Multi-knowledge graph embedding fusion method based on dyeing modeling and path optimization
By employing a coloring-based modeling and path optimization approach, a parallel optimization method for multi-knowledge graph relationship graphs is constructed, which solves the problems of low quality and efficiency in multi-knowledge graph fusion and achieves efficient multi-knowledge graph embedding and fusion.
Patent Information
- Application Number
- CN202510798410.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-16
- Publication Date
- 2025-11-21
AI Technical Summary
Existing technologies, when fusion of multiple knowledge graphs, ignore the potential relationships between the graphs, resulting in poor fusion quality and low efficiency.
A coloring-based modeling and path optimization approach is adopted to construct a relationship graph of multiple knowledge graphs. Through parallel optimization strategies and Hamiltonian path modeling, the embedding and fusion process of multiple knowledge graphs is optimized to ensure that the execution cost is minimized.
While ensuring the quality of embedding, it improves the overall efficiency of multi-knowledge graph embedding and fusion, and achieves the optimal solution for multi-knowledge graph fusion.
Smart Images
Figure CN120996151A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and more specifically, to a multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization. Background Technology
[0002] In open contexts and practical applications, entities often involve numerous different knowledge graphs, meaning that these knowledge graphs share some or all of the same knowledge. Clearly, the knowledge contained in a single knowledge graph is limited, and it cannot achieve optimal performance in reasoning and applications. Therefore, to improve the application capabilities of knowledge graphs, it is necessary to integrate multiple knowledge graphs to obtain a more complete knowledge base, thereby enhancing the overall representation quality and completing the knowledge graph completion tasks for each other.
[0003] Existing work, such as knowledge graph fusion and knowledge graph embedding fusion, is geared towards two knowledge graphs, aiming to improve the overall quality of the fused graph.
[0004] In reality, there are numerous types and quantities of knowledge graphs, and the relationships between them are intricate. For example, computer science is also applied to fields such as chemistry or geology, and there are more or less connections between them. However, when existing fusion methods are applied to multiple knowledge graphs, they need to fuse the knowledge graphs pairwise sequentially, ignoring the potential relationships between them. This results in poor quality and low efficiency of the fused graph.
[0005] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention
[0006] The purpose of this invention is to provide a multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization, which can improve the overall efficiency of multi-knowledge graph embedding and fusion while ensuring a certain embedding quality.
[0007] This invention provides a multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization, comprising the following steps: S1: Transforming the relationship graph of multiple knowledge graphs according to the similarity of multiple knowledge graphs to obtain a fusion unit relationship graph; S2: Performing coloring modeling and parallel optimization based on the fusion unit relationship graph to obtain optimized parallel units; S3: Calculating the successive execution cost based on the optimized parallel units to obtain a weighted relationship graph between parallel units; S4: Performing Hamiltonian path modeling based on the weighted relationship graph between parallel units to obtain a Hamiltonian path model; S5: Obtaining the optimal execution path based on the Hamiltonian path model using an optimization objective function.
[0008] Further, step S1 specifically includes: S11: performing similarity calculation on multiple knowledge graphs to obtain the similarity of multiple knowledge graphs; S12: constructing a relationship graph of multiple knowledge graphs based on the similarity of the multiple knowledge graphs and converting it into a fusion unit relationship graph.
[0009] Further, step S11 specifically includes: calculating the similarity of multiple knowledge graphs to obtain the similarity of the multiple knowledge graphs, as shown in the formula: , in, Represents a set and of Similarity.
[0010] Further, step S2 specifically includes: performing coloring modeling and parallel optimization based on the fusion unit relationship graph to obtain optimized parallel units, as shown in the formula: , , , , , in, It is a collection of fusion units, containing all fusion units; It is the number of fusion units; the first Each fusion unit Each graph embedding and fusion operation corresponds to an independent operation, which involves updating the knowledge graph embedding vector. Adjacency relationships describe the adjacency relationships between fusion units; Indicates fusion unit and They are adjacent to each other, meaning they cannot be executed in parallel; For color functions, Indicates the total number of colors; This indicates minimizing the number of colors used. , Represents the constraints that define the optimization objective.
[0011] Furthermore, the cost of sequential execution is calculated as follows: , , in, Represents two fusion units and The cost between them; and Representing fusion units Knowledge graph within Embedded representation and fusion unit Internal knowledge graph Embedded representation; It is cosine similarity, used to calculate the similarity between graph embeddings; Represents two parallel units and The costs of successive execution between them; This indicates taking the maximum value.
[0012] Furthermore, the Hamiltonian path model described above is as follows: , , , , , , in, For vertex set, it represents a set composed of multiple parallel units; Indicates the first Parallel units; This refers to the relationship between parallel units, specifically the execution order relationship between them, that is, the successive execution relationship between parallel units. Indicates from parallel unit arrive The execution order, if parallel units and If there is an execution order relationship between them, add an edge to the graph. ; Representing an edge Whether it is selected; if selected, the value is 1, otherwise it is 0.
[0013] Furthermore, the above optimization objective function is as follows: , in, The overall loss for knowledge graph embedding fusion is the optimization objective function; Then it is the first The loss generated by each parallel unit, where each parallel unit contains different fusion units; A dual knowledge graph embedding and fusion method; For parameter regularization terms.
[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization.
[0015] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization.
[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization.
[0017] Implementing the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization provided in this invention has the following beneficial effects: This invention constructs a relationship graph between multiple knowledge graphs during the embedding and fusion process, preserving key information between each graph. Based on this key information, the multi-knowledge graph fusion problem is decomposed into multiple pairwise graph fusion units. Considering the connections and constraints between fusion units, a parallel optimization strategy is proposed. Through graph coloring modeling and algorithm design, the optimal parallel execution strategy for each fusion unit is found. Parallel fusion units are treated as parallel units, and the interrelationships between them are analyzed. The successive execution costs between different parallel units are calculated to ensure that the execution cost of the entire fusion process is minimized. Based on the above optimization strategy, the optimal solution for multi-knowledge graph embedding and fusion is achieved, improving the overall efficiency of multi-knowledge graph embedding and fusion while maintaining a certain embedding quality. Attached Figure Description
[0018] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a flowchart of the multi-knowledge graph embedding and fusion method based on staining modeling and path optimization provided by the present invention; Figure 2 This is a schematic diagram of the relationship graph construction and transformation provided by the present invention; Figure 3 This is a structural block diagram of the computer device provided by the present invention. Detailed Implementation
[0019] To provide a clearer understanding of the technical features, objectives, and effects of the present invention, specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0020] Figure 1A schematic diagram of the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization in this embodiment is shown. In this embodiment, the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization includes the following steps: S1: Based on the similarity of multiple knowledge graphs, the relationship graphs of multiple knowledge graphs are transformed to obtain the fusion unit relationship graph; In one exemplary embodiment, step S1 specifically includes: S11: Calculate the similarity of multiple knowledge graphs to obtain the similarity of the multiple knowledge graphs; In one exemplary embodiment, step S11 specifically includes: calculating the similarity of multiple knowledge graphs to obtain the similarity of the multiple knowledge graphs, as shown in the formula: , in, Represents a set and of Similarity; S12: Based on the similarity of the multiple knowledge graphs, construct a relationship graph of the multiple knowledge graphs and transform it into a fusion unit relationship graph; S2: Based on the fusion unit relationship diagram, perform coloring modeling and parallel optimization to obtain optimized parallel units; In one exemplary embodiment, step S2 specifically includes: performing coloring modeling and parallel optimization based on the fusion unit relationship graph to obtain optimized parallel units, as shown in the formula: , , , , , in, It is a set of fusion units, that is, optimized parallel units, which include all fusion units; It is the number of fusion units; the first Each fusion unit Each graph embedding and fusion operation corresponds to an independent operation, which involves updating the knowledge graph embedding vector. Adjacency relationships describe the adjacency relationships between fusion units; Indicates fusion unit and They are adjacent to each other, meaning they cannot be executed in parallel; For color functions, Indicates the total number of colors; This indicates minimizing the number of colors used. , S3: Based on the optimized parallel units, calculate the successive execution cost to obtain the weighted relationship graph between the parallel units; In one exemplary embodiment, the sequential execution cost calculation is performed as follows: , , in, Represents two fusion units and The cost between them; and Representing fusion units Knowledge graph within Embedded representation and fusion unit Internal knowledge graph Embedded representation; It is cosine similarity, used to calculate the similarity between graph embeddings; Represents two parallel units and The costs of successive execution between them; This indicates taking the maximum value; S4: Based on the weighted relationship graph between the parallel units, Hamiltonian path modeling is performed to obtain the Hamiltonian path model; In one exemplary embodiment, the Hamiltonian path model is as follows: , , , , , , in, For vertex set, it represents a set composed of multiple parallel units; Indicates the first Parallel units; This refers to the relationship between parallel units, specifically the execution order relationship between them, that is, the successive execution relationship between parallel units. Indicates from parallel unit arrive The execution order, if parallel units and If there is an execution order relationship between them, add an edge to the graph. ; Representing an edge Whether it has been selected; if selected, the value is 1, otherwise it is 0. S5: Based on the Hamiltonian path model, the optimal execution path is obtained using the optimization objective function; In one exemplary embodiment, the optimization objective function is as follows:
[0021] in, The overall loss for knowledge graph embedding fusion is the optimization objective function; Then it is the first The loss generated by each parallel unit, where each parallel unit contains different fusion units; A dual knowledge graph embedding and fusion method; For parameter regularization terms.
[0022] In some embodiments, the above-described multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization can also be implemented in the following ways.
[0023] knowledge graph It is a set of fact triples, in which and These are entities, relations, and triples, where the set of triples is... Knowledge graphs store real-world information in the form of triples <entity, relation, entity>, describing the inherent relationship between two entities. Knowledge graph embedding aims to embed each entity... and each relationship Represented as A continuous vector. The knowledge graph embedding model defines a scoring function. Each triple is scored using its embedding vector. The embedding vector space for each knowledge graph is independent.
[0024] Given multiple knowledge graphs and its embedding vector Pre-aligned entity pairs ,in This indicates equivalence. The aforementioned multiple knowledge graphs, their embedding vectors, and aligned entity information serve as inputs to the fusion framework. The goal of this embodiment is to obtain the embedding vector representation of a new knowledge graph integrating multiple knowledge graphs, that is, to fuse the embedding vectors of multiple knowledge graphs located in different vector spaces into a unified vector space.
[0025] This embodiment provides a formal description of the method for embedding and fusing two knowledge graphs for use in subsequent work, and expresses the dual knowledge graph embedding and fusion method using formal notation:
[0026]
[0027] in It is an embedding fusion method that maps the embedding vectors of aligned entities to the fusion space. These are two independent embedding vectors for the two knowledge graphs. For the final fusion embedding.
[0028] In this embodiment, the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization includes the following steps: 1. Relationship diagram construction and transformation: In practical knowledge graph applications, relying solely on a single-source knowledge graph may not provide sufficient information, and typically, an entity is associated with multiple different knowledge graphs. Therefore, constructing a multi-knowledge graph relationship graph and effectively representing the relationships between graphs is crucial. This embodiment treats the relationship graph as an undirected graph, where vertices represent knowledge graphs, and edges between vertices indicate whether there are connections between graphs. This embodiment uses Jaccard similarity to quantify and determine the degree of connection between graphs. Jaccard similarity is a method for evaluating the similarity between two sets, and its calculation formula is as follows:
[0029] in, and These are two sets, typically sets of entities or relations in a knowledge graph. The intersection and union of the two knowledge graphs are calculated by examining their entities or relations. Similarity can quantify the similarity between knowledge graphs. By calculating the similarity between different knowledge graphs and labeling related knowledge graphs based on the results, a relationship graph can be constructed that can intuitively represent the connections between multiple knowledge graphs and identify candidate graphs with strong similarity.
[0030] In practical knowledge graph applications, relying solely on a knowledge graph from a single source may not provide sufficient information. For multi-knowledge graph relationship graphs, this embodiment considers graph transformation to facilitate subsequent graph embedding and fusion. For example... Figure 2As shown, in the relational graph, two adjacent knowledge graphs with connected edges are considered a fusion unit. Each fusion unit consists of two adjacent knowledge graphs and their edges. The edges between knowledge graphs indicate that there is some kind of connection or shared entity between them. Through the embedding fusion method, each pair of fusion units with connected edges is transformed into a new vertex. The new vertex represents a pre-merged representation of the original two knowledge graphs and contains the embedding vectors of the original graphs. At this point, the new vertex is regarded as a completely new form of the two knowledge graphs and the edges between them. After the units are transformed into new vertices, if there are identical knowledge graphs between two units, then there are edges between them. The edges reflect the commonalities or shared information between the original knowledge graphs, thus understanding the relationships and similarities between different graphs.
[0031] 2. Fusion unit coloring modeling: The parallelization problem of the fusion unit is transformed into the vertex coloring problem of an undirected graph. This model uses several key parameters to describe the graph embedding fusion operation and its constraints, specifically including the node set. Edge set and color assignment function .in, : Represents the set of fusion units, used to describe the number of graph embedding fusion operations. : Indicates the adjacency relationship between fusion units. : is a color function This is used to assign a color to each fusion unit, where It refers to the number of colors.
[0032] The model is given the following known conditions: fusion unit set :gather It includes all fusion units and is defined as follows: ,in This refers to the number of fusion units. Each fusion unit... Each corresponding independent graph embedding and fusion operation involves updating the knowledge graph embedding vector. Adjacency relationships... :gather The adjacency relationships between the fusion units are described. Specifically, This represents a set of unordered fusion unit pairs. Indicates fusion unit and They are adjacent, meaning they cannot be executed in parallel. Constraints between adjacent fusion units determine the execution of operations, ensuring that dependencies in the embedding fusion process are satisfied. Color function :function Assign a color to each fusion unit, and map it as ,in This represents the total number of colors. The goal of the color function is to ensure that adjacent fusion units are assigned different colors, i.e. For all .
[0033] The most important constraint in the model is that adjacent fusion units cannot be executed in parallel. This constraint is shown in the following formula:
[0034] This constraint indicates that for any two adjacent fusion units... and These elements must be assigned different colors, meaning they cannot be executed simultaneously. This constraint ensures correct operation during the embedding and fusion process, avoiding conflicts between operations.
[0035] The goal of this model is to minimize the amount of color used. And ensure that adjacent fusion units are not assigned the same color. The specific optimization objectives are as follows:
[0036]
[0037] 3. Calculation of costs for successive executions: In the process of embedding and fusing multiple knowledge graphs, it is necessary to consider how to optimize the execution order of each operation. In particular, when it comes to the fusion of multiple knowledge graphs, it is crucial to arrange the execution path reasonably to reduce the overall computational overhead and cost.
[0038] First, we define the sequential execution cost between parallel units. Each parallel unit is responsible for handling a portion of the knowledge graph embedding and fusion task, and the fusion units within these parallel units may have interdependencies or data-sharing relationships. Therefore, their execution order has a significant impact on the final computational cost and efficiency. This embodiment uses Wasserstein distance. To measure the cost between different fusion units. It is a commonly used distribution metric that effectively measures the distance between two probability distributions. In graph embedding, the entity embeddings of the graph are treated as probability distributions, and the distance between graph embeddings is calculated. This allows us to determine the cost between the two fusion units.
[0039] Specifically, we define the cost between the two fusion units i and j as follows:
[0040] in, and Representing fusion units Knowledge graph within Embedded representation and fusion unit Internal knowledge graph Embedded representation, This is cosine similarity, used to calculate the similarity between graph embeddings. In this way, we can obtain the similarity between graph embeddings and thus derive the corresponding cost.
[0041] For two parallel units and The execution cost between successive units is determined by taking the maximum cost among the fusion units, as shown below:
[0042] here and They represent parallel units, It is a parallel unit and The cost of successive implementation and They are fusion units Internal map and fusion unit Internal map .
[0043] This calculation method yields a weighted graph of relationships between parallel units. Nodes in the weighted graph represent parallel units, and edges represent connections between them. The weight of an edge represents the cost of sequential execution between two parallel units; a smaller edge weight indicates a lower cost for sequential execution between these two parallel units.
[0044] 4. Parallel unit Hamiltonian path modeling: Having determined the relationships between parallel units, the next challenge is selecting the optimal execution path from among them. Since the costs differ between parallel units, choosing a suitable execution order can reduce the computational overhead of embedding fusion. This embodiment models the execution order problem between parallel units as a graph path problem, specifically a Hamiltonian path problem. In this model, each node in the graph represents a parallel unit, and edges indicate connectivity between two parallel units, with edge weights representing the cost of executing them sequentially. The goal is to minimize the sum of costs among all parallel units by selecting an optimal path.
[0045] The Hamiltonian path problem aims to find a path that includes all vertices and minimizes the sum of the weights of its edges. Specifically, the problem is mathematically modeled as follows: The parameters are defined as follows: : Indicates the number of parallel units composed of multiple parallel units. E: Indicates the relationship between parallel units, specifically the execution order relationship between them. If parallel units and If there is an execution order relationship between them, add an edge to the graph. . : indicates from parallel unit Execution to parallel unit The cost reflects the successive execution cost between two parallel units.
[0046] Given the following conditions: Vertex set ,in Indicates the first Parallel units. Edge set. This indicates the sequential execution relationship between parallel units, where Indicates from parallel unit arrive The execution order. Symmetric cost: Assuming from arrive The cost and from arrive The cost Same, that is This indicates that the execution cost is symmetrical and does not depend on the specific direction of execution.
[0047] The model constraints are: unique selection constraint: for each parallel unit An edge must be chosen to connect to it. or And each parallel unit can only appear once. For example, on a path, a vertex... There must be exactly one edge selected to form a valid path. No repeated selection constraint: For each distinct pair of parallel units... and You cannot select two edges at the same time. and To avoid creating invalid loops or duplicate paths.
[0048] By modeling the execution order problem of parallel units, the goal is to find a minimum-cost Hamiltonian path. This path contains all parallel units, and each parallel unit appears only once on the path, thus guaranteeing that all operations are executed and the total execution cost is minimized. The objective function is to minimize the sum of the costs of every edge in the path, as shown below:
[0049] in Indicates from arrive The execution cost. Representing an edge Whether it is selected; if selected, the value is 1, otherwise it is 0.
[0050] The constraint is for each vertex There must be one and only one edge selected, which guarantees that each parallel unit is visited only once, as shown below:
[0051] Since the path is unidirectional, we do not allow selecting edges simultaneously. and Therefore, another constraint is shown in the following formula:
[0052] Through the above modeling, the problem of choosing the execution order of parallel units is transformed into a Hamiltonian path problem, and further research is conducted on how to efficiently solve this optimization problem.
[0053] 5. Optimization Goal: After transforming the execution order selection problem of parallel units into a Hamiltonian path problem, the optimal execution path is obtained through the optimal solution obtained by the optimization algorithm. Based on the pairwise knowledge graph embedding and fusion method, the parallel results of fusion units, and the execution paths of parallel units, this embodiment gradually decomposes the problem of embedding and fusion of multiple knowledge graphs, and based on the dual knowledge graph embedding and fusion method... And it performs corresponding actions according to the optimal fusion path. The overall training objective is:
[0054] in The overall loss for embedding and fusing multiple knowledge graphs. Then it is the first The loss generated by each parallel unit, where each parallel unit contains different fusion units.
[0055] This embodiment provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the aforementioned multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium may also include combinations of the above types of memory.
[0056] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps of the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization described above.
[0057] like Figure 3 As shown, the computer device 120 may include: at least one processor 121, such as a central processing unit (CPU), at least one communication interface 123, memory 124, and at least one communication bus 122. The communication bus 122 is used to enable communication between these components. The communication interface 123 may include a display screen and a keyboard; optionally, the communication interface 123 may also include a standard wired interface or a wireless interface. The memory 124 may be high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory 124 may also be at least one storage device located remotely from the aforementioned processor 121. The memory 124 stores application programs, and the processor 121 calls the program code stored in the memory 124 to execute any of the aforementioned method steps. The communication bus 122 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus 122 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3The term 124 is represented by a single line, but this does not imply a single bus or a single type of bus. The memory 124 may include volatile memory, such as random-access memory (RAM); it may also include non-volatile memory, such as flash memory, hard disk drive (HDD), or solid-state drive (SSD); or it may include combinations of the above types of memory. The processor 121 may be a central processing unit (CPU), a network processor (NP), or a combination of a CPU and an NP. The processor 121 may further include hardware chips. These hardware chips may be application-specific integrated circuits (ASICs), programmable logic devices (PLDs), or combinations thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof. Optionally, the memory 124 is also used to store program instructions. The processor 121 can call the program instructions to implement the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization as described in this embodiment.
[0058] This embodiment provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization.
[0059] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.
Claims
1. A multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization, characterized in that, Includes the following steps: S1: Based on the similarity of multiple knowledge graphs, the relationship graphs of multiple knowledge graphs are transformed to obtain the fusion unit relationship graph; S2: Based on the fusion unit relationship diagram, perform coloring modeling and parallel optimization to obtain optimized parallel units; S3: Based on the optimized parallel units, perform successive execution cost calculations to obtain a weighted relationship graph between the parallel units; S4: Based on the weighted relationship graph between the parallel units, Hamiltonian path modeling is performed to obtain the Hamiltonian path model; S5: Based on the Hamiltonian path model, the optimal execution path is obtained using the optimization objective function.
2. The multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization according to claim 1, characterized in that, Step S1 specifically includes: S11: Calculate the similarity of multiple knowledge graphs to obtain the similarity of the multiple knowledge graphs; S12: Based on the similarity of the multiple knowledge graphs, construct a relationship graph of the multiple knowledge graphs and transform it into a fusion unit relationship graph.
3. The multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization according to claim 2, characterized in that, Step S11 specifically includes: calculating the similarity of multiple knowledge graphs to obtain the similarity score, as shown in the formula: ， in, Represents a set and of Similarity.
4. The multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization according to claim 1, characterized in that, Step S2 specifically includes: performing coloring modeling and parallel optimization based on the fusion unit relationship diagram to obtain optimized parallel units, as shown in the formula: ， ， ， ， ， in, It is a collection of fusion units, containing all fusion units; It is the number of fusion units; the first Each fusion unit Each graph embedding and fusion operation corresponds to an independent operation, which involves updating the knowledge graph embedding vector. Adjacency relationships describe the adjacency relationships between fusion units; Indicates fusion unit and They are adjacent to each other, meaning they cannot be executed in parallel; For color functions, Indicates the total number of colors; This indicates minimizing the number of colors used. , Represents the constraints that define the optimization objective.
5. The multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization according to claim 1, characterized in that, The calculation of the successive execution cost is as follows: ， ， in, Represents two fusion units and The cost between them; and Representing fusion units Knowledge graph within Embedded representation and fusion unit Internal knowledge graph Embedded representation; It is cosine similarity, used to calculate the similarity between graph embeddings; Represents two parallel units and The costs of successive execution between them; This indicates taking the maximum value.
6. The multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization according to claim 1, characterized in that, The Hamiltonian path model is as follows: ， ， ， ， ， ， in, For vertex set, it represents a set composed of multiple parallel units; Indicates the first Parallel units; This refers to the relationship between parallel units, specifically the execution order relationship between them, that is, the successive execution relationship between parallel units. Indicates from parallel unit arrive The execution order, if parallel units and If there is an execution order relationship between them, add an edge to the graph. ; Representing an edge Whether it is selected; if selected, the value is 1, otherwise it is 0.
7. The multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization according to claim 1, characterized in that, The optimization objective function is as follows: ， in, The overall loss for knowledge graph embedding fusion is the optimization objective function; Then it is the first The loss generated by each parallel unit, where each parallel unit contains different fusion units; A dual knowledge graph embedding and fusion method; For parameter regularization terms.
8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the steps of the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization as described in any one of claims 1-7.
9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization as described in any one of claims 1-7.
10. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the multi-knowledge graph embedding and fusion method based on coloring modeling and path optimization as described in any one of claims 1-7.