Knowledge graph completion method, device, equipment and computer program product

Abstract

The embodiment of the invention relates to a knowledge graph completion method, device and equipment and a computer program product. The method includes obtaining a manufacturing knowledge graph. The method further comprises the steps of obtaining a plurality of entity embedding, a plurality of attribute embedding and a plurality of relation embedding of the manufacturing knowledge graph, wherein the plurality of entity inserts, the plurality of attribute inserts, and the plurality of relationship inserts are generated by the ontology map data, the manufacturing knowledge map data, and a mapping relationship between the ontology map data and the manufacturing knowledge map data. In addition, the method further comprises the step of complementing the manufacturing knowledge graph based on the multiple entity embedding, the multiple attribute embedding and the multiple relation embedding. According to the graph completion scheme provided by the embodiment of the invention, not only can the information of the manufacturing knowledge graph be utilized, but also the information of the ontology graph can be utilized, so that the relationship between entities can be fully mined to realize more detailed graph completion, and the comprehensiveness and accuracy of graph completion are improved.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the field of computers, and more specifically, to methods, devices, equipment, and computer program products for knowledge graph completion. Background Art

[0002] Knowledge graphs present entity data and their relationships in a graphical structure, enabling efficient organization, storage, and retrieval of knowledge. With the rapid development of technologies like artificial intelligence and natural language processing, knowledge graphs are finding widespread application in fields like search engines, recommendation systems, medical diagnosis, and industrial manufacturing. The importance of knowledge graphs continues to grow, and their widespread application presents new opportunities and possibilities for technological innovation and social development.

[0003] Knowledge graph completion refers to the process of improving and expanding the knowledge graph through automated methods, filling in missing information or relationships, and continuously optimizing the completeness and accuracy of the knowledge graph. Through knowledge graph completion, the system can more fully understand the relationships between entities, providing users with more accurate and comprehensive information retrieval and services. Summary of the Invention

[0004] Embodiments of the present disclosure provide a method, apparatus, device, and computer program product for knowledge graph completion.

[0005] According to the first aspect of the present disclosure, a method for completing a knowledge graph is provided. The method includes obtaining a manufacturing knowledge graph, wherein the manufacturing knowledge graph includes multiple entities associated with product manufacturing, multiple attributes of the multiple entities, and multiple relationships between the multiple entities. The method also includes obtaining multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings of the manufacturing knowledge graph, wherein the multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings are generated by ontology graph data, manufacturing knowledge graph data, and a mapping relationship between the ontology graph data and the manufacturing knowledge graph data. In addition, the method also includes completing the manufacturing knowledge graph based on the multiple entity embeddings, the multiple attribute embeddings, and the multiple relationship embeddings.

[0006] According to a second aspect of the present disclosure, a method for using a knowledge graph is provided. The method includes utilizing the manufacturing domain knowledge graph completed by the method of the first aspect of the present disclosure and processing a usage request related to the use of the knowledge graph.

[0007] According to a third aspect of the present disclosure, a device for completing a knowledge graph is provided. The device includes a knowledge graph acquisition unit, which is configured to acquire a manufacturing knowledge graph, wherein the manufacturing knowledge graph includes multiple entities associated with product manufacturing, multiple attributes of the multiple entities, and multiple relationships between the multiple entities. The device also includes a graph embedding acquisition unit, which is configured to acquire multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings of the manufacturing knowledge graph, wherein the multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings are generated through ontology graph data, manufacturing knowledge graph data, and a mapping relationship between the ontology graph data and the manufacturing knowledge graph data. In addition, the device also includes a knowledge graph completion unit, which is configured to complete the manufacturing knowledge graph based on multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings.

[0008] According to a fourth aspect of the present disclosure, an electronic device is provided. The electronic device includes at least one processor; and a memory coupled to the at least one processor and having instructions stored thereon, which, when executed by the at least one processor, cause the device to perform the steps of the methods of the first and second aspects of the present disclosure.

[0009] According to a fifth aspect of the present disclosure, a computer program product is provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions that, when executed, cause a computer to perform the steps of the method of the first and second aspects of the present disclosure.

[0010] According to a sixth aspect of the present disclosure, a machine-readable storage medium is provided, wherein the machine-readable storage medium stores machine-executable instructions, wherein the machine-executable instructions are executed by a processor to implement the steps of the method in the first and second aspects of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other objects, features and advantages of the present disclosure will become more apparent through a more detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings, wherein like reference numerals generally represent like components throughout the exemplary embodiments of the present disclosure.

[0012] Figure 1 A schematic diagram illustrating an example environment in which devices and / or methods according to embodiments of the present disclosure may be implemented;

[0013] Figure 2 A flowchart of a method for completing a knowledge graph according to an embodiment of the present disclosure is shown;

[0014] Figure 3A A schematic diagram illustrating a process of generating node embedding of an ontology graph according to an embodiment of the present disclosure is shown;

[0015] Figure 3B A schematic diagram of an ontology graph according to an embodiment of the present disclosure is shown;

[0016] Figure 4A A schematic diagram illustrating a process of generating node embedding of a manufacturing knowledge graph according to an embodiment of the present disclosure is shown;

[0017] Figure 4B A schematic diagram of a manufacturing knowledge graph according to an embodiment of the present disclosure is shown;

[0018] Figure 4C A schematic diagram of a bipartite graph according to an embodiment of the present disclosure is shown;

[0019] Figure 4D A schematic diagram of an ontology entity graph according to an embodiment of the present disclosure is shown;

[0020] Figure 5 A flowchart illustrating a process of training a knowledge graph completion model according to an embodiment of the present disclosure is shown;

[0021] Figure 6 A flowchart showing a process of completing a manufacturing knowledge graph according to an embodiment of the present disclosure is shown;

[0022] Figure 7 A schematic diagram illustrating an apparatus for completing a knowledge graph according to an embodiment of the present disclosure is shown; and

[0023] Figure 8 Illustrated is a schematic block diagram of an example device suitable for implementing embodiments of the present disclosure.

[0024] In the various drawings, the same or corresponding reference numerals denote the same or corresponding parts. DETAILED DESCRIPTION

[0025] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0026] As mentioned earlier, knowledge graphs are playing an increasingly important role in fields such as knowledge storage and artificial intelligence. Knowledge graph completion can improve, update, and expand knowledge graphs, and therefore also plays a very important role. When implementing knowledge graph completion, related technologies for knowledge graph completion often only use the structural information of the knowledge graph itself to generate embeddings of entities, relationships, and attributes in the knowledge graph, and then use the generated multiple embeddings to train the completion model and perform knowledge graph completion. Performing knowledge graph completion in this way only uses the knowledge graph itself and does not utilize the information of the knowledge graph's ontology. The ontology is a collection of abstract concepts in the knowledge graph that can describe the common characteristics of entities within a certain range and the relationships between entities.

[0027] To this end, an embodiment of the present disclosure proposes a knowledge graph completion solution, which uses multiple embeddings generated through ontology graphs, knowledge graphs, and mapping relationships between ontology graphs and knowledge graphs to complete the knowledge graph. Specifically, a manufacturing knowledge graph is first obtained, which includes entities related to product manufacturing, entity attributes, and relationships between entities; then, entity embeddings, entity attribute embeddings, and relationship embeddings of entities in the manufacturing knowledge graph are obtained, and the entity embeddings, entity attribute embeddings, and relationship embeddings are generated using ontology graph data, manufacturing knowledge graph data, and mapping relationships between ontology graph data and manufacturing knowledge graph data; then, the manufacturing knowledge graph is completed using entity embeddings, entity attribute embeddings, and relationship embeddings.

[0028] Therefore, according to the solution of the embodiment of the present disclosure, when completing the manufacturing knowledge graph, not only the information of the manufacturing knowledge graph itself can be utilized, but also the information of the ontology graph. The introduction of the ontology graph makes it possible to fully explore the relationship between entities in the process of completing the manufacturing knowledge graph, and achieve more detailed information completion. By fully integrating the manufacturing knowledge graph and the ontology graph, the comprehensiveness and accuracy of the graph completion can be improved, the information integrity of the graph can be improved, and the accuracy of the graph query can be improved.

[0029] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. Figure 1 An example environment 100 is shown in which devices and / or methods of embodiments of the present disclosure may be implemented.

[0030] like Figure 1 As shown, the example environment 100 includes a manufacturing knowledge graph 102, which may include a plurality of entities, relationships, and attributes, wherein the entities, relationships, and attributes may be collectively referred to as nodes. Figure 1The nodes of the manufacturing knowledge graph in the subsequent figures include letters and numbers, where the letter E indicates that the node is an entity, the letter R indicates that the node is a relationship between entities, the letter A indicates that the node is an attribute of the entity, and the number is the serial number. The manufacturing knowledge graph is a knowledge graph applied in the industrial and manufacturing fields. It is built based on knowledge graph technology and aims to integrate and organize information related to industrial production, manufacturing processes, equipment, materials, products, etc. The manufacturing knowledge graph can contain various entities (such as machines, equipment, materials, locations), relationships (such as production processes, supply relationships, product composition), attributes (for example, generation time, identifiers), and other information.

[0031] refer to Figure 1 , the manufacturing knowledge graph 102 includes entities 104, relationships 106, and entities 108, etc. It should be understood that for the purpose of illustration, Figure 1 The manufacturing knowledge graph 102 in the figure only shows a small number of nodes, and the manufacturing knowledge graph actually used may include more nodes. In some embodiments, entity 104 may be a tire, entity 108 may be a vehicle, and relationship 106 may be "belongs to product", indicating that entity 104 is part of entity 108, and the relationship between the two is "belongs to vehicle". In addition, the manufacturing knowledge graph 102 also includes multiple different nodes such as relationship 110, attribute 112, entity 114, and entity 116. For example, entity 116 may be a factory, and relationship 110 may be "manufactured in factory", indicating that entity 104 (tire) is manufactured in entity 116 (factory). Entity 114 may be assembly equipment, and its attribute 112 may be production time. It should be understood that other entities may also have corresponding attributes, and only attribute 112 of entity 114 is listed here as an example of entity attributes.

[0032] The example environment 100 shows an input computing device 134, examples of which include but are not limited to computers, laptops, tablets, smartphones, servers, cloud services, and the like. A knowledge graph completion model 118 can be deployed on the computing device 134, which can receive the manufacturing knowledge graph 102 and complete it. During the manufacturing knowledge graph completion process, the knowledge graph completion model 118 can obtain entity embeddings, relationship embeddings, and attribute embeddings in the manufacturing knowledge graph 102 from the memory 120. Embedding is a technology that maps high-dimensional data to a low-dimensional space in order to capture relationships and patterns in the data. Embedding can map information such as entities, relationships, and attributes to a low-dimensional vector space, so that the computing device 134 and the knowledge graph completion model 118 can better understand and process these entities, relationships, and attributes. For example, entity embeddings, relationship embeddings, and attribute embeddings can be 128-dimensional embedding vectors, or embedding vectors of other dimensions, and the dimensionality of the three embedding vectors can be the same or different, and the present disclosure does not limit this.

[0033] refer to Figure 1 , the embodiments of the present disclosure can use the ontology graph 122, the manufacturing knowledge graph 124 and the bipartite graph 126 to generate entity embeddings, relationship embeddings and attribute embeddings, and save them in the memory 120. The ontology graph 122 is a collection of abstract concepts in the manufacturing field, which can describe the common characteristics of entities in the manufacturing field and the relationships between entities. For example, entity 104 in the manufacturing knowledge graph 102 is a tire, entity 108 is a vehicle, and entity 116 is a factory. In the ontology graph 122, the category corresponding to entity 104 can be a component, the category corresponding to entity 108 can be a product, and the category corresponding to entity 116 can be a production location. In other words, the ontology graph 122 can specify the categories of entities in the manufacturing knowledge graph 102 and the manufacturing knowledge graph 124. In addition, the ontology graph 122 can also specify ontology relationships and ontology attributes. It should be understood that both manufacturing knowledge graph 102 and manufacturing knowledge graph 124 are manufacturing knowledge graphs, with manufacturing knowledge graph 102 being the manufacturing knowledge graph that needs to be completed, and manufacturing knowledge graph 124 being the manufacturing knowledge graph used to generate entity embeddings, relationship embeddings, and attribute embeddings. Bipartite graph 126 can represent the mapping relationship between ontology graph 122 and manufacturing knowledge graph 124. For example, bipartite graph 126 can be used to map entity 104 in manufacturing knowledge graph 124 to a component category in ontology graph 122, and entity 108 in manufacturing knowledge graph 124 to a product category in ontology graph 122, and so on.

[0034] Continue to refer Figure 1After receiving the manufacturing knowledge graph 102 and obtaining the corresponding entity embeddings, relationship embeddings, and attribute embeddings, the knowledge graph completion model 118 can complete the manufacturing knowledge graph 102 to generate a completed manufacturing knowledge graph 128. For example, the knowledge graph completion model 118 can mine the relationship 130 between entity 108 and entity 116, indicating that entity 108 is manufactured by entity 116. Similarly, the knowledge graph completion model 118 can mine the relationship 132 between entity 114 and entity 116, indicating that entity 114 is manufactured by entity 116. Therefore, compared with the manufacturing knowledge graph 102, the completed manufacturing knowledge graph 128 can have richer entity, relationship, and attribute information.

[0035] Combined with the above Figure 1 An example environment 100 is described in which embodiments of the present disclosure can be implemented. Figure 2 A flowchart describing a method 200 for knowledge graph completion according to an embodiment of the present disclosure.

[0036] Figure 2 A flow chart of a method 200 for completing a knowledge graph according to an embodiment of the present disclosure is shown. At block 202, a manufacturing knowledge graph may be obtained, wherein the manufacturing knowledge graph includes a plurality of entities associated with product manufacturing, a plurality of attributes of the plurality of entities, and a plurality of relationships between the plurality of entities. For example, Figure 1 As shown, the computing device 134 can obtain the manufacturing knowledge graph 102, which is the manufacturing knowledge graph to be completed. The manufacturing knowledge graph 102 includes multiple entities (e.g., entity 104, entity 108, and entity 116), multiple attributes (e.g., attribute 112 and other attributes not shown), and multiple relationships (e.g., relationship 106 and relationship 110).

[0037] At block 204, multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings of the manufacturing knowledge graph may be obtained, wherein the multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings are generated by ontology graph data, manufacturing knowledge graph data, and mapping relationships between the ontology graph data and the manufacturing knowledge graph data. Figure 1 As shown, the computing device 134 can obtain multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings from the memory 120. It should be understood that the multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings can also be obtained from other places, and the present disclosure is not limited to this. The multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings can be generated through the ontology graph 122, the manufacturing knowledge graph 124, and the bipartite graph 126 (i.e., mapping relationship).

[0038] At block 206, the manufacturing knowledge graph may be completed based on multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings. Figure 1 The computing device 134 may utilize the knowledge graph completion model 118 to complete the manufacturing knowledge graph 102 based on multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings to generate a manufacturing knowledge graph 128 .

[0039] Therefore, according to the method 200 of the embodiment of the present disclosure, when completing the manufacturing knowledge graph, not only the information of the manufacturing knowledge graph itself can be utilized, but also the information of the ontology graph. The introduction of the ontology graph makes it possible to fully explore the relationship between entities in the process of completing the manufacturing knowledge graph, and achieve more detailed information completion. By fully integrating the manufacturing knowledge graph and the ontology graph, the comprehensiveness and accuracy of the graph completion can be improved, the information integrity of the graph can be improved, and the accuracy of the graph query can be improved.

[0040] Figure 3A A schematic diagram of a process 300A for generating node embeddings of an ontology graph according to an embodiment of the present disclosure is shown. Figure 3B A schematic diagram of an ontology graph 300B according to an embodiment of the present disclosure is shown. Figure 3B A process 300A for generating node embeddings for an ontology graph is described. At block 302, ontology graph data may be obtained. Figure 3B , need explanation, Figure 3B The nodes of the ontology graph in the subsequent figures include letters and numbers, where the letters OC indicate that the node is an ontology class, OR indicates that the node is an ontology relationship between ontology classes, and OA indicates that the node is an ontology attribute of an ontology class. The numbers are numbers. For example, Figure 3B In the example, category 320 can be a component category, category 324 can be a product category, and ontology relationship 322 can be a "belongs to product" relationship. Furthermore, category 328 can be a location category, ontology relationship 326 can be a "manufactured in" relationship, category 332 can be a device category, ontology attribute 330 can be a time attribute of category 332, and ontology relationship 334 can be a "manufactured by device" relationship. It should be understood that only a small number of nodes are shown here; actual ontology graphs include many more nodes.

[0041] At box 304, category embedding, ontology relationship embedding, and ontology attribute embedding can be trained based on the acquired ontology graph data. These three embeddings can be collectively referred to as node embedding. In some embodiments, node embedding can be generated by a random walk algorithm. For example, a node (category node, ontology relationship node, and ontology attribute node) can be randomly selected in the ontology graph as the starting node, and then a neighbor node is randomly selected for movement. The walk path is then recorded, and the node embedding is iteratively trained based on the walk path. In addition, the embodiments of the present disclosure can also use other methods to generate category embedding, ontology relationship embedding, and ontology attribute embedding. The present disclosure does not limit the generation method.

[0042] At block 306, trained category embeddings, ontology relationship embeddings, and ontology attribute embeddings may be generated. For example, trained node embeddings may be generated by iteratively executing a random walk algorithm. In some embodiments, multiple category embeddings, multiple ontology attribute embeddings, and multiple ontology relationship embeddings in the ontology graph data may be generated based on the ontology graph data.

[0043] Figure 4A A schematic diagram of a process 400A for generating node embeddings of a manufacturing knowledge graph according to an embodiment of the present disclosure is shown. Figure 4B A schematic diagram of a manufacturing knowledge graph 400B according to an embodiment of the present disclosure is shown. Figure 4C A schematic diagram of a bipartite graph 400C according to an embodiment of the present disclosure is shown. Figure 4D A schematic diagram of an ontology entity graph 400D according to an embodiment of the present disclosure is shown. Figure 3B as well as Figures 4B to 4D Describe the process 400A of generating node embeddings for a manufacturing knowledge graph.

[0044] like Figure 4A As shown, at block 402, ontology graph data may be obtained. For example, referring to Figure 3B As shown, the ontology graph 300B can be obtained. At block 404, the manufacturing knowledge graph 400B can be obtained. For example, referring to Figure 4B Entity 420 can be a tire, entity 424 can be a vehicle, and relationship 422 can be a "belongs to vehicle" relationship. Furthermore, entity 428 can be a factory, relationship 326 can be a "manufactured in factory" relationship, entity 432 can be assembly equipment, and attribute 430 can be the production time attribute of entity 432.

[0045] At block 406, bipartite graph data may be obtained. For example, referring to Figure 4C , the bipartite graph 400C can represent the mapping relationship between the ontology graph and the manufacturing knowledge graph. Figure 3B and Figure 4B For example, the bipartite graph 400C may indicate that the entity 420 corresponds to the category 320. As previously mentioned, the entity 420 may be a tire, and the category 320 may be a component. This mapping relationship indicates that the entity 420 is a specific instance of the category 320. In some embodiments, the category 320 may also have other specific instances, such as an engine, a transmission, a wheel, etc. Figure 4C As shown, bipartite graph 400C may also indicate mappings between entity relationships and ontology relationships, for example, entity relationship 422 may be mapped to ontology relationship 322; and indicate mappings between entity attributes and ontology attributes, for example, entity attribute 430 may be mapped to ontology attribute 330. It should be understood that the bipartite graph shown here is for illustrative purposes only, and the bipartite graph may include more mappings.

[0046] At block 408, the ontology graph and the manufacturing knowledge graph may be combined using a bipartite graph to generate an ontology entity graph. Figure 4D , the ontology graph 300B and the manufacturing knowledge graph 400B can be combined using the bipartite graph 400C to generate an ontology entity graph 400D. It can be understood that in addition to retaining the previous nodes and previous relationships of the ontology graph and the manufacturing knowledge graph, the ontology entity graph also utilizes the mapping relationship provided by the bipartite graph to connect the corresponding nodes. Therefore, the ontology entity graph can provide richer graph information. In some embodiments, ontology entity graph data including ontology graph data and manufacturing knowledge graph data can be generated based on the ontology graph data, the manufacturing knowledge graph data, and the mapping relationship.

[0047] At block 410, the category embedding, ontology relationship embedding, and ontology attribute embedding may be obtained. Figure 3A As shown, the trained category embedding, ontology relationship embedding, and ontology attribute embedding generated at box 306 can be obtained. Since there is a mapping relationship between the ontology graph and the manufacturing knowledge graph, the category embedding, ontology relationship embedding, and ontology attribute embedding can be used as prior information for the manufacturing knowledge graph. At box 412, the embedding of all nodes in the ontology entity graph can be initialized. For example, the node embedding of the trained ontology graph can be used as the node embedding of all nodes in the ontology entity graph that are related to the ontology graph. In addition, the node embedding of the trained ontology graph can also be used to initialize the node embedding of all nodes in the ontology entity graph that are related to the manufacturing knowledge graph. In some embodiments, multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings can be generated based on multiple category embeddings, multiple ontology attribute embeddings, and multiple ontology relationship embeddings.

[0048] refer to Figure 4D, the entity embedding of entity 420 can be initialized using the category embedding of category 320. In some embodiments, the category embedding of category 320 can be a 120-dimensional embedding vector, the first 64 dimensions of the category embedding can be used as the first 64-dimensional embedding vector of the entity embedding, and the second 64-dimensional vector of the entity embedding can be randomly initialized. In some embodiments, multiple initial entity embeddings, multiple initial attribute embeddings, and multiple initial relationship embeddings can be generated based on multiple category embeddings, multiple ontology attribute embeddings, and multiple ontology relationship embeddings.

[0049] Continue to refer Figure 4A , at box 414, entity embeddings, multiple attribute embeddings, and multiple relationship embeddings can be generated. These three embeddings can be collectively referred to as node embeddings. In some embodiments, node embeddings can be generated by a random walk algorithm. For example, a node (category node, ontology relationship node, ontology category node, entity node, relationship node, and attribute node) can be randomly selected in the ontology entity graph as the starting node, and then a neighbor node is randomly selected for movement. The walk path is then recorded, and the node embedding is iteratively trained based on the walk path. In addition, the embodiments of the present disclosure can also generate entity embeddings, relationship embeddings, and attribute embeddings using other methods. The present disclosure does not limit the generation method. In some embodiments, multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings can be generated based on ontology entity graph data, multiple initial entity embeddings, multiple initial attribute embeddings, and multiple initial relationship embeddings.

[0050] Through process 400 of an embodiment of the present disclosure, initialization using trained category embeddings, ontology attribute embeddings, and ontology relationships can improve the computational efficiency of entity node embeddings and provide rich prior information. In addition, using ontology entity graphs to generate entity embeddings, attribute embeddings, and relationship embeddings can improve the quality of entity node embeddings due to the more comprehensive and rich training data.

[0051] Figure 5 FIG. 5 is a flowchart of a process 500 for training a knowledge graph completion model according to an embodiment of the present disclosure. At block 502, manufacturing knowledge graph data may be obtained. For example, Figure 4B Manufacturing knowledge graph 400B in FIG. At block 504 , the manufacturing knowledge graph data may be divided into training data, validation data, and test data. In some embodiments, a knowledge graph completion model may be trained based on the manufacturing knowledge graph data, multiple entity embeddings, multiple attribute embeddings, and multiple relationship embeddings. In some embodiments, the manufacturing knowledge graph data may be divided into training data, validation data, and test data.

[0052] At box 506, the knowledge graph completion model can be trained using training data and verification data. For example, the weight parameters in the knowledge graph completion model can be adjusted using training data, and model indicators such as accuracy, recall rate, and completion rate can be generated using verification data to judge the adjustment effect of the knowledge graph completion model. If the model effect meets the predetermined metric threshold, the training can be completed. If the model effect does not meet the predetermined metric threshold, the model needs to be trained. In some embodiments, the knowledge graph completion model can be trained based on training data, verification data, and test data. In some embodiments, the knowledge graph completion model can be adjusted based on the training data. In some embodiments, a verification metric of the adjusted knowledge graph completion model can be generated based on the verification data. In some embodiments, it can be determined whether the adjusted knowledge graph completion model meets the metric threshold based on the verification metric. In some embodiments, the training of the knowledge graph completion model can be completed in response to determining that the adjusted knowledge graph completion model meets the metric threshold.

[0053] At block 508, the effectiveness of the knowledge graph completion model can be determined using the test data. For example, the knowledge graph completion model can be used to complete the test data, and a completion rate can be calculated as a metric to evaluate the effectiveness of the knowledge graph completion model. In some embodiments, a test metric can be generated using the trained knowledge graph completion model based on the test data.

[0054] Figure 6 A flowchart of a process 600 for completing a manufacturing knowledge graph according to an embodiment of the present disclosure is shown. At box 602, a manufacturing knowledge graph may be obtained. For example, a manufacturing knowledge graph to be completed may be obtained. At box 604, the entity nodes and attribute nodes may be completed using a knowledge graph completion model. For example, for entity nodes and attribute nodes that lack association relationships, the knowledge graph completion model may be used to calculate the matching scores between the entity nodes and attribute nodes in the manufacturing knowledge graph, and the entity nodes and attribute nodes whose matching scores are higher than a score threshold may be completed. In some embodiments, a plurality of first matching scores between the plurality of entities and the plurality of attributes may be determined based on the plurality of entity embeddings and the plurality of attribute embeddings. In some embodiments, the plurality of entities and the plurality of attributes may be completed based on the plurality of first matching scores.

[0055] At box 606, the knowledge graph completion model can be used to complete the entity nodes and relationship nodes. For example, for the triple <first entity node, relationship node, second entity node> that lacks an association relationship, the knowledge graph completion model can be used to determine the matching score of the triple, and for the triple with a matching score higher than the score threshold, the first entity node and the second entity node are completed using the relationship node. In some embodiments, based on the multiple entity embeddings and the multiple relationship embeddings, multiple second matching scores between the multiple relationships and the first entity and the second entity in the multiple entities can be determined. In some embodiments, based on the multiple second matching scores, the multiple relationships can be used to complete the first entity and the second entity in the multiple entities. In some embodiments, the completed manufacturing field knowledge graph can be used to process usage requests related to the use of the knowledge graph. For example, the completed manufacturing field knowledge graph can be used to process query requests and return corresponding query results.

[0056] Figure 7 A schematic diagram of an apparatus 700 for completing a knowledge graph according to an embodiment of the present disclosure is shown. The apparatus 700 includes a knowledge graph acquisition unit 702, configured to acquire a manufacturing knowledge graph, wherein the manufacturing knowledge graph includes a plurality of entities associated with product manufacturing, a plurality of attributes of the plurality of entities, and a plurality of relationships between the plurality of entities. The apparatus 700 also includes a graph embedding acquisition unit 704, configured to acquire a plurality of entity embeddings, a plurality of attribute embeddings, and a plurality of relationship embeddings of the manufacturing knowledge graph, wherein the plurality of entity embeddings, the plurality of attribute embeddings, and the plurality of relationship embeddings are generated by ontology graph data, manufacturing knowledge graph data, and a mapping relationship between the ontology graph data and the manufacturing knowledge graph data. In addition, the apparatus 700 also includes a knowledge graph completion unit 706, configured to complete the manufacturing knowledge graph based on the plurality of entity embeddings, the plurality of attribute embeddings, and the plurality of relationship embeddings.

[0057] In some embodiments, the knowledge graph completion unit 706 includes: an entity attribute completion unit configured to complete the multiple entities and the multiple attributes in the manufacturing knowledge graph based on the multiple entity embeddings and the multiple attribute embeddings; an entity relationship completion unit configured to complete the multiple entities and the multiple relationships in the manufacturing knowledge graph based on the multiple entity embeddings and the multiple relationship embeddings.

[0058] In some embodiments, the entity attribute completion unit includes: a first score determination unit, configured to determine multiple first matching scores between the multiple entities and the multiple attributes based on the multiple entity embeddings and the multiple attribute embeddings; and a graph first completion unit, configured to complete the multiple entities and the multiple attributes based on the multiple first matching scores.

[0059] In some embodiments, the knowledge graph completion unit 706 also includes: a second score determination unit, configured to determine multiple second matching scores between the multiple relationships and the first entity and the second entity among the multiple entities based on the multiple entity embeddings and the multiple relationship embeddings; and a graph second completion unit, configured to complete the first entity and the second entity among the multiple entities using the multiple relationships based on the multiple second matching scores.

[0060] In some embodiments, the multiple entity embeddings, the multiple attribute embeddings, and the multiple relationship embeddings are generated through the ontology graph data, the manufacturing knowledge graph data, and the mapping relationship, and the entity embedding generation unit includes: an ontology embedding generation unit, configured to generate multiple category embeddings, multiple ontology attribute embeddings, and multiple ontology relationship embeddings in the ontology graph data based on the ontology graph data; and an entity embedding second generation unit, configured to generate the multiple entity embeddings, the multiple attribute embeddings, and the multiple relationship embeddings based on the multiple category embeddings, the multiple ontology attribute embeddings, and the multiple ontology relationship embeddings.

[0061] In some embodiments, the second entity embedding generation unit includes: an ontology entity graph generation unit, configured to generate ontology entity graph data including the ontology graph data and the manufacturing knowledge graph data based on the ontology graph data, the manufacturing knowledge graph data and the mapping relationship; and an initial embedding generation unit, configured to generate multiple initial entity embeddings, multiple initial attribute embeddings and multiple initial relationship embeddings based on the ontology entity graph data, the multiple category embeddings, the multiple ontology attribute embeddings and the multiple ontology relationship embeddings; and a third entity embedding generation unit, configured to generate the multiple entity embeddings, the multiple attribute embeddings and the multiple relationship embeddings based on the ontology entity graph data, the multiple initial entity embeddings, the multiple initial attribute embeddings and the multiple initial relationship embeddings.

[0062] In some embodiments, the completion of the manufacturing knowledge graph is performed by a knowledge graph completion model, and the completion model training unit includes: a second completion model training unit, configured to train the knowledge graph completion model based on the manufacturing knowledge graph data, the multiple entity embeddings, the multiple attribute embeddings and the multiple relationship embeddings.

[0063] In some embodiments, the second training unit of the completion model includes: a graph data division unit, configured to divide the manufacturing knowledge graph data into training data, verification data and test data; and a third training unit of the completion model, configured to train the knowledge graph completion model based on the training data, the verification data and the test data.

[0064] In some embodiments, the third training unit of the completion model includes: a completion model adjustment unit, configured to adjust the knowledge graph completion model based on the training data; a verification metric generation unit, configured to generate a verification metric of the adjusted knowledge graph completion model based on the verification data; a metric threshold determination unit, configured to determine whether the adjusted knowledge graph completion model meets the metric threshold based on the verification metric; and a model training completion unit, configured to complete the training of the knowledge graph completion model in response to determining that the adjusted knowledge graph completion model meets the metric threshold.

[0065] In some embodiments, the third training unit of the completion model further includes: a test metric generation unit, which generates test metrics based on the test data using the trained knowledge graph completion model.

[0066] Figure 8 1 shows a schematic block diagram of a device 800 that can be used to implement an embodiment of the present disclosure. The device 800 can be a device or apparatus described in an embodiment of the present disclosure. Figure 8 As shown, the device 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) 802 or computer program instructions loaded from a storage unit 808 into a random access memory (RAM) 803. Various programs and data required for the operation of the device 800 can also be stored in the RAM 803. The CPU 801, ROM 802, and RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0067] Various components in device 800 are connected to I / O interface 805, including an input unit 806, such as a keyboard, mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a magnetic disk, optical disk, etc.; and a communication unit 809, such as a network card, modem, wireless communication transceiver, etc. The communication unit 809 allows device 800 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0068] The various methods or processes described above may be performed by the processing unit 801. For example, in some embodiments, the methods may be implemented as computer software programs that are tangibly embodied in a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed onto the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the CPU 801, one or more steps or actions in the methods or processes described above may be performed.

[0069] In some embodiments, the methods and processes described above may be implemented as a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present disclosure.

[0070] Computer-readable storage medium can be a tangible device that can keep and store the instructions used by the instruction execution device.Computer-readable storage medium can be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device or any suitable combination thereof.More specific examples (non-exhaustive list) of computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, for example, a punch card or a convex structure in a groove having instructions stored thereon, and any suitable combination thereof.Computer-readable storage medium used herein is not interpreted as a transient signal itself, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagated by waveguides or other transmission media (for example, light pulses by fiber optic cables), or electrical signals transmitted by wires.

[0071] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0072] The computer program instructions for performing the disclosed operation can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data or source code or the object code written in any combination of one or more programming languages, wherein the programming languages ​​include object-oriented programming languages, and conventional procedural programming languages.Computer-readable program instructions can be performed completely on the user's computer, partially on the user's computer, performed as an independent software package, partly on the user's computer and partly on a remote computer, or performed completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer by any type of network-including local area network (LAN) or wide area network (WAN), or can be connected to an external computer (such as utilizing an Internet service provider to connect by the Internet). In certain embodiments, by utilizing the state information of computer-readable program instructions to carry out personalized customization electronic circuits, such as programmable logic circuits, field programmable gate arrays (FPGAs) or programmable logic arrays (PLA), this electronic circuit can perform computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0073] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine such that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0074] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0075] The flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram can represent a part of a module, program segment or instruction, and the part of the module, program segment or instruction contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the blocks can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive blocks can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of the blocks in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0076] While various embodiments of the present disclosure have been described above, the above descriptions are illustrative, non-exhaustive, and not intended to be limiting of the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or technical improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method (200) for completing a knowledge graph, comprising: Obtaining (202) a manufacturing knowledge graph, wherein the manufacturing knowledge graph includes a plurality of entities associated with product manufacturing, a plurality of attributes of the plurality of entities, and a plurality of relationships between the plurality of entities; Obtaining (204) a plurality of entity embeddings, a plurality of attribute embeddings, and a plurality of relationship embeddings of the manufacturing knowledge graph, wherein the plurality of entity embeddings, the plurality of attribute embeddings, and the plurality of relationship embeddings are generated by ontology graph data, manufacturing knowledge graph data, and a mapping relationship between the ontology graph data and the manufacturing knowledge graph data; as well as The manufacturing knowledge graph is completed based on the multiple entity embeddings, the multiple attribute embeddings, and the multiple relationship embeddings (206).

2. The method (200) according to claim 1, wherein completing the manufacturing knowledge graph comprises: Based on the multiple entity embeddings and the multiple attribute embeddings, completing the multiple entities and the multiple attributes in the manufacturing knowledge graph; Based on the multiple entity embeddings and the multiple relationship embeddings, the multiple entities and the multiple relationships in the manufacturing knowledge graph are completed.

3. The method (200) according to claim 2, wherein completing the plurality of entities and the plurality of attributes in the manufacturing knowledge graph comprises: determining a plurality of first matching scores between the plurality of entities and the plurality of attributes based on the plurality of entity embeddings and the plurality of attribute embeddings; as well as The entities and the attributes are completed based on the first matching scores.

4. The method (200) according to claim 2, wherein completing the plurality of entities and the plurality of relationships in the manufacturing knowledge graph comprises: determining a plurality of second matching scores between the plurality of relationships and a first entity and a second entity of the plurality of entities based on the plurality of entity embeddings and the plurality of relationship embeddings; as well as Based on the plurality of second matching scores, a first entity and a second entity in the plurality of entities are completed using the plurality of relationships.

5. The method (200) of claim 1, wherein generating the plurality of entity embeddings, the plurality of attribute embeddings, and the plurality of relationship embeddings comprises: Based on the ontology graph data, generating a plurality of category embeddings, a plurality of ontology attribute embeddings, and a plurality of ontology relationship embeddings in the ontology graph data; as well as Based on the multiple category embeddings, the multiple ontology attribute embeddings, the multiple ontology relationship embeddings and the manufacturing knowledge graph data, the multiple entity embeddings, the multiple attribute embeddings and the multiple relationship embeddings are generated.

6. The method (200) of claim 5, wherein generating the plurality of entity embeddings, the plurality of attribute embeddings, and the plurality of relationship embeddings comprises: Based on the ontology graph data, the manufacturing knowledge graph data and the mapping relationship, generating ontology entity graph data including the ontology graph data and the manufacturing knowledge graph data; generating a plurality of initial entity embeddings, a plurality of initial attribute embeddings, and a plurality of initial relationship embeddings based on the plurality of category embeddings, the plurality of ontology attribute embeddings, and the plurality of ontology relationship embeddings; as well as Based on the ontology entity graph data, the multiple initial entity embeddings, the multiple initial attribute embeddings, and the multiple initial relationship embeddings, the multiple entity embeddings, the multiple attribute embeddings, and the multiple relationship embeddings are generated.

7. The method (200) according to claim 6, wherein completing the manufacturing knowledge graph is performed by a knowledge graph completion model, and training the knowledge graph completion model comprises: The knowledge graph completion model is trained based on the manufacturing knowledge graph data, the multiple entity embeddings, the multiple attribute embeddings, and the multiple relationship embeddings.

8. The method (200) according to claim 7, wherein training the knowledge graph completion model comprises: Dividing the manufacturing knowledge graph data into training data, verification data, and test data; as well as Based on the training data and the verification data, the knowledge graph completion model is trained.

9. The method (200) according to claim 8, wherein training the knowledge graph completion model comprises: Adjusting the knowledge graph completion model based on the training data; Generating a validation metric of the adjusted knowledge graph completion model based on the validation data; Based on the verification metric, determining whether the adjusted knowledge graph completion model meets a metric threshold; as well as In response to determining that the adjusted knowledge graph completion model satisfies the metric threshold, training of the knowledge graph completion model is completed.

10. The method (200) of claim 8, further comprising: Based on the test data, test metrics are generated using the trained knowledge graph completion model.

11. A method for using a knowledge graph, comprising: Using the manufacturing field knowledge graph completed according to the method described in any one of claims 1-10, processing usage requests related to the use of the knowledge graph.

12. A device for completing a knowledge graph, comprising: a knowledge graph acquisition unit configured to acquire a manufacturing knowledge graph, wherein the manufacturing knowledge graph includes a plurality of entities associated with product manufacturing, a plurality of attributes of the plurality of entities, and a plurality of relationships between the plurality of entities; a graph embedding acquisition unit configured to acquire a plurality of entity embeddings, a plurality of attribute embeddings, and a plurality of relationship embeddings of the manufacturing knowledge graph, wherein the plurality of entity embeddings, the plurality of attribute embeddings, and the plurality of relationship embeddings are generated by ontology graph data, manufacturing knowledge graph data, and a mapping relationship between the ontology graph data and the manufacturing knowledge graph data; as well as The knowledge graph completion unit is configured to complete the manufacturing knowledge graph based on the multiple entity embeddings, the multiple attribute embeddings and the multiple relationship embeddings.

13. An electronic device comprising: at least one processor; as well as A memory coupled to the at least one processor and having instructions stored thereon, the instructions, when executed by the at least one processor, causing the apparatus to perform the method according to any one of claims 1 to 11.

14. A computer program product tangibly stored on a non-transitory computer readable medium and comprising machine-executable instructions for performing the method according to any one of claims 1 to 11.

Images