Cloud service robot knowledge graph completion method and system based on hierarchical modeling

Abstract

This invention belongs to the field of knowledge completion learning technology for service robots, and provides a method and system for knowledge graph completion of cloud service robots based on hierarchical modeling. The method includes: acquiring knowledge graph triple information of the cloud service robot in a corresponding scenario; representing the knowledge graph triple information as two levels of information: semantic information and structural information; mapping the semantic information to a semantic measurement space and the structural information to a structural measurement space; extracting knowledge graph semantic features from the semantic measurement space and knowledge graph structural features from the structural measurement space; fusing the knowledge graph semantic features and knowledge graph structural features, and using a knowledge graph link prediction scoring network to process the fused features and predict the knowledge graph link score probability; and selecting the knowledge graph link with the highest score probability to output, thereby completing the knowledge graph.

Description

Technical Field

[0001] This invention belongs to the field of knowledge completion and learning technology for service robots, and particularly relates to a method and system for completing the knowledge graph of cloud service robots based on hierarchical modeling. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Service robots are widely used in various real-world scenarios such as greeting guests and contactless disinfection and delivery. In these scenarios, representation learning technology based on semantic information plays an indispensable role in human-computer interaction. The ability of service robots to answer natural questions posed by service recipients is strong evidence of their cognitive capabilities. Knowledge graphs (KGs) have become an important component of artificial intelligence (AI) and are applied in many downstream applications, such as visual question answering and recommendation systems. In complex scene interactions, the amount of information contained in knowledge graphs is considerable. Through information extraction, robots can even extract key information from each entity and relation node in the knowledge graph and provide answers and corresponding services. Combining prior information in the knowledge graph allows for better prediction of user needs. By developing interactive AI, this has greatly promoted the development of online education, contextual analysis, and video content retrieval. However, it is difficult to guarantee the completeness and accuracy of knowledge graphs during their construction.

[0004] During continuous human-computer interaction with service robots, the robots accumulate a large number of subtle deviations due to their constant semantic and visual interactions with the outside world and their semantic collection and entity localization of each node in the knowledge graph. When these deviations reach a certain saturation point, the effectiveness of responding to user questions will significantly decrease. Furthermore, because service robots operate in complex environments, such as hospitals where a large amount of semantic information is received simultaneously, it is difficult for the robots to accurately clean up noise and focus on the main information. This results in the question-and-answer results relying excessively on prior knowledge, insufficient visual information capture and discrimination in actual scenarios, poor service performance robustness, and limited scenario considerations. Traditional single optimization algorithms also suffer from poor stability and are difficult to widely use.

[0005] Knowledge graph completion prediction aims to provide accurate reasoning and prediction answers for complex human-computer interaction scenarios and related problems, requiring a high level of understanding of both visual and textual information. During the research and development process, the inventors discovered that existing knowledge graph completion algorithms suffer from poor stability, low prediction accuracy, and high costs, making it difficult to apply theoretical analysis to practical evaluation. Summary of the Invention

[0006] To address the technical problems mentioned above, this invention provides a method and system for completing a knowledge graph for cloud service robots based on hierarchical modeling. The completed knowledge graph can improve the accuracy of human-computer interaction question answering in actual service robots, understand different aspects and granularities of videos, and solve reasoning problems from fine-grained to coarse-grained in the time and space domains, thus improving the user experience of service robots.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The first aspect of this invention provides a method for completing a knowledge graph for cloud service robots based on hierarchical modeling.

[0009] A knowledge graph completion method for cloud service robots based on hierarchical modeling, comprising:

[0010] Obtain knowledge graph triple information of cloud service robots in corresponding scenarios;

[0011] The knowledge graph triple information is represented by two levels of information: semantic information and structural information.

[0012] The semantic information is mapped to the semantic measurement space, and the structural information is mapped to the structural measurement space;

[0013] Extract semantic features of knowledge graphs from the semantic measurement space, and extract structural features of knowledge graphs from the structural measurement space;

[0014] The semantic and structural features of the knowledge graph are integrated, and the integrated features are processed by the knowledge graph link prediction score network to predict the probability of knowledge graph link scores.

[0015] The knowledge graph links with the highest scoring probabilities are selected and output to complete the knowledge graph.

[0016] A second aspect of the present invention provides a knowledge graph completion system for cloud service robots based on hierarchical modeling.

[0017] A knowledge graph completion system for cloud service robots based on hierarchical modeling, comprising:

[0018] The knowledge graph information acquisition module is used to acquire knowledge graph triple information of cloud service robots in corresponding scenarios;

[0019] The hierarchical information representation module is used to represent knowledge graph triple information in two hierarchical information formats: semantic information and structural information.

[0020] An information mapping module is used to map the semantic information to a semantic measurement space and the structural information to a structural measurement space.

[0021] The feature extraction module is used to extract semantic features of the knowledge graph from the semantic measurement space and structural features of the knowledge graph from the structural measurement space.

[0022] The link score prediction module is used to fuse semantic features and structural features of the knowledge graph, and uses the knowledge graph link prediction score network to process the fused features and predict the link score probability of the knowledge graph.

[0023] The knowledge graph completion module is used to filter out the knowledge graph links with the highest scoring probability to complete the knowledge graph.

[0024] A third aspect of the present invention provides a computer-readable storage medium.

[0025] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the hierarchical modeling-based cloud service robot knowledge graph completion method described above.

[0026] A fourth aspect of the present invention provides a computer device.

[0027] A computer device includes 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 in the hierarchical modeling-based cloud service robot knowledge graph completion method described above.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] (1) In order to improve the accuracy and robustness of the service robot knowledge graph completion system, this invention proposes a cloud service robot knowledge graph completion method and system based on hierarchical modeling. By using a dual-space modeling strategy and a hierarchical mechanism, a high-quality service robot knowledge graph completion system is constructed. The completed knowledge graph can improve the accuracy of human-computer interaction question answering in actual service robots, understand the different aspects and granularities contained in the input information of service objects, and solve reasoning problems from fine-grained to coarse-grained in the time and space domains, thus making the service robot experience better.

[0030] (2) This invention effectively introduces an adjustment coefficient to balance and fuse the semantic and distance measurement feature information of the lower and higher layers, optimize the proportion of scoring elements, make full use of reference sample information, strengthen entity embedding, eliminate semantic bias, improve the accuracy of the completed knowledge graph, and ultimately improve the accuracy of human-computer interaction question answering on the actual server.

[0031] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0032] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0033] Figure 1 This is a structural diagram of the distance measurement space and semantic measurement space of the cloud service robot in Embodiment 1 of the present invention;

[0034] Figure 2 This is the architecture diagram of the cloud service robot knowledge graph completion model based on hierarchical modeling in Embodiment 1 of the present invention;

[0035] Figure 3 is a line graph of the model prediction results at different levels in Embodiment 1 of the present invention;

[0036] Figure 4 This is a statistical chart of the model prediction results after introducing adjustment coefficients in Embodiment 1 of the present invention.

[0037] Figure 5 This is an overall flowchart of the knowledge graph completion method for cloud service robots based on hierarchical modeling in Embodiment 1 of the present invention. Detailed Implementation

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0039] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0040] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0041] When service robots use existing knowledge graphs, they need to infer and complete the prior information in the knowledge graph before iterative training. Knowledge graph completion refers to identifying and filling in missing entities and relationships in the knowledge graph. This requires utilizing a large amount of additional data, such as text, web pages, and images, to identify missing entities and relationships. Most service robots have low utilization rates of prior graph information. During human-computer dialogue, service robots simply collect semantic information of objects, failing to utilize the vast amount of important information in the knowledge graph to provide accurate answers in complex scenarios. This results in poor human-computer interaction capabilities and large errors in question-and-answer results for most service robots, limiting their application to specific semantic scenarios and significantly restricting their development. Currently, the most popular knowledge graph completion method is link prediction. Link prediction methods identify the attribute information of missing entities and match them with existing entities, allowing service robots to obtain potentially richer and higher-quality visual and textual information with a temporal dimension. By predicting the links between nodes in a knowledge graph, service robots can predict a series of potential intentions of service recipients and quickly provide corresponding appearance and motion information services, reducing service response time and computational costs. This is currently a key challenge in semantic prediction technology for service robots. To improve the model's generalization ability to handle complex structured data in environments such as homes and hospitals, and to enhance the service robot's ability to accurately respond to user questions in complex scenarios by combining prior knowledge graph knowledge, this invention proposes a missing semantic information prediction function based on the semantic information of prior knowledge graphs. Simultaneously, it utilizes the node structure and semantic information contained in the knowledge graph to obtain prior data (speech, video, etc. descriptions), scene information, entities, and spatial relationships in human-computer dialogue modalities, thereby better improving the robot's service targeting and humanization, making the robot's question-and-answer service function more intelligent, and better understanding visual information intentions.

[0042] Example 1

[0043] This embodiment illustrates the application of the method to a server. It is understood that the method can also be applied to terminals, and can be applied to systems including terminals, servers, and other components, and implemented through interaction between the terminal and the server. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communication, middleware services, domain name services, CDN security services, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, which is not limited herein.

[0044] The following is based on Figure 2 and Figure 5 This paper details the specific implementation process of a knowledge graph completion method for cloud service robots based on hierarchical modeling in this embodiment.

[0045] In practical implementation, a knowledge graph completion method for cloud service robots based on hierarchical modeling includes:

[0046] Step 1: Obtain the knowledge graph triple information of the cloud service robot in the corresponding scenario. For example, each triple of the knowledge graph is (h, r, t). h, r, and t represent the head entity, the relationship between the head entity and the tail entity, and the tail entity, respectively.

[0047] Step 2: Represent the knowledge graph triple information in terms of semantic information and structural information.

[0048] This embodiment models the triples as distance measurement and semantic measurement spaces. The basic distance measurement space and semantic measurement space structures are as follows: Figure 1 As shown. For each entity or relation, we consider it to contain two parts of information. First, the structural information of entity distance measurements in the distance space under a specific relation. Second, a method for identifying semantic information in the semantic space. Specifically, this invention preserves the structural information of the origin triples as auxiliary information when measuring the distance space, which can be used to determine a reliable long-path triple when dealing with complex relation properties.

[0049] Step 3: Map the semantic information to the semantic measurement space, and map the structural information to the structural measurement space.

[0050] Step 4: Extract the semantic features of the knowledge graph from the semantic measurement space, and extract the structural features of the knowledge graph from the structural measurement space.

[0051] In step 4, the first transformation matrix is ​​used to extract the semantic features of the knowledge graph from the semantic measurement space. The second transformation matrix is ​​used to extract the structural features of the knowledge graph from the structural measurement space.

[0052] Specifically, first, entities and relations are randomly initialized as k-dimensional representations and then equally divided into two parts. The head entity h is divided into h0 and h1, where h0 consists of structural information and h1 consists of semantic information. Second, h0 and h1 are projected into the distance measurement space and semantic measurement space, respectively. In this distance space, we obtain the corresponding matrices... and The element-order projection measurement space in distance is defined as follows:

[0053]

[0054] Where h0, t0, and r0 represent the first segment of h, r, and t, respectively. It is a matrix that projects h0 onto the structural distance measurement space. and This indicates the same operation on t0 and r0; in the first paragraph, and The superscript s is 1, where s indicates that there are s segments.

[0055] Using flexible transformation matrices This captures the structural features between entities and relationships, rather than using rotation or translation operations for knowledge graph representation. For example... Figure 1 As shown, this invention proposes a soft transformation matrix, represented by a solid line, to flexibly capture more transformation information between entities and relationships to handle complex relationships. This is achieved through transformation operators. Store all possible transformation and move operation parameters:

[0056]

[0057] Where M0 is used to generate the transformation operation matrix, and the dimension of M0 is 0.5k*1, therefore, the distance scoring function at the basic level of the knowledge graph completion model... as follows:

[0058]

[0059] in, This represents the L1 / L2 norm.

[0060] After completing the structural distance space modeling, this embodiment performs semantic metric space modeling. In this semantic space, the invention uses sub-semantics to project entities and relationships in a fine-grained manner. The projection in this space is defined as follows:

[0061]

[0062] Where h1, t1, and r1 represent the second segment of h, r, and t, respectively. It is a semantic extractor that projects h1 into a fine-grained sub-semantic metric space. and This indicates the same operation on t1 and r1.

[0063] In order to be in Figure 4 middle, It is the predicted tail entity. It is an ideal relational embedding, the goal of which is determined by the following equation: The projection combination representation in the constructed semantic space, where This represents a mixed semantic information representation, including type, concept, and other semantic information. Formula (5) is used to obtain the semantic score of the triples.

[0064]

[0065] in, This represents the L2 norm.

[0066] Subsequently, based on the projections of the two spaces mentioned above, the joint scoring function at the basic structural level is defined as follows:

[0067]

[0068] Here, α is a learnable weight parameter used to utilize distance scores in the base level space. and semantic score

[0069] The semantic matrix and the structure extraction matrix are defined as follows:

[0070]

[0071] Among them, M pe M represents a structured extraction matrix from which latent information can be obtained from shallow layers. se This is the semantic extraction matrix. More in-depth information can be obtained through this matrix.

[0072] Furthermore, by performing a similar operation on new embeddings in the deep representation space, the distance function can be defined as:

[0073]

[0074] Among them, This represents the element-level projection in the distance measurement space. This is the transformation operation parameter matrix.

[0075] Similar to formula (3), the corresponding scoring measurement functions can be obtained in different hierarchical spaces as follows:

[0076]

[0077] The semantic functions at different deeper levels are defined as follows:

[0078]

[0079] Step 5: Integrate the semantic features and structural features of the knowledge graph, and use the knowledge graph link prediction score network to process the integrated features and predict the probability of the knowledge graph link score.

[0080] By introducing an adjustment coefficient, the semantic and distance metric feature information from the lower and higher levels are fused in a balanced manner, and the scoring function in the deeper layers is defined as follows:

[0081]

[0082] From this point on, the final scoring function f of the cloud service robot knowledge graph completion model in Example 1 is... r (h,t) is defined as follows:

[0083]

[0084] Where λ is in constraint 1 = ∑λ k The weighting of the embedded scores at the next level.

[0085] Step 6: Filter out the knowledge graph links with the highest scoring probability to complete the knowledge graph.

[0086] The knowledge graph completion prediction aims to provide correct reasoning and prediction answers for complex human-computer interaction scenarios and related questions. This embodiment proposes a function to predict missing semantic information based on the semantic information of the prior knowledge graph, and at the same time utilizes the node structure information and semantic information contained in the knowledge graph to obtain prior data in the human-computer dialogue modality.

[0087] The training process of a model generates a large amount of raw data, which contains a lot of missing data and noise, seriously affecting the quality of the data and causing some difficulties in mining effective information. Applying some methods, such as data segmentation, can improve the quality of the data.

[0088] Experiments were conducted on the publicly available datasets WN18, WN18RR, FB15k-237, and YAGO3-10 to evaluate the robustness of the knowledge graph completion model for cloud service robots. The model was evaluated based on computational accuracy.

[0089] The hierarchical modeling-based knowledge graph completion model for cloud service robots was trained with multiple baseline models under the same pre-trained model parameters. For all implemented knowledge graph completion methods, we initialized the knowledge graph structure and semantic information embeddings through the pre-trained model. Taking the WN18 and WN18RR datasets as examples, the model prediction results at different levels are shown in Figures 3(a) and 3(b). As the level increases, the batch size and embedding dimension decrease accordingly, while the initial weights of the levels are equal, thus realizing four hierarchical knowledge graph completion models at different levels.

[0090] An adjustment coefficient is introduced to achieve a balanced fusion of semantic and distance metric feature information from the lower and higher levels. Taking the WN18RR dataset as an example, for instance... Figure 4 As shown, the final prediction result of the prediction model can be changed by adjusting the value of the coefficient λ.

[0091] The model was trained using the aforementioned publicly available dataset, and then tuned and evaluated using the relationships found in the validation and test data. Experiments compared the scores of several existing models on the same dataset, with evaluation metrics based on prediction accuracy, as follows: mean rank (MR), mean reciprocal rank (MRR), and Hits@k (k = 1, 3, and 10).

[0092] Taking the WN18 and YAGO3-10 datasets as examples, the link prediction scores of the baseline model and this model are compared on these datasets. The performance of all models on these datasets is shown in Tables 1 and 2 below:

[0093] Table 1. Comparison of prediction scores for each model on the WN18 dataset.

[0094]

[0095] Table 2 shows the comparison of prediction scores for each model on the YAGO3-10 dataset.

[0096]

[0097]

[0098] The experimental results are shown in the table above. It is clear that compared with traditional robot knowledge graph completion methods, this model achieves better performance on both datasets. The cloud service robot knowledge graph completion system based on hierarchical modeling is more suitable for solving the human-computer interaction prediction problem of service robots, and provides accurate answers by combining structural and semantic information.

[0099] Example 2

[0100] This embodiment implements a cloud service robot knowledge graph completion system based on hierarchical modeling, which includes:

[0101] (1) Knowledge graph information acquisition module, which is used to acquire knowledge graph triple information of cloud service robots in corresponding scenarios;

[0102] (2) Hierarchical information representation module, which is used to represent knowledge graph triple information in two hierarchical information: semantic information and structural information;

[0103] (3) An information mapping module, which is used to map the semantic information to the semantic measurement space and the structural information to the structural measurement space;

[0104] (4) Feature extraction module, which is used to extract knowledge graph semantic features from the semantic measurement space and knowledge graph structural features from the structural measurement space;

[0105] (5) Link score prediction module, which is used to fuse the semantic features and structural features of the knowledge graph, and use the knowledge graph link prediction score network to process the fused features and predict the link score probability of the knowledge graph.

[0106] (6) Knowledge graph completion module, which is used to filter out the knowledge graph links with the highest score probability to complete the knowledge graph.

[0107] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0108] Example 3

[0109] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the hierarchical modeling-based cloud service robot knowledge graph completion method described above.

[0110] Example 4

[0111] 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 in the hierarchical modeling-based cloud service robot knowledge graph completion method described above.

[0112] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0113] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0114] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0115] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for completing a cloud service robot knowledge graph based on hierarchical modeling, characterized in that, include: Obtain knowledge graph triple information of cloud service robots in corresponding scenarios; The knowledge graph triple information is represented by two levels of information: semantic information and structural information. The semantic information is mapped to the semantic measurement space, and the structural information is mapped to the structural measurement space; Extract semantic features of knowledge graphs from the semantic measurement space, and extract structural features of knowledge graphs from the structural measurement space; This method integrates semantic and structural features of a knowledge graph, and uses a knowledge graph link prediction scoring network to process the integrated features. The network then predicts the probability of a knowledge graph link score based on the final scoring function. Specifically, in the knowledge graph link prediction scoring network, an adjustment coefficient is introduced to balance and fuse semantic and distance metric feature information from the bottom and high-level layers. The scoring function at deeper levels is the sum of the distance metric and semantic features at different depths, along with their corresponding learning weights. The final scoring function is the sum of the embedding scores and their weights at each level. The knowledge graph links with the highest scoring probabilities are selected and output to complete the knowledge graph. 2.The cloud service robot knowledge graph completion method based on hierarchical modeling according to claim 1, wherein, The first transformation matrix is ​​used to extract the semantic features of the knowledge graph from the semantic measurement space. 3.The cloud service robot knowledge graph completion method based on hierarchical modeling according to claim 1 or 2, wherein, The knowledge graph structural features are extracted from the structural measurement space using the second transformation matrix.

4. A cloud service robot knowledge graph completion system based on hierarchical modeling, characterized in that, include: The knowledge graph information acquisition module is used to acquire knowledge graph triple information of cloud service robots in corresponding scenarios; The hierarchical information representation module is used to represent knowledge graph triple information in two hierarchical information formats: semantic information and structural information. An information mapping module is used to map the semantic information to a semantic measurement space and the structural information to a structural measurement space. The feature extraction module is used to extract semantic features of the knowledge graph from the semantic measurement space and structural features of the knowledge graph from the structural measurement space. The link score prediction module is used to fuse semantic features and structural features of the knowledge graph, and to process the fused features using a knowledge graph link prediction scoring network. Based on the final scoring function, it predicts the probability of the knowledge graph link score. Specifically, in the knowledge graph link prediction scoring network, an adjustment coefficient is introduced to balance the fusion of semantic and distance metric features from the bottom and high levels. The scoring function in the deeper layers is the sum of the distance metrics and semantics at different depths, along with their corresponding learning weights. The final scoring function is the sum of the embedding scores and their weights at each level. The knowledge graph completion module is used to filter out the knowledge graph links with the highest scoring probability to complete the knowledge graph. 5.The hierarchical modeling based cloud service robot knowledge graph completion system of claim 4, wherein, The first transformation matrix is ​​used to extract the semantic features of the knowledge graph from the semantic measurement space. 6.The hierarchical modeling based cloud service robot knowledge graph completion system of claim 4 or 5, wherein, The knowledge graph structural features are extracted from the structural measurement space using the second transformation matrix.

7. A computer readable storage medium having stored thereon a computer program, characterized in that When the program is executed by the processor, it implements the steps in the cloud service robot knowledge graph completion method based on hierarchical modeling as described in any one of claims 1-3.

8. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the cloud service robot knowledge graph completion method based on hierarchical modeling as described in any one of claims 1-3.

Images