or system components. The processing device 120 may execute program instructions based on these data, information and / or processing results to perform one or more functions described in this application. For example, the processing device 120 may obtain two or more knowledge graphs from two or more of the servers 110-1, 110-2, 110-3, ... through the network 130 to obtain a shared knowledge graph that integrates knowledge graphs of multiple business fields. The processing device 120 may select several nodes and their edges from the shared knowledge graph based on one or more entity types involved in the target business domain to obtain a target sub-graph, and process the target sub-graph to extract one or more graph features, and provide the graph features to the servers 110-1, 110-2, 110-3, ... In some embodiments, the processing device 120 may include one or more sub-processing devices (e.g., a single-core processing device or a multi-core multi-core processing device). By way of example only, the processing device 120 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), an application specific instruction processor (ASIP), a graphics processing unit (GPU), a physical processing unit (PPU), a digital signal processor (DSP), a field programmable gate array (FPGA), a programmable logic device (PLD), a controller, a microcontroller unit, a reduced instruction set computer (RISC), a microprocessor, etc., or any combination of the above.

[0027] The network 130 can connect the components of the system and / or connect the system with external parts. The network 130 allows the components of the system and the system and external parts to communicate with each other and promote the exchange of data and / or information. In some embodiments, the network 130 can be any one or more of a wired network or a wireless network. For example, the network 130 can include a cable network, an optical fiber network, a telecommunications network, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), a public switched telephone network (PSTN), a Bluetooth network, a ZigBee network (ZigBee), a near field communication (NFC), a bus within a device, a line within a device, a cable connection, etc. or any combination thereof. In some embodiments, the network connection between the various parts of the system can adopt one of the above methods or multiple methods. In some embodiments, the network 130 can be a point-to-point, shared, central, etc., various topological structures or a combination of multiple topological structures. In some embodiments, the network 130 can include one or more network access points. For example, network 130 may include wired or wireless network access points, such as base stations and / or network switching points 130-1, 130-2, ..., through which one or more components of system 100 may connect to network 130 to exchange data and / or information.

[0028] Figure 2 It is a block diagram of a knowledge graph processing system according to some embodiments of this specification.

[0029] In some embodiments, the knowledge graph processing system 200 can be implemented on the processing device 120 .

[0030] In some embodiments, the knowledge graph processing system 200 may include a sub-graph determination module 210, a graph feature acquisition module 220, and a task processing module 230. In some embodiments, the knowledge graph processing system 200 may further include a recall module 240. In some embodiments, the graph feature acquisition module 220 may further include a graph decomposition unit 221 and a homogeneous graph feature acquisition unit 222.

[0031] In some embodiments, the subgraph determination module 210 can be used to select several nodes and their edges from the shared knowledge graph based on one or more entity types involved in the target business domain to obtain a target subgraph; the shared knowledge graph is obtained by fusion of knowledge graphs based on one or more business domains. In some embodiments, the subgraph determination module 210 can also be used to obtain macro features of the target subgraph, the macro features include one or more of the following: number of entities, degree distribution of the graph, connectivity distribution of the graph, data quality score of the graph; based on the macro features, determine whether the target subgraph meets the requirements, if not, modify the target subgraph or re-acquire the target subgraph from the shared knowledge graph.

[0032] In some embodiments, the graph feature acquisition module 220 can be used to process the target sub-graph to extract one or more graph features, wherein the graph features include some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, and graph rule features. In some embodiments, the graph structure features include one or more of the following: degree information, PageRank value, point clustering coefficient, close centrality, eigenvector centrality, common neighbor index, Katz index, and random walk similarity.

[0033] In some embodiments, the target sub-graph may be a heterogeneous graph. In some embodiments, the graph decomposition unit 221 may be used to decompose the target sub-graph into a plurality of homogeneous graphs.

[0034] In some embodiments, the homogeneous spectrum feature acquisition unit 222 can be used to process the homogeneous spectrum separately to extract one or more spectrum features.

[0035] In some embodiments, the task processing module 230 can be used to provide the graph features to a target data processing task of a target business domain; wherein the graph features are used together with the task customization features as input features of the target data processing task to implement the target data processing task. In some embodiments, the target data processing task is entity classification, prediction of relationships between entities, or entity set mining.

[0036] In some embodiments, the recall module 240 can be used to recall several candidate nodes from the shared knowledge graph based on the target data processing task; the candidate nodes are the processing objects of the target data processing task; wherein the recall method includes querying in the shared knowledge graph based on the retrieval conditions to obtain the candidate nodes, or obtaining the candidate nodes from the shared knowledge graph through vector retrieval based on the target vector.

[0037] It should be understood that the system and its modules shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art will understand that the above-mentioned methods and systems can be implemented using computer executable instructions and / or included in processor control codes, such as such as carrier media such as disks, CDs or DVD-ROMs, programmable memories such as read-only memories (firmware), or data carriers such as optical or electronic signal carriers. Such codes are provided on the system and its modules of this specification. Not only can the hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc. be implemented, they can also be implemented with software such as executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software (e.g., firmware).

[0038] It should be noted that the above description of the system and its modules is only for convenience of description and does not limit the present specification to the scope of the embodiments. It is understandable that, after understanding the principle of the system, those skilled in the art may arbitrarily combine the modules or form a subsystem to connect with other modules without deviating from the principle.

[0039] Figure 3 It is an exemplary flowchart of a knowledge graph processing method shown in some embodiments of this specification.

[0040] In some embodiments, the method 300 may be performed by the processing device 120. In some embodiments, the method 300 may be implemented by the knowledge graph processing system 200 deployed on the processing device 120.

[0041] like Figure 3 As shown, the method 300 may include:

[0042] Step 310, select several nodes and their edges from the shared knowledge graph based on one or more entity types involved in the target business domain to obtain a target subgraph.

[0043] In some embodiments, step 310 may be performed by the sub-graph determination module 210 .

[0044] A knowledge graph refers to a knowledge base consisting of a series of nodes representing entities and edges representing the relationships between entities.

[0045] Among them, entity is a broad abstraction of objective individuals, which can refer to tangible objects in the physical world, such as natural persons, cars, merchants, etc., or intangible objects, such as payment accounts, identity identifiers, wifi accounts, funds, program codes, etc. Entities can have multiple types. For example, natural persons, merchants, payment accounts, identity identifiers, and wifi accounts can all be a type of entity. Among them, each entity can correspond to multiple entity instances. For example, the natural person entity can include data instances such as Zhang San, Li Ming, and Wang Nian.

[0046] There can be relationships between entities, for example, merchant A has business dealings with merchant B, merchant C is a sub-merchant of merchant A, and Zhang San is the manager of merchant A. There can be many types of relationships between entities, such as belonging to a relationship, employment relationship, payment relationship, etc.

[0047] A shared knowledge graph refers to a knowledge graph that can be used in multiple business fields. In some embodiments, a shared knowledge graph can be obtained by merging two or more knowledge graphs, wherein the data in the shared knowledge graph can come from one business field or from multiple business fields. In some embodiments, the multiple business fields can be related business fields, such as finance, payment, security, and other similar or overlapping business fields.

[0048] In some embodiments, the shared knowledge graph can be obtained by fusing knowledge graphs of multiple business domains (such as insurance, payment, etc.). Specifically, knowledge graphs of multiple business domains can be obtained, and the node attributes or edge attributes in each knowledge graph can be standardized by an attribute standardization operator, or the same nodes in each knowledge graph can be fused by an entity fusion operator, or a new relationship can be established between nodes by a link finger operator, etc. Figure 5 As shown, knowledge graph 1 and knowledge graph 2 can be fused, wherein machine A and tool A are fused into one entity tool A, and a shared knowledge graph 3 including the two knowledge graphs is obtained, and the graph data of the two knowledge graphs are expressed in a unified manner and the data is connected and shared.

[0049] Generally speaking, shared knowledge graph data covers a wide range, but its scale is also large. For example, the number of nodes reaches hundreds of billions. The target business domain refers to the business domain corresponding to the business application to be carried out, which can be one of the multiple business domains involved in the graph fusion, or it can be other business areas. In the business domain, one or more entity types may be involved. For example, in the payment field, multiple entity types such as merchants, natural persons, payment accounts, identity identifiers, business licenses, mobile phone numbers, etc. are involved. In some embodiments, the multiple entity types involved can be determined based on the common data processing tasks in the target business domain, and then based on the one or more entity types involved in the target business domain, the several nodes and their edges are selected from the shared knowledge graph.

[0050] Selecting several nodes and their edges from the shared knowledge graph based on one or more entity types involved in the target business domain means selecting nodes corresponding to the entity types involved in the target business domain and the edges associated with these nodes from the shared knowledge graph. For example, if the target business domain is the payment field, the entity types involved include merchants, natural persons, payment accounts, identity identifiers, business licenses, and mobile phone numbers. Then, the nodes corresponding to the entities belonging to merchants (such as the first convenience store, the second supermarket, etc.), the entities belonging to natural persons (such as Zhang San, Li Ming, Wang Nian, etc.), the entities belonging to payment accounts (such as account 51522, account 51324, etc.), the entities belonging to identity identifiers (such as identity identifier 3123, identity identifier 3224, etc.), the entities belonging to business licenses (such as business license number 321, business license number 311, etc.), and the entities belonging to mobile phone numbers (mobile phone number 212367, mobile phone number 212346, etc.) can be selected, and the edges associated with these nodes can also be selected.

[0051] In some embodiments, the selected nodes and edges may include entity types and inter-entity relationships involved or related to various data processing tasks in the target business domain. For example, in the payment field, data processing tasks may include determining whether multiple payment accounts are from the same merchant, determining whether multiple merchants are from the same operator, etc., and the selected nodes and edges include the nodes and edges corresponding to the entity types and inter-entity relationships that the aforementioned data processing tasks may involve or be related to.

[0052] In some embodiments, the required nodes and edges can be selected from the shared knowledge graph by various selection methods. For example, the shared knowledge graph can be queried based on search conditions such as entity type, relationship information, etc. to obtain the required nodes and edges.

[0053] In this specification, a knowledge graph consisting of several nodes and edges selected from a shared knowledge graph based on one or more entity types involved in a target business domain can be called a target subgraph.

[0054] In some embodiments, after obtaining the target sub-graph, the macro features of the target sub-graph can be obtained. Macro features refer to features that can reflect the overall characteristics or statistical information of the graph. In some embodiments, the macro features may include one or more of the following: the number of entities, the degree distribution of the graph, the connectivity distribution of the graph, the data quality score of the graph, etc.

[0055] Among them: the number of entities refers to the number of entities included in the graph, for example, 10,000. The number of entities can reflect the data scale of the target subgraph. The degree distribution of the graph refers to the degree of each node or entity in the graph. The degree refers to the number of edges connected to a node, and also refers to the number of other nodes connected to a node. The degree can include out-degree and in-degree. The out-degree refers to the number of edges pointing to other nodes from the node, and the in-degree refers to the number of edges pointing to the node. The degree distribution can reflect the hotspot distribution of the graph (such as the target subgraph). The connectivity distribution of the graph refers to the distribution of the connectivity of each node. Connectivity means that there are edges between nodes. The associations of nodes in areas with better connectivity are more extensive or more complete, and the associations of nodes in areas with poor connectivity are relatively simple. The data quality score of the graph refers to the score of the data quality of the nodes and edges of the graph. The more data gaps and / or the lower the data accuracy, the lower the data quality score.

[0056] In some embodiments, the target sub-graph can be evaluated based on the macro features of the target sub-graph. In some embodiments, it can be determined based on the macro features whether the target sub-graph meets the requirements, such as whether the number of entities is greater than a threshold, whether the degree distribution of the graph meets the conditions, whether the connectivity distribution of the graph meets the conditions, whether the data quality score of the graph is greater than a threshold, etc. Among them, the requirements can be determined based on actual needs or experience, and this embodiment does not limit this.

[0057] In some embodiments, if it is determined based on the macro features that the target subgraph does not meet the requirements, the target subgraph can be modified (for example, nodes, edges, relevant information of entities corresponding to the nodes, relevant information of the relationships between entities corresponding to the edges, etc. are supplemented from the shared knowledge graph) or the target subgraph can be re-acquired from the shared knowledge graph based on new retrieval conditions to obtain a target subgraph that meets the requirements.

[0058] Step 320: Process the target sub-graph to extract one or more graph features.

[0059] In some embodiments, step 320 may be performed by the graph feature acquisition module 220 .

[0060] Graph features refer to the characteristic information contained in the knowledge graph, which may include some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, and graph rule features.

[0061] A node representation vector refers to a vector used to represent node information (such as node type, node attribute information), and an edge representation vector refers to a vector used to represent edge information (such as edge type, edge attribute information). In some embodiments, a target subgraph can be represented and learned by a graph neural network model such as GNN, GCN, Graph LSTM, etc. to obtain a node representation vector for each node and an edge representation vector for each edge.

[0062] Graph structural features refer to the structural information features in the graph, which can characterize structural information with specific meanings.

[0063] In some embodiments, the graph structure features may include one or more of the following: degree information, PageRank value, point clustering coefficient, close centrality, eigenvector centrality, common neighbor index, Katz index, random walk similarity, etc. In some embodiments, the graph structure features may be obtained by statistics, graph structure algorithms, model calculations, and the like.

[0064] Degree information refers to the degree information of a node. Degree refers to the number of edges connected to a node, and also refers to the number of other nodes connected to a node. Degree can include out-degree and in-degree. Out-degree refers to the number of edges pointing from the node to other nodes, and in-degree refers to the number of edges pointing to the node.

[0065] The PageRank value is an indicator used to characterize the importance of a node in a graph. In some embodiments, the PageRank value of a node can be determined based on the situation of the edge to which the node is connected and the PageRank values ​​of other nodes to which the node is connected. For example, if node Zhang San is connected to an edge, then the node can be given extra points. If node Zhang San is not connected to an edge, then the node can be subtracted from the node. The PageRank values ​​of each neighboring node of node Zhang San can be weighted averaged to obtain a score for node Zhang San. Based on the scores, extra points, and subtractions obtained by weighted average of neighboring nodes, a comprehensive score for node Zhang San can be obtained and used as the PageRank value of node Zhang San.

[0066] The point clustering coefficient is an indicator used to describe the degree of association between the adjacent nodes of a node. For example, it is used to describe the degree of interconnection between the adjacent nodes of a node. The more edges between adjacent nodes, the higher the point clustering coefficient. For example, many fans are adjacent nodes of star A, but there is little or no association between fans. The point clustering coefficient of the node of star A is low.

[0067] Closeness centrality is an indicator used to represent the length of the path from a node to other nodes. It can be represented by the average shortest path from a node to all other nodes. The smaller the value of the average shortest path, the shorter the path from this point to all other points, which means that this node is closer to all other nodes.

[0068] Eigenvector centrality (or feature vector centrality) is used to represent the likelihood of a node being visited in a random walk of infinite length. Eigenvector centrality can be characterized by the eigenvector score of a node. A node that is connected to other adjacent nodes with higher eigenvector scores will have a higher score than a node that has adjacent nodes with lower eigenvector scores. A higher eigenvector score can mean that the node is connected to many nodes with higher eigenvector scores, that is, closer to the center.

[0069] The common neighbor index refers to an index used to characterize the potential relationship and proximity between two nodes. It can be obtained through various common neighbor algorithms. For example, the common neighbor algorithm is used to obtain the common neighbor nodes of two nodes, and further infer the potential relationship and proximity between the two nodes based on the neighbor nodes.

[0070] The Katz index is an index used to describe how many paths can reach another node from one node. For example, you can get the set of all paths from one node to another node, sum the path lengths of the path set, and get the path length from one node to another node (for example, a node A to another node B needs to pass through 3 edges, then the set of paths from node A to node B is 3 paths, one of which can be considered to have a length of 1, and the path lengths of the path set are summed to get the path length from node A to node B as 3). Based on the path length, you can get the corresponding Katz value.

[0071] Random walk similarity refers to the similarity between nodes determined based on a random walk method that starts from a specific node and walks along randomly selected adjacent edges for multiple steps to reach other nodes. In some embodiments, the random walk similarity can be calculated by a random walk model, a local random walk model, etc. For example, the transition probability vector of the node can be obtained by a random walk model, and the similarity of the two nodes can be obtained by calculating the relative entropy of the transition probability vectors of the two nodes. Among them, the transition probability refers to the probability of reaching other nodes in the network after multiple steps of random walk from a specific node, and the transition probability vector is a vector representation of the probability of reaching all other nodes in the network after multiple steps of random walk from a specific node.

[0072] The semantic features of text information in a graph refer to the semantic features of text information in the graph data (e.g., attribute information of nodes, attribute information of edges, etc.). In some embodiments, the semantic features of text information in a graph can be obtained by various methods such as natural language processing models, feature extraction algorithms, or feature representation algorithms. For example, the text information can be input into a natural language processing model such as BERT, RNN, Transformer, ESIM, etc. to obtain the corresponding text representation vector, and the semantic features of the text information can be represented by the text representation vector.

[0073] Graph rule features refer to the features of various graph rules. Graph rules refer to the relevant rules of nodes and edges or the laws of graph data, which can be used for business reasoning, decision-making, verification, and constraints. For example, father's father is grandfather, and a company can only have one legal person, which are all graph rules. In some embodiments, graph rule features can be obtained by manual mining. For example, graph rule features can be obtained by rule mining algorithms such as SFE.

[0074] Through this embodiment, multiple and diverse graph features can be uniformly obtained for various common data processing tasks in the target business domain. Among them, multiple graph features such as graph structure features, graph rule features, and semantic features can provide specific meanings for the feature information of the graph, making the graph features more concrete and explainable.

[0075] In some embodiments, the target sub-graph may include multiple different types of nodes corresponding to multiple different types of entities (such as people, payment accounts, wifi accounts, etc.) and multiple different types of edges corresponding to multiple different types of relationships between entities (such as belonging relationships, employment relationships, payment relationships, etc.), that is, the target sub-graph is a heterogeneous graph.

[0076] In some embodiments, when the target sub-graph is a heterogeneous graph, the target sub-graph can be decomposed into multiple homogeneous graphs, and the homogeneous graphs are processed separately to extract one or more graph features. For more details on decomposing the target sub-graph into multiple homogeneous graphs to extract one or more graph features, please refer to Figure 4 and related instructions.

[0077] Step 330, providing the graph features to the target data processing task of the target business domain; wherein the graph features are used together with the task customized features as input features of the target data processing task to implement the target data processing task.

[0078] In some embodiments, step 330 may be performed by the task processing module 230 .

[0079] The target data processing task refers to the data processing task to be performed, which may include various data processing tasks of business applications.

[0080] In some embodiments, the target data processing task may be entity classification, prediction of relationships between entities, entity set mining, etc.

[0081] Entity classification refers to the task of classifying entities (such as binary classification, multi-classification, etc.). For example, for an entity XX Technology Company, determine its corresponding risk category.

[0082] Entity relationship prediction refers to the task of predicting the association between entities. For example, for entity Zhang San and multiple companies, predict the association between Zhang San and the company to find out which company Zhang San works for. For another example, for multiple payment accounts, predict the association between multiple payment accounts to determine whether multiple payment accounts belong to the same merchant.

[0083] Entity set mining refers to mining entity sets consisting of multiple entities to understand relevant group information such as the group situation and the roles of each entity in the group. For example, for an entity set consisting of multiple natural persons, determine whether the multiple persons are a criminal gang, and determine which of the multiple persons are the core members of the gang.

[0084] In some embodiments, the target data processing task may be implemented by various task processing methods, such as a graph reasoning method, a model prediction method, etc. This embodiment does not limit this.

[0085] In some embodiments, the obtained one or more graph features are provided to the target business domain, and the target business party can select one or more graph features (for example, node representation vectors and edge representation vectors, or one or more graph structure features, or node representation vectors, edge representation vectors and graph rule features, or node representation vectors, edge representation vectors and semantic features) from the graph features that can be used for the target data processing task, and use the selected graph features as input features of the target data processing task to implement the target data processing task and obtain the processing task results.

[0086] In some embodiments, the selected graph features and the task customized features can be used together as input features of the target data processing task to implement the target data processing task and obtain the processing task result.

[0087] In the target business domain, graph features are used as a supplementary input or background knowledge, and the corresponding features are task-customized features. Task-customized features refer to features extracted from existing data in the target business domain based on the target data processing task. Generally speaking, they are more relevant to the target data processing task or have a more intuitive interpretability in relation to the processing results. In some embodiments, the input features of the target data processing task other than the graph features provided by the knowledge graph platform may be referred to as task-customized features. In other words, task-customized features are generally generated in the target business domain, or extracted by the target business party based on the target data processing task. To some extent, it can be understood that different target data processing tasks have different task-customized features, but can share graph features as auxiliary features. For an exemplary description of task-customized features, see also Figure 1 The relevant content will not be repeated here.

[0088] As an example, if the target data processing task is to determine whether multiple payment accounts belong to the same merchant, the graph structure features of the payment account (for example, reflecting whether the terminal device that transfers money to the payment account is connected to the same media such as wifi when transferring money), graph rule features and semantic features can be selected from the various graph features obtained, and the selected graph features and task customization features are used as input features of the target data processing task to determine whether multiple payment accounts belong to the same merchant. For another example, for the type annotation of word segmentation in the text, in addition to using the semantic features of the word segmentation (i.e., task customization features), the node representation vector corresponding to the word segmentation of the synonym graph can also be obtained, and it can be used as auxiliary information together with the semantic features as the input feature of the word segmentation type annotation. Through this embodiment, by constructing a target sub-graph related to the target business domain and producing a variety of graph features, it can bring gains to specific processing tasks in the business domain, complete data processing tasks more efficiently and effectively, and obtain more accurate task results. Moreover, based on graph features with specific meanings such as graph structure features, graph rule features and semantic features, we can provide explanations for the impact or effect of various data information on the results of data processing tasks, and further improve the implementation effect of tasks (such as the accuracy of prediction tasks, recognition tasks, etc.).

[0089] In some embodiments, the implementation of the target data processing task may also include recalling several candidate nodes from the shared knowledge graph based on the target data processing task to perform the target data processing task on the candidate nodes and obtain the processing task results for the candidate nodes. For example, for the target data processing task that needs to predict which company Zhang San is employed in, in order to reduce the data processing pressure, multiple candidate company nodes where Zhang San may be employed can be selected from the shared knowledge graph based on certain principles. After that, Zhang San's personal information and the relevant features of each candidate node can be obtained as task customization features, and the graph structure features between the natural person and the company node can be obtained from the graph features as auxiliary information, and then the company where Zhang San is located can be predicted from multiple candidate company nodes. In other words, the candidate node can be the processing object of the target data processing task. For another example, multiple groups of nodes can be obtained from the shared knowledge graph, and then multiple candidate node sets can be obtained to predict the properties of these candidate node sets. In some embodiments, recalling candidate nodes can be performed by the recall module 240.

[0090] In some embodiments, the shared knowledge graph may be queried based on the search conditions to obtain candidate nodes. For example, the recall module 240 may query multiple company nodes with the same geographic information as candidate nodes from the shared knowledge graph based on Zhang San's residence information. For another example, the recall module 240 may query multiple groups of nodes with larger point clustering coefficients from the shared knowledge graph, thereby obtaining multiple candidate node sets.

[0091] In some embodiments, the candidate node can also be obtained from the shared knowledge graph through vector retrieval based on the target vector. For example, if the target data processing task is to determine whether multiple payment accounts belong to the same merchant, a target payment account (such as the payment account of the target merchant) can be determined, and a corresponding feature representation vector, i.e., a target vector, is generated for the target payment account. The target vector is matched with the node representation vector of each node in the knowledge graph to obtain a node representation vector that matches or is similar to the target vector (such as a vector distance less than a set threshold), and the payment account node corresponding to the matching or similar node representation vector is used as a candidate node.

[0092] In some embodiments, the node representation vector of each node in the graph (including the payment account node) can be obtained by learning the representation of the graph. Since the node representation vector integrates the information of the neighbor nodes or edges of each node, as well as various graph features such as graph structure features, it can improve the accuracy and coverage of matching, and thus improve the accuracy and coverage of candidate node recall. At the same time, vector-based retrieval can greatly shorten the recall time and further improve work efficiency.

[0093] Figure 4 It is a schematic diagram of a method for processing a target sub-map according to some embodiments of this specification.

[0094] In some embodiments, the method 400 may be implemented by the processing device 120. In some embodiments, the method 400 may be implemented by the graph decomposition unit 221 and the homogeneous graph feature acquisition unit 222 in the knowledge graph processing system 200.

[0095] like Figure 4 As shown, the method 400 may include:

[0096] Step 410, decomposing the target sub-graph into multiple homogeneous graphs.

[0097] In some embodiments, step 410 may be performed by the graph decomposition unit 221 .

[0098] In some embodiments, the target subgraph is a heterogeneous graph, and the target subgraph can be decomposed into multiple homogeneous graphs, wherein a homogeneous graph refers to a knowledge graph that includes only one entity type and one relationship type. For example, the target subgraph is a heterogeneous graph a that includes several entity types such as people, payment accounts, and wifi accounts, and multiple types of relationships between these entities. The target subgraph a can be decomposed into the following homogeneous graphs: a homogeneous graph that only includes people and interpersonal relationships between people (can be called a social graph) b, a homogeneous graph that only includes payment accounts and payment relationships between payment accounts (can be called a payment graph) c, and a homogeneous graph that only includes wifi accounts and binding relationships between wifi accounts (can be called a usage medium graph) d.

[0099] In some embodiments, the target sub-graph as a heterogeneous graph can be decomposed into multiple homogeneous graphs by various graph decomposition methods or graph extraction methods. For example, the corresponding homogeneous graph can be constructed by extracting nodes belonging to the same entity type and edges belonging to the same relationship type.

[0100] Step 420 , processing the homogeneous graph to extract one or more graph features.

[0101] In some embodiments, step 420 may be performed by the homogeneous spectrum feature acquisition unit 222 .

[0102] In some embodiments, the homogeneous spectrum obtained by decomposition can be subjected to spectrum data processing to extract a variety of spectrum features. For more information about the method of performing spectrum data processing on the spectrum to obtain a variety of spectrum features, please refer to step 320 and its related description, which will not be repeated here. For multiple different homogeneous spectra, spectrum data processing can be performed on multiple homogeneous spectra respectively to obtain corresponding multiple sets of spectrum features (a set of spectrum features can include one or more spectrum features of the homogeneous spectrum).

[0103] Different homogeneous graphs may have different graph structure meanings (such as a person's social relationship structure, a payment account's payment relationship structure, and a wifi account's ownership relationship structure), and the corresponding graph features may represent different graph structure meanings. Through this embodiment, for the target sub-graph as a heterogeneous graph, a more detailed and concrete set of graph features with multiple different structural meanings can be obtained.

[0104] It should be noted that the above description of the processes and methods is only for illustration and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the processes and methods under the guidance of this specification. However, these modifications and changes are still within the scope of this specification. For example, the order of steps in the processes and methods is changed, and steps in different processes and methods are combined.

[0105] The embodiments of this specification also provide a knowledge graph processing device, including at least one storage medium and at least one processor, wherein the at least one storage medium is used to store computer instructions; the at least one processor is used to execute the computer instructions to implement a knowledge graph processing method. The method may include: selecting a number of nodes and their edges from a shared knowledge graph based on one or more entity types involved in a target business domain to obtain a target sub-graph; the shared knowledge graph is obtained based on the fusion of knowledge graphs of multiple business domains; processing the target sub-graph to extract one or more graph features, wherein the graph features include some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, and graph rule features; providing the graph features to a target data processing task of a target business domain; wherein the graph features are used together with task customization features as input features of the target data processing task to implement the target data processing task.

[0106] The beneficial effects that may be brought about by the embodiments of this specification include but are not limited to: (1) When performing data processing tasks based on knowledge graphs, a shared knowledge graph that integrates data from multiple business domains is used to use more complete knowledge graph data to perform data processing tasks, and the various information included in the knowledge graph data is used more effectively to obtain multiple and diverse graph features. For different business applications, the required graph features can be selected from them to use more complete feature data to achieve the target data processing tasks, which is more efficient and can greatly improve the implementation effect of business tasks or goals; (2) Based on the target sub-graph determined from the shared knowledge graph, multiple and diverse graph features with specific meanings are obtained, such as graph structure features, graph rule features, and semantic features, which can provide explanations for the influence or role of various data information in business application results, and can effectively help further improve the implementation effect of business tasks or goals (such as the accuracy of prediction tasks, recognition tasks, etc.). It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other beneficial effects that may be obtained.

[0107] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of this specification. Although not explicitly stated here, those skilled in the art may make various modifications, improvements and corrections to this specification. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

[0108] At the same time, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures or characteristics in one or more embodiments of this specification can be appropriately combined.

[0109] In addition, it will be understood by those skilled in the art that various aspects of this specification may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of this specification may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of this specification may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0110] A computer storage medium may include a propagated data signal containing computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0111] The computer program code required for the operation of each part of this specification can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, Visual Basic, Fortran2003, Perl, COBOL2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program code can be run entirely on the user's computer, or run on the user's computer as an independent software package, or run partially on the user's computer and partially on a remote computer, or run entirely on a remote computer or processing device. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0112] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this specification, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing processing device or mobile device.

[0113] Similarly, it should be noted that in order to simplify the description disclosed in this specification and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this specification, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the subject matter of this specification are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0114] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0115] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, documents, etc., cited in this specification are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the contents of this specification, documents that limit the broadest scope of the claims of this specification (currently or later attached to this specification) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the materials attached to this specification are inconsistent or conflicting with the contents described in this specification, the descriptions, definitions, and / or use of terms in this specification shall prevail.

[0116] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. A knowledge graph processing method, comprising: Select several nodes and their edges from the shared knowledge graph based on one or more entity types involved in the target business domain to obtain a target subgraph; The shared knowledge graph is obtained by fusion of knowledge graphs based on one or more business domains; Processing the target sub-graph to extract one or more graph features, wherein the graph features include some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, and graph rule features; The graph features are provided to the target data processing tasks in the target business domain; wherein the graph features are used together with the task customized features as input features of the target data processing tasks to implement the target data processing tasks; in the target business domain, the graph features are used as input supplements or background knowledge, and the task customized features are extracted by the target business party based on the target data processing tasks, and are input features outside the shared knowledge graph.

2. According to the method as claimed in claim 1, the graph structure features include one or more of the following: degree information, PageRank value, point clustering coefficient, closeness centrality, eigenvector centrality, common neighbor index, Katz index, and random walk similarity.

3. The method of claim 1, further comprising: Recalling a number of candidate nodes from the shared knowledge graph based on the target data processing task; The candidate node is a processing object of the target data processing task; Among them, the recall method includes querying in the shared knowledge graph based on the retrieval condition to obtain the candidate node, or obtaining the candidate node from the shared knowledge graph through vector retrieval based on the target vector.

4. The method according to claim 1, wherein the step of selecting a plurality of nodes and their edges from a shared knowledge graph based on one or more entity types involved in the target business domain to obtain a target subgraph further comprises: Obtaining macro features of the target sub-graph, wherein the macro features include one or more of the following: the number of entities, the degree distribution of the graph, the connectivity distribution of the graph, and the data quality score of the graph; Based on the macro features, it is determined whether the target sub-graph meets the requirements. If not, the target sub-graph is modified or the target sub-graph is re-acquired from the shared knowledge graph.

5. The method according to claim 1, wherein the target sub-graph is a heterogeneous graph; and the processing of the target sub-graph to extract one or more graph features comprises: Decompose the target sub-graph into multiple homogeneous graphs; The homogeneous spectra are processed separately to extract one or more spectra features.

6. The method according to claim 1, wherein the target data processing task is entity classification, prediction of relationships between entities, or entity set mining.

7. A knowledge graph processing system, comprising: A sub-graph determination module is used to select a number of nodes and their edges from a shared knowledge graph based on one or more entity types involved in a target business domain to obtain a target sub-graph; the shared knowledge graph is obtained by fusion of knowledge graphs of one or more business domains; A graph feature acquisition module, used to process the target sub-graph to extract one or more graph features, wherein the graph features include some or all of the following: node representation vectors, edge representation vectors, graph structure features, semantic features of text information in the graph, and graph rule features; A task processing module is used to provide the graph features to the target data processing task in the target business domain; wherein the graph features are used together with the task customized features as input features of the target data processing task to implement the target data processing task; in the target business domain, the graph features are used as input supplements or background knowledge, and the task customized features are extracted by the target business party based on the target data processing task, and are input features outside the shared knowledge graph.

8. In the system of claim 7, the graph structure features include one or more of the following: degree information, PageRank value, point clustering coefficient, closeness centrality, eigenvector centrality, common neighbor index, Katz index, random walk similarity.

9. The system according to claim 7, further comprising a recall module, wherein the recall module is configured to: Recalling a number of candidate nodes from the shared knowledge graph based on the target data processing task; the candidate nodes are processing objects of the target data processing task; in, The recall method includes querying in the shared knowledge graph based on the retrieval condition to obtain the candidate node, or obtaining the candidate node from the shared knowledge graph through vector retrieval based on the target vector.

10. The system according to claim 7, wherein the sub-graph determination module is further configured to: Obtaining macro features of the target sub-graph, wherein the macro features include one or more of the following: the number of entities, the degree distribution of the graph, the connectivity distribution of the graph, and the data quality score of the graph; Based on the macro features, it is determined whether the target sub-graph meets the requirements. If not, the target sub-graph is modified or the target sub-graph is re-acquired from the shared knowledge graph.

11. The system according to claim 7, wherein the target sub-graph is a heterogeneous graph; the graph feature acquisition module further comprises a graph decomposition unit and a homogeneous graph feature acquisition unit; The graph decomposition unit is used to decompose the target sub-graph into multiple homogeneous graphs; The homogeneous spectrum feature acquisition unit is used to process the homogeneous spectrum respectively to extract one or more spectrum features.

12. In the system as claimed in claim 7, the target data processing task is entity classification, prediction of the relationship between entities or entity set mining.

13. A knowledge graph processing device, comprising at least one storage medium and at least one processor, wherein the at least one storage medium is used to store computer instructions; the at least one processor is used to execute the computer instructions to implement the knowledge graph processing method as described in any one of claims 1-6.

Images