Knowledge graph construction method, device and storage medium for cloud computing

Abstract

The embodiment of the present application provides a knowledge graph construction method, device and storage medium for cloud computing. In the embodiment of the present application, a knowledge graph relationship graph for cloud computing is pre-generated, and multiple question-type entity types, answer-type entity types and edge relationships between these entity types in the cloud computing after-sales scenario are defined in the knowledge graph relationship graph. Based on this, by mining question-type entity phrases from the original after-sales questions corresponding to the cloud computing service and the standard after-sales questions in the after-sales knowledge base, mining answer-type entity phrases from the standard solutions in the after-sales knowledge base, and adding edges between question-type entity phrases and answer-type entity phrases, a knowledge graph for cloud computing after-sales service is obtained. Afterwards, the knowledge graph is applied to the after-sales service of cloud computing services, which can solve various problems raised by users in the process of using cloud computing services online and in a timely manner, reduce the cost of human resources, and improve the after-sales quality of cloud computing services.

Description

Technical Field

[0001] The present application relates to the field of cloud computing technology, and in particular to a method, device and storage medium for constructing a knowledge graph for cloud computing. Background Art

[0002] Cloud computing is a type of distributed computing that solves the problem of computing power by distributing tasks and merging computing results. With the help of cloud computing technology, tens of thousands of data can be processed in a very short time, thus achieving powerful network services. In view of the advantages of cloud computing, more and more corporate users are beginning to use cloud computing services. The types of cloud computing services that cloud computing operators can provide are increasing, and the functions of cloud computing services are becoming more and more powerful. However, with this comes the problem of after-sales service. More and more users need to consult cloud computing service-related issues, or when problems occur during use, they also need to be solved through customer service. However, the existing manual customer service method requires a lot of human resources, while the online customer service method can reduce the cost of human resources, but the existing online customer service lacks relevant knowledge reserves and cannot well solve various problems raised by users at any time. This is a major technical problem facing the cloud computing field. Summary of the invention

[0003] Multiple aspects of the present application provide a knowledge graph construction method, device and storage medium for cloud computing, which are used to provide a knowledge graph for cloud computing after-sales service, so as to effectively solve various problems raised by users in the process of using cloud computing services.

[0004] An embodiment of the present application provides a method for constructing a knowledge graph for cloud computing, including: mining entity phrases for original after-sales questions corresponding to cloud computing services and standard after-sales questions in an after-sales knowledge base to obtain multiple question-type entity phrases; determining the question-type entity types to which the multiple question-type entity phrases belong based on multiple question-type entity types defined in a pre-generated knowledge graph relationship graph and in combination with attribute information of the multiple question-type entity phrases; selecting at least one answer-type entity phrase belonging to the answer-type entity type from standard solutions in an after-sales knowledge base according to the answer-type entity types defined in the knowledge graph relationship graph; adding edges between the multiple question-type entity phrases and between the multiple question-type entity phrases and the at least one answer-type entity phrase according to edge relationships between question-type entity types and between question-type entity types and answer-type entity types defined in the knowledge graph relationship graph to obtain a knowledge graph for cloud computing after-sales service.

[0005] An embodiment of the present application also provides a method for generating a knowledge graph relationship diagram for cloud computing, including: generating a plurality of question-type entity types based on the analysis results of after-sales problems of cloud computing services, wherein different question-type entity types are used to describe different types of after-sales problems or to describe after-sales problems from different dimensions; generating an answer-type entity type, wherein the answer-type entity type is used to describe solutions to after-sales problems; determining edge relationships between the plurality of question-type entity types and between the plurality of question-type entity types and the answer-type entity types based on the association relationships between the after-sales problems described by the plurality of question-type entity types and the demand information for solutions to the after-sales problems; generating a knowledge graph relationship diagram based on the plurality of question-type entity types, the answer-type entity types and the edge relationships, wherein the knowledge graph relationship diagram is used to generate a knowledge graph required for cloud computing after-sales services.

[0006] An embodiment of the present application also provides an electronic device, including: a memory and a processor; the memory is used for computer programs, and the processor is coupled to the memory for executing the computer program to implement the steps in the knowledge graph construction method for cloud computing provided in an embodiment of the present application.

[0007] An embodiment of the present application also provides an electronic device, comprising: a memory and a processor; the memory is used for computer programs, and the processor is coupled to the memory for executing the computer program, so as to implement the steps in the knowledge graph relationship diagram generation method for cloud computing provided in an embodiment of the present application.

[0008] An embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor is enabled to implement the steps in the knowledge graph construction method for cloud computing provided in the embodiment of the present application.

[0009] An embodiment of the present application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor is enabled to implement the steps in the knowledge graph relationship diagram generation method for cloud computing provided in an embodiment of the present application.

[0010] In the embodiment of the present application, a knowledge graph relationship diagram for cloud computing is generated in advance, and multiple question-type entity types, answer-type entity types, and edge relationships between these entity types in the cloud computing after-sales scenario are defined in the knowledge graph relationship diagram. Based on this, question-type entity phrases are mined from the original after-sales questions corresponding to the cloud computing service and the standard after-sales questions in the after-sales knowledge base, answer-type entity phrases are mined from the standard solutions in the after-sales knowledge base, and edges are added between question-type entity phrases and answer-type entity phrases to obtain a knowledge graph for cloud computing after-sales service. Afterwards, the knowledge graph is applied to the after-sales service of cloud computing services, which can solve various problems raised by users in the process of using cloud computing services online and in a timely manner, reduce the cost of human resources, and improve the after-sales quality of cloud computing services. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0012] Figure 1 A schematic diagram of a process for generating a knowledge graph relationship diagram for cloud computing provided by an exemplary embodiment of the present application;

[0013] Figure 2 An example of a knowledge graph relationship diagram for cloud computing provided for an exemplary embodiment of the present application;

[0014] Figure 3 The exemplary embodiment of the present application is provided based on Figure 2 Schematic diagram of the knowledge graph generated by the knowledge graph relationship graph shown;

[0015] Figure 4 A schematic diagram of a process for constructing a knowledge graph for cloud computing provided in an exemplary embodiment of the present application;

[0016] Figure 5 The exemplary embodiments of the present application provide Figure 4 A schematic diagram of the detailed implementation process of step 401;

[0017] Figure 6 The exemplary embodiments of the present application provide Figure 4 A schematic diagram of the detailed implementation process of step 402;

[0018] Figure 7 The exemplary embodiments of the present application provide Figure 4 A schematic diagram of the detailed implementation process of step 403;

[0019] Figure 8A schematic diagram of the structure of a knowledge graph relationship diagram generating device for cloud computing provided by an exemplary embodiment of the present application;

[0020] Fig. 9 A schematic diagram of the structure of a knowledge graph construction device for cloud computing provided by an exemplary embodiment of the present application;

[0021] Fig.10 A schematic structural diagram of an electronic device provided for an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0023] In order to solve the technical problem that various problems raised by existing users in the process of using cloud computing services cannot be solved in a timely and effective manner, in an embodiment of the present application, the concept of knowledge graph is applied to the after-sales field of cloud computing services to obtain a knowledge graph for cloud computing after-sales services. Thereafter, the knowledge graph is applied to the after-sales services of cloud computing services, which can solve various problems raised by users in the process of using cloud computing services online and in a timely manner, reduce human resource costs, and improve the after-sales quality of cloud computing services.

[0024] It should be emphasized that there is no precedent for the definition and construction of knowledge graphs for the after-sales field of cloud computing services in the cloud computing industry. In the embodiment of the present application, a new method for defining knowledge graphs for the after-sales field of cloud computing services is provided. Based on the definition, a knowledge graph relationship graph for cloud computing is pre-generated, and a variety of question-type entity types, answer-type entity types, and edge relationships between these entity types in the after-sales scenario of cloud computing services are defined through the knowledge graph relationship graph; then, various after-sales problems accumulated and precipitated in the after-sales service process of cloud computing services and some existing solutions are used as data basis for in-depth analysis and mining, and question-type entity phrases under various question-type entity types and answer-type entity phrases under answer-type entity types are obtained. Based on the edge relationship between each entity type defined in the knowledge graph relationship graph, edges are added between question-type entity phrases and answer-type entity phrases to obtain a knowledge graph for after-sales service of cloud computing. Afterwards, the knowledge graph is applied to the after-sales service of cloud computing services, which can solve various problems raised by users in the process of using cloud computing services online and in a timely manner, reduce the cost of human resources, and improve the after-sales quality of cloud computing services.

[0025] The technical solutions provided by various embodiments of the present application are described in detail below in conjunction with the accompanying drawings.

[0026] Figure 1 A schematic diagram of a method for generating a knowledge graph relationship diagram for cloud computing provided by an exemplary embodiment of the present application. Figure 1 As shown, the method includes:

[0027] 101. Generate multiple problem-type entity types based on the analysis results of after-sales problems of cloud computing services. Different problem-type entity types are used to describe different types of after-sales problems or describe after-sales problems from different dimensions.

[0028] 102. Generate an answer-type entity type, where the answer-type entity type is used to describe a solution to an after-sales problem;

[0029] 103. According to the association relationship between after-sales problems described by the various problem-type entity types and the demand information for solutions of the after-sales problems, determine the edge relationships between the various problem-type entity types and between the various problem-type entity types and the answer-type entity types;

[0030] 104. Generate a knowledge graph relationship diagram based on multiple question-type entity types, answer-type entity types and the above-mentioned edge relationships. The knowledge graph relationship diagram is used to generate the knowledge graph required for cloud computing after-sales service.

[0031] In the embodiment of the present application, after-sales problems corresponding to cloud computing services are collected. The after-sales problems here are after-sales problems in a broad sense, including various original after-sales problems raised by users in the process of using cloud computing services, and also standard after-sales problems formed after sorting and solving. Among them, one of the differences between the original after-sales problem and the standard after-sales problem is the expression of the problem. The original after-sales problem is a problem directly raised by the user, which is more casual, flexible or diversified in expression, while the standard after-sales problem has been standardized in expression. For example, an original after-sales problem may be: "My ECS server with Linux system is running full of CPU. Is there any monitoring tool available?" Correspondingly, the standard after-sales problem may be "For the ECS server with Linux system, is there a monitoring tool for monitoring the CPU usage status?" In this example, the ECS server is a server product that provides ECS services, and the ECS service is a cloud computing service. In addition, the second difference between the original after-sales problem and the standard after-sales problem is that the original after-sales problem does not necessarily have a solution, while the standard after-sales problem is a part of the after-sales problem with a solution. In order to facilitate distinction and description, the solution corresponding to the standard after-sales problem is called a standard solution. Among them, with the continuous accumulation of standard after-sales problems and their corresponding standard solutions, an after-sales knowledge base corresponding to cloud computing services can be formed, and the after-sales knowledge base stores standard after-sales problems and standard solutions corresponding to cloud computing services.

[0032] In this embodiment, after-sales problems corresponding to cloud computing services can be analyzed in multiple dimensions. For example, information about the cloud computing services specifically involved in each after-sales problem can be analyzed from the dimension of the products involved in the after-sales problem. The type of problem can be analyzed to determine which type of problem each after-sales problem belongs to. The severity of the problem can be analyzed from the dimension of the severity of the problem. The detailed information of each after-sales problem can be analyzed from the dimension of the details of the problem. The accuracy of problem expression can be used to determine whether each after-sales problem has been accurately described, and so on.

[0033] Based on the analysis results of after-sales problems corresponding to cloud computing services, it is possible to determine from which dimensions it is more reasonable to define the types of problems involved in the after-sales of cloud computing services, and give definitions of these problem types, which can be used as the problem-type entity types required to build the knowledge graph. In addition to defining the types of problems involved in the after-sales of cloud computing services, it is also necessary to define the solution from the after-sales dimension of cloud computing services, and use the definition of the solution as the answer-type entity type required to build the knowledge graph. Based on this, according to the analysis results of after-sales problems of cloud computing services, a variety of problem-type entity types can be generated, where different problem-type entity types are used to describe different types of after-sales problems or describe after-sales problems from different dimensions; and an answer-type entity type is generated, which is used to describe solutions to after-sales problems.

[0034] Furthermore, after obtaining a plurality of question-type entity types and answer-type entity types, on the one hand, the edge relationships between the plurality of question-type entity types can be determined based on the association relationships between the after-sales problems described by the plurality of question-type entity types. For example, there is an edge relationship between different question-type entity types that involve the same cloud computing service for the described after-sales problems, there is an edge relationship between question-type entity types that describe the same or related after-sales problems from different dimensions, and so on. On the other hand, the edge relationships between the plurality of question-type entity types and the answer-type entity types can also be determined based on the demand information for solutions to the after-sales problems described by the plurality of question-type entity types. Among them, the demand information for solutions to the after-sales problems described by the question-type entity type refers to whether the after-sales problems described by the question-type entity type require a solution. If a solution is required, there is an edge relationship between the question-type entity type and the answer-type entity type. If a solution is not required, there is no edge relationship between the question-type entity type and the answer-type entity type.

[0035] After determining the edge relationships between multiple question-type entity types and between multiple question-type entity types and answer-type entity types, a knowledge graph relationship diagram can be generated based on multiple question-type entity types, answer-type entity types, and these edge relationships. The knowledge graph relationship diagram is the basis for generating the knowledge graph required for cloud computing after-sales service. The question-type entity types, answer-type entity types, and the edge relationships between these entity types are defined in the knowledge graph relationship diagram.

[0036] Further optionally, in the process of determining the edge relations between multiple question-type entity types and between multiple question-type entity types and answer-type entity types, the type of edge relations can also be determined, and the type of edge relations is used to describe the association relationship between the two entity types corresponding to the edge relations. In addition, in an embodiment of the present application, the names of multiple question-type entity types and the names of the edge relations between multiple question-type entity types and between multiple question-type entity types and answer-type entity types can also be determined based on the after-sales problems described by the multiple question-type entity types. In an embodiment of the present application, the naming method of the names of each entity type and the names of each edge relation is not limited, and any naming method that can distinguish these entity types and edge relations is applicable to the embodiment of the present application. Accordingly, a name with a unique identification function can also be determined for the answer-type entity type.

[0037] Further optionally, in the embodiment of the present application, cloud computing services related to after-sales problems described by multiple question-type entity types can also be obtained, and the cloud computing services can be used as attribute information of corresponding question-type entity types and / or answer-type entity types. Of course, the names of the above-mentioned question-type entity types can also be used as attribute information of the question-type entity types, and similarly, the names of the answer-type entity types can also be used as their attribute information.

[0038] In the above-mentioned embodiments of the present application, the number and specific implementation of problem-type entity types are not limited, and all entity types that can reflect after-sales problems of cloud computing services are applicable to the embodiments of the present application. In an optional embodiment of the present application, a variety of problem-type entity types are divided into two categories, the first category is problem expression-type entity types, and the second category is problem description-type entity types. Each category can include one or more (i.e., at least one) entity types, that is, a variety of problem-type entity types include at least one problem expression-type entity type and at least one problem description-type entity type. Among them, each problem expression-type entity type is used to describe a type of after-sales problem, and each problem description-type entity type is used to describe the detailed information of the after-sales problem from one dimension.

[0039] Furthermore, in an optional embodiment of the present application, 7 entity types are proposed for the after-sales field of cloud computing services, and names are configured for these 7 entity types. Specifically, these 7 entity types include: "Problem phenomenon", "Fuzzy problem", "Details item", "Error item", "Diagnosis item", "System version", and "Solution". It should be noted that the names of these 7 entity types are not limited to the examples here, and other names that can express the same definition are also applicable to the embodiments of the present application. The definitions of these 7 entity types are shown in Table 1 below:

[0040] Table 1

[0041]

[0042]

[0043] Among the above 7 entity types, problem phenomenon, fuzzy problem, detail item, error item, diagnosis item and system version belong to the problem entity type; further, problem phenomenon and fuzzy problem belong to the problem expression entity type, detail item, error item, diagnosis item and system version belong to the problem description entity type. Solution belongs to the answer entity type.

[0044] Further, for the above-mentioned 7 entity types, the above-mentioned step 103, i.e., determining the edge relationships between the various question-type entity types and between the various question-type entity types and the answer-type entity types based on the associations between the after-sales problems described by the various question-type entity types and the demand information for solutions of the after-sales problems, can be implemented in the following manner: on the one hand, based on the associations between the after-sales problems described by the various question-type entity types, determining the edge relationships between the problem phenomenon and the diagnosis item, between the error item and the problem phenomenon, between the fuzzy problem and the problem phenomenon, and between the fuzzy problem and the error item; on the other hand, based on the demand information for solutions of the after-sales problems described by the various question-type entity types, determining the edge relationships between the problem phenomenon and the solution, between the detail item and the solution, between the system version and the solution, between the diagnosis item and the solution, and between the error item and the solution. Among them, the edge relationships between the 7 entity types are shown in Table 2 below:

[0045] Table 2

[0046]

[0047] In the above Table 2, the edge relationship is from the source node to the target node, and the source node or the target node represents an entity type. In addition, in the above Table 2, the names of various edge relationships are given, and the names can indicate the type of edge relationship. In detail, the edge relationship from the problem phenomenon to the solution is a problem-solving edge relationship, which indicates that a solution needs to be given for the key phenomenon of the after-sales problem described in the problem phenomenon; the edge relationship from the detail item to the solution is a detail description edge relationship, which indicates that a solution needs to be given for the detail information of the after-sales problem described in the detail item; the edge relationship from the system version to the solution is a system-type edge relationship, which indicates that a solution needs to be given for the operating system or database information involved in the after-sales problem described in the system version; the edge relationship from the problem phenomenon to the diagnosis item is a problem diagnosis edge relationship, which indicates that the key phenomenon of the after-sales problem described in the problem phenomenon needs to be diagnosed; The edge relationship from the diagnosis item to the solution is a diagnosis-solution type edge relationship, which indicates that a solution needs to be given for the diagnosis information obtained from the diagnosis item; the edge relationship from the error item to the problem phenomenon is an error phenomenon type edge relationship, which indicates that it is necessary to determine the problem phenomenon to which the error information in the after-sales problem described in the error item belongs; the edge relationship from the error item to the solution is an error-solution type edge relationship, which indicates that a solution needs to be given for the error information in the after-sales problem described in the error item; the edge relationship from the fuzzy problem to the problem phenomenon or error item is an inclusion type edge relationship, which indicates that an incomplete or ambiguous after-sales problem described in the fuzzy problem may belong to a problem phenomenon or error item.

[0048] For the 7 entity types given in Table 1 and the edge relationships between the 7 entity types given in Table 2, each entity type is taken as a node, and edges are established between the nodes according to the edge relationships between these entity types. Figure 2 The knowledge graph relationship diagram shown in Figure 1. Figure 2 In the problem model, product nodes appear as attributes of problem phenomenon nodes and fuzzy problem nodes.

[0049] After obtaining the knowledge graph relationship diagram, we can mine entity phrases from various after-sales problems and existing solutions corresponding to cloud computing services, obtain entity phrases under various entity types, and further extract the edge relationships between these entity phrases to construct a specific knowledge graph. Figure 2 The knowledge graph relationship shown in Figure 3 The knowledge graph shown. It should be noted that Figure 3 The graph shown can be a complete knowledge graph or a part of the complete knowledge graph, depending on the number of entity phrases mined. Figure 2In , the entity types represented by the nodes are distinguished by the grayscale of the nodes. Nodes with different grayscales represent different entity types, and nodes with the same grayscale represent the same entity type. Similarly, in Figure 3 In , the entity type of the entity phrase represented by the node is also distinguished by the grayscale of the node. Nodes with different grayscales represent entity phrases belonging to different entity types, and nodes with the same grayscale represent entity phrases belonging to the same entity type. In addition, Figure 2 and Figure 3 , nodes with the same grayscale represent the same entity type.

[0050] exist Figure 3 The knowledge graph shown includes the following entity phrases: Linux, monitoring tool, CPU diagnosis, CPU overload, server overload, memory overload, bandwidth full, ECS, Linux server CPU overload monitoring program; among them, "Linux server CPU overload monitoring program" belongs to Figure 2 The entity phrase under "Solution" shown, "Linux" belongs to Figure 2 The entity phrase under "System Version" shown below, "Monitoring Tools" belongs to Figure 2 The entity phrase "CPU Diagnosis" under the "Details" shown belongs to Figure 2 The entity phrase "Server load is too high" under the "Diagnostic Item" shown belongs to Figure 2 The entity phrase under "Fuzzy Question" shown below, "CPU Load Too High" belongs to Figure 2 The entity phrase under "Problem phenomenon" shown in the figure, "memory is too high" belongs to Figure 2 The entity phrase under the "problem phenomenon" shown in the figure, "bandwidth full" belongs to Figure 2 The entity phrase under the "problem phenomenon" shown in the figure, "ECS" belongs to Figure 2 Entity phrases under the "Product" attribute shown.

[0051] about Figure 3 The following is an example of how the knowledge graph can be used: When a user asks the following after-sales question: "The CPU of my Linux ECS server is fully utilized. Is there any monitoring tool I can use?", various entity phrases such as "Linux", "ECS server", "CPU fully utilized", and "monitoring tool" can be decomposed from the after-sales question. Figure 3 Reasoning in the knowledge graph shown can locate the solution of "Monitoring program for excessive CPU load on Linux servers", and then give accurate responses to after-sales questions raised by users. This can not only improve the after-sales experience of cloud computing services, but also improve the accuracy of identifying after-sales problems, the success rate of solving them, and the efficiency of solving them.

[0052] In the following embodiments, the process of constructing a knowledge graph based on a knowledge graph relationship graph provided in the embodiments of the present application will be described in detail.

[0053] Figure 4 A flow chart of a method for constructing a knowledge graph for cloud computing provided for an exemplary embodiment of the present application. The method for constructing a knowledge graph provided in this embodiment belongs to the cold-start construction process of a knowledge graph. The process of cold-starting a knowledge graph refers to the process of directly constructing a domain knowledge graph for cloud computing based on a knowledge graph relationship graph without a basic knowledge graph. In this embodiment, the knowledge graph relationship graph only defines a variety of question-type entity types, answer-type entity types, and edge relationships between question-type entity types and edge relationships between question-type entity types and answer-type entity types for cloud computing after-sales service. Among them, cloud computing after-sales service refers to the after-sales service provided for cloud computing services. Specifically, such as Figure 4 As shown, the method for constructing the knowledge graph includes:

[0054] 401. Mining entity phrases for original after-sales questions corresponding to the cloud computing service and standard after-sales questions in the after-sales knowledge base to obtain multiple question-type entity phrases;

[0055] 402. Based on the multiple question-type entity types defined in the pre-generated knowledge graph relationship diagram, combined with the attribute information of the multiple question-type entity phrases, determine the question-type entity types to which the multiple question-type entity phrases belong;

[0056] 403. According to the answer-type entity type defined in the knowledge graph relationship diagram, select at least one answer-type entity phrase belonging to the answer-type entity type from the standard solutions in the after-sales knowledge base;

[0057] 404. According to the edge relationships between question-type entity types and between question-type entity types and answer-type entity types defined in the knowledge graph relationship diagram, edges are added between multiple question-type entity phrases and between multiple question-type entity phrases and at least one answer-type entity phrase to obtain a knowledge graph for cloud computing after-sales service.

[0058] In this embodiment, on the one hand, the original after-sales questions corresponding to the cloud computing service and the standard after-sales questions in the after-sales knowledge base are used as the data basis to mine entity phrases to obtain multiple question-type entity phrases; the multiple question-type entity types defined in the knowledge graph relationship diagram are combined with the attribute information of the multiple question-type entity phrases to classify the multiple question-type entity phrases and determine the question-type entity type to which each question-type entity phrase belongs. On the other hand, according to the answer-type entity type defined in the knowledge graph relationship diagram, at least one answer-type entity phrase belonging to the answer-type entity type is selected from the standard solutions in the after-sales knowledge base. Afterwards, according to the edge relationships between the question-type entity types and between the question-type entity type and the answer-type entity type defined in the knowledge graph relationship diagram, edges are added between the multiple question-type entity phrases and between the multiple question-type entity phrases and at least one answer-type entity phrase to obtain a knowledge graph for cloud computing after-sales service.

[0059] In the embodiment of the present application, the implementation method of step 401 is not limited, and any implementation method that can mine entity phrases from original after-sales questions and standard after-sales questions is applicable to the embodiment of the present application. In an optional embodiment of the present application, an implementation method of step 401 is given, such as Figure 5 As shown, the implementation includes the following steps: data source selection, phrase mining, phrase vectorization, phrase clustering and entity phrase formation.

[0060] Data source selection: In the embodiment of the present application, it is necessary to construct a knowledge graph for the after-sales field of cloud computing services. Combined with the above-mentioned knowledge graph relationship diagram, it can be seen that to construct the knowledge graph in the after-sales field, it is necessary to mine entity phrases of the problem-type entity type. Therefore, the after-sales problems of cloud computing services are selected as the data source required to construct the knowledge graph. In the embodiment of the present application, the after-sales problems of cloud computing services include two parts. One part is the original after-sales problems submitted by users, which are mainly after-sales problems submitted by users online or through work orders. This part of the original after-sales problems is relatively casual and complex in expression; the other part is the standard after-sales problems that have been precipitated in the after-sales knowledge base corresponding to the cloud computing service. The standard after-sales problems are relatively high-quality problem expressions after processing in expression. They are similar problems to the original after-sales problems, and the standard after-sales problems have corresponding standard solutions. The standard solutions in the after-sales knowledge base can also be regarded as knowledge points.

[0061] Phrase Mining: The purpose is to automatically extract high-quality entity phrases from a large amount of text corpus, mainly to solve the problem of insufficient professional dictionaries in professional fields and reduce manual sorting costs. In an embodiment of the present application, after determining the data source required for mining entity phrases, entity phrases suitable as question-type entity types are mined from the original after-sales questions and standard after-sales questions contained in the data source to solve the problem of insufficient question-type entity phrases faced in building a knowledge graph in the after-sales field for cloud computing services. Optionally, hot or high-frequency entity phrases can be mined from these original after-sales questions and standard after-sales questions. The mining of entity phrases specifically includes the following steps:

[0062] First, a syntactic dependency analysis is performed for each original after-sales question or standard after-sales question, so as to structure the original after-sales question or standard after-sales question and obtain the words contained in the original after-sales question or standard after-sales question; then, the frequency of the combination of words is counted according to whether there is a relationship between the two words; then, according to the frequency of the combination of words, a word pair with a frequency greater than the set first frequency threshold or the highest frequency is selected from each word, and these word pairs are extracted according to the word order to form phrases. These phrases can be used as entity phrases of question-type entity types. For ease of description, the entity phrases of question-type entity types are called question-type entity phrases. Further optionally, some words with a frequency greater than the second frequency threshold or the highest frequency can also be extracted, and these words can also be used as part of the question-type entity phrases. For example, Windows, CentOS, etc. belong to single words, but they are also meaningful entity phrases in the knowledge graph. In other words, the question-type entity phrase in the embodiment of the present application is not necessarily a phrase, but can also be a word.

[0063] It should be noted that in addition to using the method provided in the above embodiment to mine entity phrases, text ranking (TextRank) algorithm, AutoPhrase, deep neural network (Deep Neural Networks, DNN), named entity recognition (Named Entity Recognition, NER) and other methods can also be used to mine entity phrases.

[0064] Phrase vectorization: After phrase mining, the original after-sales questions or standard after-sales questions become question-type entity phrases. These question-type entity phrases are input into the pre-trained language representation model for vectorization processing to obtain vector expressions of multiple question-type entity phrases. In the embodiment of the present application, the language representation model is not limited. For example, a neural network model that encodes text using a bidirectional Transformer structure can be used. Specifically, a BERT (Pre-training of Deep Bidirectional Transformers) model after pre-training and fine-tuning can be used. The model is an autoencoder language model (Autoencoder LM), which uses a bidirectional Transformer structure and designs two tasks to pre-train the model. The first task is to train the language model using MaskLM, that is, when entering a sentence, randomly select some words to be predicted, and then replace them with a special symbol [MASK], and then let the model learn the words that should be filled in these places according to the given labels. The second task adds a sentence-level continuity prediction task based on the bidirectional Transformer language model, that is, predicting whether the two text segments input into BERT are continuous texts. Introducing this task can better enable the model to learn the relationship between continuous text segments, and finally represent the words as mathematical vector expressions through the embedding model. Of course, in addition to the BERT model, neural network models such as ELMo and T5 (Transfer Text-to-Text Transformer) can also be used to vectorize problem-type entity phrases.

[0065] Phrase clustering: After vectorizing the problem entity phrases extracted above, multiple rounds of density clustering can be performed on the problem entity phrases according to the vector expression of the problem entity phrases to obtain multiple groups of problem entity phrases. The density clustering algorithm can be but is not limited to: DBSCAN (Density-Based Spatial Clustering of Applications with Noise). In this embodiment, density clustering refers to calculating the distribution density between problem entity phrases through the vector expression of the problem entity phrases, analyzing the connectivity between the problem entity phrases based on the distribution density, and continuously expanding the clustering clusters based on the connectivity between the problem entity phrases, and finally obtaining multiple clusters. The problem entity phrases in each cluster are a group of problem entity phrases. The problem entity phrases in the same group are entity phrases that express the same or similar meanings, such as "cannot connect", "cannot connect", and "cannot connect". These entity phrases express the same or similar meanings and will be clustered into the same group. Further optionally, when calculating the distribution density between the problem entity phrases through the vector expression of the problem entity phrases, the cosine similarity between the problem entity phrases can be calculated according to the vector expression of the problem entity phrases, and the problem entity phrases are subjected to multiple rounds of density clustering based on the cosine similarity between the problem entity phrases; wherein, in each round of clustering, the problem entity phrases whose cosine similarity is greater than the cosine similarity threshold used for clustering in that round are divided into the same clustering cluster, and in addition, except for the first round of clustering, the clustering of other rounds is based on the clustering results of the previous round, and further clustering is performed for each cluster obtained in the previous round. It should be noted that the cosine similarity threshold used in different rounds of clustering is not the same, and the threshold will gradually increase from low to high as the number of clustering rounds increases. For example, assuming that there are M question entity phrases, in the first round of clustering, the first cosine similarity threshold A1 is used, and the M question entity phrases are clustered into three clusters B1; in the second round of clustering, the second cosine similarity threshold A2 is used to further cluster the question entity phrases in each cluster B1 to obtain 2 or more clusters. The number of new clusters obtained by further clustering different clusters B1 may not be the same; and so on, until the clustering end condition is met, and the final cluster is obtained. It should be noted that in addition to using cosine similarity in the above clustering process, Euclidean distance, Ming-style distance, etc. can also be used to characterize the distribution density between question entity phrases.

[0066] In addition, in addition to the density clustering method, the above clustering process may also use clustering algorithms such as K-means and spectral clustering, which are not limited.

[0067] Entity formation: After the above steps, the clustering results of question entity phrases can be obtained, that is, multiple groups of question entity phrases. Further, each group of question entity phrases can be manually reviewed, and some question entity phrases that are not suitable for building a knowledge graph or clustering errors can be adjusted to obtain better quality question entity phrases.

[0068] It is to be noted that the above-mentioned phrase vectorization, phrase clustering and entity formation are all optional operations, which can be completed in the entity phrase mining stage or in other subsequent stages when needed, and there is no limitation on this. In this embodiment, the above-mentioned operations are completed in the entity phrase mining stage as an example for explanation.

[0069] In the above embodiment, after mining entity phrases, multiple question-type entity phrases can be obtained. Further, the multiple question-type entity phrases can be classified in combination with the multiple question-type entity types defined in the knowledge graph relationship diagram to determine the question-type entity types to which the multiple question-type entity phrases belong. In the embodiment of the present application, the implementation method of determining the question-type entity types to which the multiple question-type entity phrases belong is not limited. In an optional embodiment, the detailed implementation process of the above step 402 includes: according to the meaning expression of multiple question-type entity phrases, the question-type entity phrases expressing the same or similar meaning are divided into a group to obtain multiple groups of question-type entity phrases; according to the frequency of occurrence of each group of question-type entity phrases in the standard after-sales problems and standard solutions in the after-sales knowledge base, the entity type discrimination information corresponding to each group of question-type entity phrases is determined; according to the entity type discrimination information corresponding to each group of question-type entity phrases, the question-type entity type to which each group of question-type entity phrases belongs is determined.

[0070] It is hereby explained that in the embodiments of the present application, the above-mentioned implementation method of obtaining multiple groups of question-type entity phrases is not limited. In an optional embodiment, the clustering method mentioned in the above-mentioned phrase clustering part can be adopted, specifically including: inputting multiple question-type entity phrases into the language representation model for vectorization processing to obtain vector expressions of multiple question-type entity phrases; performing multiple rounds of density clustering on the multiple question-type entity phrases according to the vector expressions of the multiple question-type entity phrases to obtain multiple groups of question-type entity phrases. It should be noted that when the clustering method mentioned in the above-mentioned phrase clustering part is adopted to divide the question-type entity phrases expressing the same or similar meaning into a group to obtain multiple groups of question-type entity phrases, the operation is performed either in the above-mentioned entity phrase mining stage or in the entity phrase classification stage described in step 402, and there is no limitation on this.

[0071] In an optional embodiment, the entity type discrimination information corresponding to each group of problem-type entity phrases is determined according to the frequency of occurrence of each group of problem-type entity phrases in standard after-sales problems and standard solutions in the after-sales knowledge base, including: for each group of problem-type entity phrases, the frequency of occurrence of the group of problem-type entity phrases in standard after-sales problems is used as the first entity type discrimination information, and the ratio of the frequency of occurrence of the group of problem-type entity phrases in standard after-sales problems to the frequency of occurrence in standard solutions is calculated as the second entity type discrimination information.

[0072] Further optionally, the multiple question-type entity types include: question-expressing entity types and question-describing entity types; according to the entity type discrimination information corresponding to each group of question-type entity phrases, the question-type entity type to which each group of question-type entity phrases belongs is determined, including: for each group of question-type entity phrases, when the first entity type discrimination information is greater than a set third frequency threshold, and the second entity type discrimination information is greater than a set ratio threshold, determining that the group of question-type entity phrases belongs to the question-expressing entity type; correspondingly, when the first entity type discrimination information is less than or equal to the set frequency threshold, or the second entity type discrimination information is less than or equal to the set ratio threshold, determining that the group of question-type entity phrases belongs to the question-describing entity type.

[0073] Further optionally, the problem expression type entity type includes at least problem phenomena. The definition of the problem phenomenon entity type can be found in the above-mentioned embodiment and will not be repeated here. Based on this, when the first entity type discrimination information is greater than the set third frequency threshold, and the second entity type discrimination information is greater than the set ratio threshold, it is determined that the group of problem entity phrases belongs to the problem expression type entity type, including: when the first entity type discrimination information is greater than the set third frequency threshold, and the second entity type discrimination information is greater than the set ratio threshold, it is determined that the problem entity type to which the group of problem entity phrases belong is a problem phenomenon.

[0074] Further optionally, the above-mentioned problem description type entity type at least includes: detail item, error item and system version. The definition of the centralized problem description type entity type can be found in the above-mentioned embodiment, which will not be repeated here. Based on this, when the first entity type discrimination information is less than or equal to the set third frequency threshold, or the second entity type discrimination information is less than or equal to the set ratio threshold, it is determined that the group of problem entity phrases belongs to the problem description type entity type, including: when the first entity type discrimination information is greater than the set frequency threshold, and the second entity type discrimination information is less than or equal to the set ratio threshold, the problem entity type to which the group of problem entity phrases belongs is determined to be the system version; when the first entity type discrimination information is less than or equal to the set third frequency threshold, and the problem entity phrase appears in the error reporting feature rule table, the problem entity type to which the group of problem entity phrases belongs is determined to be the error reporting item; when the first entity type discrimination information is less than or equal to the set third frequency threshold, and the problem entity phrase does not appear in the error reporting feature rule table, the problem entity type to which the group of problem entity phrases belongs is determined to be the detail item.

[0075] In an optional embodiment of the present application, the multiple problem-type entity types defined in the knowledge graph relationship diagram include: at least one problem expression type entity type and at least one problem description type entity type; wherein at least one problem expression type entity type includes: problem phenomenon and fuzzy problem; at least one problem description type entity type includes: detail item, error item, diagnosis item and system version. On this basis, the embodiment of the present application gives a detailed implementation method of the above step 402 when determining the problem-type entity type to which multiple problem-type entity phrases belong. The detailed implementation method is also a comprehensive description of the above optional embodiments. Specifically, Figure 6 As shown, the detailed implementation process of step 402 includes the steps of data source selection, frequency statistics, decision classification and classification results.

[0076] Data source selection: The data sources required for this step include two parts. One part is the multiple question-type entity phrases obtained above, and the other part is the standard after-sales questions and standard solutions in the after-sales knowledge base.

[0077] Frequency statistics: This step requires statistics on the frequency of occurrence of the multiple problem-type entity phrases obtained above in the standard after-sales questions in the after-sales knowledge base. For example, assuming that the problem-type entity phrase "ECS cannot be connected" has appeared in 5 standard after-sales questions, the frequency of occurrence of the problem-type entity phrase in the standard after-sales questions can be recorded as 5; the sum of the frequency of occurrence of the problem-type entity phrases in the standard after-sales questions of the same group is used as the frequency of occurrence of the problem-type entity phrases in the standard after-sales questions of this group. In addition, it is also necessary to count the frequency of occurrence of the multiple problem-type entity phrases obtained above in the standard solutions (i.e., knowledge points) in the after-sales knowledge base. For example, assuming that the problem-type entity phrase "ECS cannot be connected" has appeared in 2 standard solutions as knowledge points, the frequency of occurrence of the problem-type entity phrase in the standard solution can be recorded as 2, and the sum of the frequency of occurrence of the problem-type entity phrases in the standard solutions as knowledge points of the same group is used as the frequency of occurrence of the problem-type entity phrases in the standard solutions as knowledge points.

[0078] Decision classification: This step is a key step to determine which type of question entity the question entity phrase belongs to. In this embodiment, the following classification steps are specifically adopted:

[0079] Step 1: For each group of question-type entity phrases, use their frequency of occurrence in standard after-sales questions as the first entity type discrimination information, and determine whether their frequency of occurrence in standard after-sales questions is greater than a preset third frequency threshold. If so, execute steps 2 and 3; if not, execute step 4.

[0080] Step 2: According to the frequency of occurrence of each group of problem-type entity phrases in the standard after-sales questions and the frequency of occurrence in the standard solution as a knowledge point, the formula d=f1 / f2 is used to calculate the second entity type discrimination information d corresponding to the group of problem-type entity phrases. Among them, f1 represents the frequency of occurrence of each group of problem-type entity phrases in the standard after-sales questions, and f2 represents the frequency of occurrence of each group of problem-type entity phrases in the standard solution as a knowledge point.

[0081] Step 3: For each group of problem-type entity phrases, according to a pre-set ratio threshold, if the second entity type discrimination information d corresponding to the group of problem-type entity phrases is greater than or equal to the ratio threshold, it means that the group of problem-type entity phrases appears in many standard after-sales problems corresponding to a standard solution, and the problem-type entity type to which the group of problem-type entity phrases belongs is determined to be "problem phenomenon"; otherwise, it means that the group of problem-type entity phrases appears more frequently in many standard solutions, and the problem-type entity type to which the group of problem-type entity phrases belongs is determined to be "system version".

[0082] Step 4: For each group of problem entity phrases, if the frequency of occurrence of the group of problem entity phrases in the standard after-sales questions is less than or equal to the preset third frequency threshold, first determine whether the problem entity type to which the group of problem entity phrase entities belong is an "error item" according to the error feature rule table. For example, the error feature rule table includes: error words such as error, exception, response timeout, and failure, error codes such as 404 and 505, or various error rule features such as super-long continuous English phrases and phrases containing "error is" before the phrase. If a group of problem entity phrases contains error words such as error, exception, time out, and failed, or contains error codes such as 404 and 505, or super-long continuous English phrases and phrases containing "error is", etc., then it can be determined that the problem entity type to which the group of problem entity phrases belong is an error item. For the other remaining groups of problem entity phrases, it can be determined that the problem entity type to which they belong is a detail item.

[0083] Classification results: According to the above steps, this embodiment classifies multiple problem-type entity phrases into four problem-type entity types: problem phenomenon, system version, error item, and detail item in groups. Further optionally, the classification results can be manually reviewed and verified later. It should be noted that, with respect to the two problem-type entity types of diagnostic items and fuzzy questions, the problem-type entity phrases required for these two entity types can be manually constructed, but it is not limited to this. For example, in the above step 4, the two problem-type entity types of diagnostic items and fuzzy questions can be further taken into consideration for further subdivision.

[0084] In an embodiment of the present application, in addition to determining the question-type entity type to which multiple question-type entity phrases belong, it is also necessary to combine the answer-type entity type defined in the knowledge graph relationship diagram, and select at least one answer-type entity phrase belonging to the answer-type entity type from the standard solutions in the after-sales knowledge base, to provide another data condition for constructing the knowledge graph. In an embodiment of the present application, the implementation method of selecting at least one answer-type entity phrase belonging to the answer-type entity type from the standard solutions in the after-sales knowledge base is not limited. In an optional embodiment, according to the answer-type entity type defined in the knowledge graph relationship diagram, at least one answer-type entity phrase belonging to the answer-type entity type is selected from the standard solutions in the after-sales knowledge base, that is, an implementation method of the above step 403 includes:

[0085] Combine question-type entity phrases belonging to different question-type entity types to obtain multiple entity phrase combinations, and count the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base; select at least one standard solution from the standard solutions in the after-sales knowledge base according to the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base as at least one answer-type entity phrase.

[0086] Further optionally, the above-mentioned selecting at least one standard solution from the standard solutions in the after-sales knowledge base according to the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base respectively as an implementation method of at least one answer-type entity phrase includes: for each entity phrase combination, if the co-occurrence frequency of the entity phrase combination in the original after-sales question is greater than a first co-occurrence frequency threshold and the co-occurrence frequency in the after-sales knowledge base is greater than a second co-occurrence frequency threshold, then using the standard solution in the after-sales knowledge base that has co-occurred with the entity phrase combination as the answer-type entity phrase corresponding to the entity phrase combination.

[0087] Further, in an optional embodiment of the present application, a detailed implementation of the above step 403 when selecting at least one answer-type entity phrase is given, and the detailed implementation is also a comprehensive description of the above optional embodiments. Specifically, Figure 7 As shown, a detailed implementation process of step 403 includes the steps of data source selection, first co-occurrence statistics, second co-occurrence statistics and entity phrase selection.

[0088] Data source selection: In this step, the data source consists of three parts. One part is the multiple question-type entity phrases obtained above, another part is the original after-sales questions, and the third part is the standard after-sales questions and standard solutions in the after-sales knowledge base.

[0089] First co-occurrence statistics: After obtaining the data source, this step first combines the problem-type entity phrases classified under the problem-type entity types such as "problem phenomenon" and "detail items", "error items", and "system versions" to obtain multiple entity phrase combinations. Optionally, the problem-type entity phrases classified under the problem-type entity types such as "problem phenomenon" and "detail items", "error items", and "system versions" can be combined in permutations and combinations, and each permutation and combination obtains an entity phrase combination. For example, "ECS cannot connect", combined with "password reset", and "Linux", can obtain an entity phrase combination. For each entity phrase combination, the frequency of simultaneous appearance of the entity phrase combination in the original after-sales problem, that is, the co-occurrence frequency, can be counted and recorded.

[0090] Second co-occurrence statistics: This step is mainly used to count the co-occurrence frequencies of each entity phrase combination mentioned above in each standard after-sales problem and the corresponding standard solution in the after-sales knowledge base, that is, the co-occurrence frequency, and record it. For the same entity phrase combination, the co-occurrence frequency in each standard after-sales problem and the co-occurrence frequency in each standard solution are summed as the co-occurrence frequency of the entity phrase combination in the after-sales knowledge base.

[0091] Entity phrase selection: This step can also be called relationship extraction. It mainly identifies answer-type entity phrases from the standard solutions in the after-sales knowledge base based on the co-occurrence frequencies of each entity phrase combination obtained in the above steps in the original after-sales question and the after-sales knowledge base. For each entity phrase combination, according to its co-occurrence frequency in the original after-sales question and the after-sales knowledge base, one of the following three situations will occur:

[0092] Case 1: For any entity phrase combination, its co-occurrence frequency in the original after-sales problem is high, for example, greater than the first co-occurrence frequency threshold, and at the same time, its co-occurrence frequency in the after-sales knowledge base is also high, for example, greater than the second co-occurrence frequency threshold, which means that the standard solution co-occurred by the entity phrase combination can be used as the solution to the original after-sales problem co-occurred by the entity phrase combination. Therefore, the standard solution co-occurred by the entity phrase combination can be used as the answer-type entity phrase corresponding to the entity phrase combination. It is hereby explained that in the embodiment of the present application, the values ​​of the first co-occurrence frequency threshold and the second co-occurrence frequency threshold are not limited, and can be flexibly set according to application requirements, and the size relationship between the two co-occurrence frequency thresholds is not limited.

[0093] Case 2: For any entity phrase combination, its co-occurrence frequency in the original after-sales question is high, for example, greater than the first co-occurrence frequency threshold, but its co-occurrence frequency in the after-sales knowledge base is also low, for example, less than or equal to the second co-occurrence frequency threshold, or even has never co-occurred in the after-sales knowledge base, which indicates that the after-sales knowledge base may lack standard solutions corresponding to the original after-sales questions that have co-occurred with the entity phrase combination. Further optionally, standard solutions corresponding to the original after-sales questions that have co-occurred with the entity phrase combination can be added to the after-sales knowledge base, and the newly added standard solutions can be used as the answer-type entity phrases corresponding to the entity phrase combination.

[0094] Case 3: For any entity phrase combination, its co-occurrence frequency in the original after-sales question is low, for example, less than or equal to the first co-occurrence frequency threshold, and its co-occurrence frequency in the after-sales knowledge base is also low, for example, less than or equal to the second co-occurrence frequency threshold, or even has never co-occurred in the after-sales knowledge base, which indicates that the entity phrase combination may not form a valuable question-type entity phrase, and there is no corresponding solution. This type of entity phrase combination can be retained without paying attention to the corresponding answer-type entity phrase. It should be noted that not all entity phrase combinations will correspond to answer-type entity phrases.

[0095] After obtaining a plurality of question-type entity phrases and the question-type entity types to which they belong, and at least one answer-type entity phrase, edges can be added between the plurality of question-type entity phrases and between the plurality of question-type entity phrases and the at least one answer-type entity phrase according to the edge relationships between the question-type entity types and between the question-type entity types and the answer-type entity types defined in the knowledge graph relationship graph, thereby obtaining a knowledge graph for cloud computing after-sales service. In an optional embodiment, an implementation method for obtaining a knowledge graph for cloud computing after-sales service includes: for each entity phrase combination corresponding to an answer-type entity phrase, according to the edge relationships between the question-type entity types defined in the knowledge graph relationship graph, edges are added between the question-type entity phrases in the entity phrase combination; further, according to the edge relationships between the question-type entity types and the answer-type entity types defined in the knowledge graph relationship graph, edges are added between the question-type entity phrases in the entity phrase combination and the answer-type entity phrase corresponding to the entity phrase combination, thereby obtaining a knowledge graph for cloud computing after-sales service.

[0096] Example: Combination Figure 2 As shown in the knowledge graph relationship diagram, suppose an entity phrase combination is: "Linux, CPU diagnosis, CPU load is too high", and its corresponding answer-type entity phrase is "Linux server CPU load is too high monitoring program", "Linux" belongs to the entity phrase of system version type, "CPU diagnosis" belongs to the entity phrase of diagnosis item type, and "CPU load is too high" belongs to the entity phrase of problem phenomenon type. Figure 2 From the edge relationships between the problem entity types shown, we can see that we need to add an edge between the problem entity phrase "CPU load is too high" and the problem entity phrase "CPU diagnosis". Figure 2The edge relationship between the question entity type and the answer entity type is shown. An edge is added between the question entity phrase "CPU load is too high" and the answer entity phrase "Monitoring program for Linux server CPU load is too high", an edge is added between the question entity phrase "CPU diagnosis" and the answer entity phrase "Monitoring program for Linux server CPU load is too high", and an edge is added between the question entity phrase "Linux" and the answer entity phrase "Monitoring program for Linux server CPU load is too high". After processing all entity phrase combinations with corresponding answer entity phrases in the same way, the following can be obtained: Figure 3 The knowledge graph shown.

[0097] It should be noted that after obtaining the knowledge graph, it can also be manually reviewed and corrected to correct various entity phrases and the edge relationships between them to obtain a more complete knowledge graph.

[0098] In the embodiments of the present application, an architectural definition of the knowledge graph in the field of cloud computing, namely, the knowledge graph relationship diagram, is innovatively proposed for the first time. A construction method for the cold start phase of the knowledge graph in the field of cloud computing is also innovatively proposed for the first time. The data source only needs the original after-sales questions raised by users and the existing after-sales knowledge base, and there is no need to collect a large amount of sample data. This can reduce the requirements for the data source for knowledge graph construction and improve the efficiency of knowledge graph construction.

[0099] Furthermore, after obtaining the knowledge graph in the field of cloud computing, the constructed knowledge graph can be applied to the after-sales customer robot of cloud computing services to provide timely, accurate and efficient solutions to after-sales issues raised by users during the use of cloud computing services, thereby improving users' experience of cloud computing services and further promoting the use of cloud computing services.

[0100] It should be noted that the execution subject of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 401 to 404 can be device A; for another example, the execution subject of steps 401 and 402 can be device A, and the execution subject of steps 403 and 404 can be device B; and so on.

[0101] In addition, in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel, and the sequence numbers of the operations, such as 401, 402, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0102] Figure 8 A schematic diagram of a knowledge graph relationship diagram generating device for cloud computing provided by an exemplary embodiment of the present application. Figure 8 As shown, the device includes: a first generating module 81, a determining module 82 and a second generating module 83.

[0103] The first generation module 81 is used to generate a plurality of problem-type entity types according to the analysis results of after-sales problems of cloud computing services, and different problem-type entity types are used to describe different types of after-sales problems or describe after-sales problems from different dimensions; and to generate an answer-type entity type, and the answer-type entity type is used to describe solutions to after-sales problems.

[0104] The determination module 82 is used to determine the edge relationships between multiple question-type entity types and between multiple question-type entity types and answer-type entity types based on the association relationships between after-sales problems described by the multiple question-type entity types and the demand information for solutions of the after-sales problems.

[0105] The second generation module 83 is used to generate a knowledge graph relationship diagram based on multiple question-type entity types, answer-type entity types and edge relationships. The knowledge graph relationship diagram is used to generate the knowledge graph required for cloud computing after-sales service.

[0106] In an optional embodiment, the second generating module 83 is further used to: obtain cloud computing services involved in after-sales problems described by the multiple question-type entity types, and use the cloud computing services as attribute information corresponding to the question-type entity type and / or answer-type entity type.

[0107] In an optional embodiment, the second generation module 83 is specifically used to: determine the existence of edge relationships between problem phenomena and diagnosis items, between error items and problem phenomena, between fuzzy problems and problem phenomena, and between fuzzy problems and error items based on the association relationships between after-sales problems described by the multiple problem-type entity types; determine the existence of edge relationships between problem phenomena and solutions, between detail items and solutions, between system versions and solutions, between diagnosis items and solutions, and between error items and solutions based on the demand information for solutions for after-sales problems described by the multiple problem-type entity types; wherein the solution is the name of the answer-type entity type.

[0108] Fig. 9 The schematic diagram of the structure of a knowledge graph construction device for cloud computing provided by the exemplary embodiment of the present application. Fig. 9 As shown, the device includes: a mining module 91, a determination module 92, a selection module 93 and a generation module 94.

[0109] The mining module 91 is used to mine entity phrases for the original after-sales questions corresponding to the cloud computing service and the standard after-sales questions in the after-sales knowledge base to obtain a plurality of question-type entity phrases.

[0110] The determination module 92 is used to determine the question-type entity types to which multiple question-type entity phrases belong based on multiple question-type entity types defined in a pre-generated knowledge graph relationship diagram and in combination with attribute information of multiple question-type entity phrases.

[0111] The selection module 93 is used to select at least one answer-type entity phrase belonging to the answer-type entity type from the standard solutions in the after-sales knowledge base according to the answer-type entity type defined in the knowledge graph relationship diagram.

[0112] A generation module 94 is used to add edges between multiple question-type entity phrases and between multiple question-type entity phrases and at least one answer-type entity phrase according to the edge relationships between question-type entity types and between question-type entity types and answer-type entity types defined in the knowledge graph relationship diagram, so as to obtain a knowledge graph for cloud computing after-sales service.

[0113] In an optional embodiment, when determining the problem-type entity types to which multiple problem-type entity phrases belong, the determination module 92 is specifically used to: divide the problem-type entity phrases expressing the same or similar meanings into a group according to the meanings expressed by the multiple problem-type entity phrases, so as to obtain multiple groups of problem-type entity phrases; determine the entity type discrimination information corresponding to each group of problem-type entity phrases according to the frequency of occurrence of each group of problem-type entity phrases in standard after-sales problems and standard solutions in the after-sales knowledge base; determine the problem-type entity type to which each group of problem-type entity phrases belongs according to the entity type discrimination information corresponding to each group of problem-type entity phrases.

[0114] Further optionally, when the determination module 92 obtains multiple groups of question-type entity phrases, it is specifically used to: input the multiple question-type entity phrases into the language representation model for vectorization processing to obtain vector expressions of the multiple question-type entity phrases; based on the vector expressions of the multiple question-type entity phrases, perform multiple rounds of density clustering on the multiple question-type entity phrases to obtain multiple groups of question-type entity phrases.

[0115] Further optionally, when determining the entity type discrimination information corresponding to each group of problem-type entity phrases, the determination module 92 is specifically used to: for each group of problem-type entity phrases, use the frequency of occurrence of the group of problem-type entity phrases in standard after-sales questions as the first entity type discrimination information, and calculate the ratio of the frequency of occurrence of the group of problem-type entity phrases in standard after-sales questions and the frequency of occurrence in standard solutions as the second entity type discrimination information.

[0116] Further optionally, the above-mentioned multiple question-type entity types include: question-expression-type entity types and question-description-type entity types. Based on this, when determining the entity type discrimination information corresponding to each group of question-type entity phrases, the determination module 92 is specifically used to: for each group of question-type entity phrases, when the first entity type discrimination information is greater than a set frequency threshold, and the second entity type discrimination information is greater than a set ratio threshold, determine that the group of question-type entity phrases belongs to the question-expression-type entity type; when the first entity type discrimination information is less than or equal to the set frequency threshold, or the second entity type discrimination information is less than or equal to the set ratio threshold, determine that the group of question-type entity phrases belongs to the question-description-type entity type.

[0117] Furthermore, the problem expression type entity type at least includes problem phenomena, which are entity types used to describe key phenomena manifested by after-sales problems. Based on this, when determining that the group of problem entity phrases belongs to the problem expression type entity type, the determination module 92 is specifically used to: when the first entity type discrimination information is greater than a set frequency threshold, and the second entity type discrimination information is greater than a set ratio threshold, determine that the problem entity type to which the group of problem entity phrases belongs is a problem phenomenon.

[0118] Furthermore, the entity type of the problem description type includes at least: detail item, error item and system version; detail item is an entity type used to describe the detail information of the after-sales problem, error item is an entity type used to describe the error information appearing in the after-sales problem, and system version is an entity type used to describe the operating system information involved in the after-sales problem. Based on this, when determining that the group of problem entity phrases belongs to the problem description type entity type, the determination module 92 is specifically used to: when the first entity type discrimination information is greater than the set frequency threshold, and the second entity type discrimination information is less than or equal to the set ratio threshold, determine that the problem entity type to which the group of problem entity phrases belongs is the system version; when the first entity type discrimination information is less than or equal to the set frequency threshold, and the problem entity phrase appears in the error feature rule table, determine that the problem entity type to which the group of problem entity phrases belongs is the error item; when the first entity type discrimination information is less than or equal to the set frequency threshold, and the problem entity phrase does not appear in the error feature rule table, determine that the problem entity type to which the group of problem entity phrases belongs is the detail item.

[0119] In an optional embodiment, when selecting at least one answer-type entity phrase belonging to an answer-type entity type, the selection module 83 is specifically used to: combine question-type entity phrases belonging to different question-type entity types to obtain multiple entity phrase combinations, and count the co-occurrence frequencies of each entity phrase combination in the original after-sales question and the after-sales knowledge base; select at least one standard solution from the standard solutions in the after-sales knowledge base as at least one answer-type entity phrase based on the co-occurrence frequencies of each entity phrase combination in the original after-sales question and the after-sales knowledge base.

[0120] Further optionally, when the selection module 83 selects at least one standard solution from the standard solutions in the after-sales knowledge base as at least one answer-type entity phrase, it is specifically used to: for each entity phrase combination, if the co-occurrence frequency of the entity phrase combination in the original after-sales question is greater than a first co-occurrence frequency threshold and the co-occurrence frequency in the after-sales knowledge base is greater than a second co-occurrence frequency threshold, then the standard solution in the after-sales knowledge base that has co-occurred with the entity phrase combination is used as the answer-type entity phrase corresponding to the entity phrase combination.

[0121] In an optional embodiment, when the generation module 84 obtains the knowledge graph for cloud computing after-sales service, it is specifically used to: for each entity phrase combination corresponding to an answer-type entity phrase, add edges between each question-type entity phrase in the entity phrase combination according to the edge relationship between question-type entity types defined in the knowledge graph relationship graph; add edges between each question-type entity phrase in the entity phrase combination and the answer-type entity phrase corresponding to the entity phrase combination according to the edge relationship between question-type entity types and answer-type entity types defined in the knowledge graph relationship graph, so as to obtain the knowledge graph for cloud computing after-sales service.

[0122] Fig.10 The following is a schematic diagram of the structure of an electronic device provided by an exemplary embodiment of the present application. Fig.10 As shown, the electronic device includes: a storage 1001 , a processor 1002 and a communication component 1003 .

[0123] The memory 1001 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions, messages, pictures, videos, etc. for any application or method operating on the electronic device.

[0124] Memory 1001 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0125] Processor 1002 is coupled to memory 1001 and is used to execute a computer program in memory 1001, so as to: mine entity phrases for original after-sales questions corresponding to cloud computing services and standard after-sales questions in an after-sales knowledge base to obtain multiple question-type entity phrases; determine the question-type entity types to which the multiple question-type entity phrases belong based on multiple question-type entity types defined in a pre-generated knowledge graph relationship graph and in combination with attribute information of multiple question-type entity phrases; select at least one answer-type entity phrase belonging to the answer-type entity type from standard solutions in the after-sales knowledge base according to the answer-type entity types defined in the knowledge graph relationship graph; add edges between multiple question-type entity phrases and between multiple question-type entity phrases and at least one answer-type entity phrase according to edge relationships between question-type entity types and between question-type entity types and answer-type entity types defined in the knowledge graph relationship graph to obtain a knowledge graph for cloud computing after-sales service.

[0126] In an optional embodiment, when determining the question-type entity types to which multiple question-type entity phrases belong, the processor 1002 is specifically used to: divide the question-type entity phrases expressing the same or similar meanings into a group according to the meanings expressed by the multiple question-type entity phrases, so as to obtain multiple groups of question-type entity phrases; determine the entity type discrimination information corresponding to each group of question-type entity phrases according to the frequency of occurrence of each group of question-type entity phrases in standard after-sales problems and standard solutions in the after-sales knowledge base; determine the question-type entity type to which each group of question-type entity phrases belongs according to the entity type discrimination information corresponding to each group of question-type entity phrases.

[0127] Further optionally, when the processor 1002 obtains multiple groups of question-type entity phrases, it is specifically used to: input the multiple question-type entity phrases into the language representation model for vectorization processing to obtain vector expressions of the multiple question-type entity phrases; based on the vector expressions of the multiple question-type entity phrases, perform multiple rounds of density clustering on the multiple question-type entity phrases to obtain multiple groups of question-type entity phrases.

[0128] Further optionally, when determining the entity type discrimination information corresponding to each group of problem-type entity phrases, the processor 1002 is specifically used to: for each group of problem-type entity phrases, use the frequency of occurrence of the group of problem-type entity phrases in standard after-sales questions as the first entity type discrimination information, and calculate the ratio of the frequency of occurrence of the group of problem-type entity phrases in the standard after-sales questions and the frequency of occurrence in the standard solutions as the second entity type discrimination information.

[0129] Further optionally, the above-mentioned multiple question-type entity types include: question-expression-type entity types and question-description-type entity types. Based on this, when the processor 1002 determines the entity type discrimination information corresponding to each group of question-type entity phrases, it is specifically used to: for each group of question-type entity phrases, when the first entity type discrimination information is greater than a set frequency threshold, and the second entity type discrimination information is greater than a set ratio threshold, determine that the group of question-type entity phrases belongs to the question-expression-type entity type; when the first entity type discrimination information is less than or equal to the set frequency threshold, or the second entity type discrimination information is less than or equal to the set ratio threshold, determine that the group of question-type entity phrases belongs to the question-description-type entity type.

[0130] Furthermore, the problem expression type entity type at least includes problem phenomena, which are entity types used to describe key phenomena manifested by after-sales problems. Based on this, when the processor 1002 determines that the group of problem entity phrases belongs to the problem expression type entity type, it is specifically used to: when the first entity type discrimination information is greater than the set frequency threshold, and the second entity type discrimination information is greater than the set ratio threshold, determine that the problem entity type to which the group of problem entity phrases belong is a problem phenomenon.

[0131] Furthermore, the entity type of the problem description type includes at least: a detail item, an error item, and a system version; the detail item is an entity type used to describe the detail information of the after-sales problem, the error item is an entity type used to describe the error information that appears in the after-sales problem, and the system version is an entity type used to describe the operating system information involved in the after-sales problem. Based on this, when the processor 1002 determines that the group of problem-type entity phrases belongs to the problem expression type entity type, it is specifically used to: when the first entity type discrimination information is greater than the set frequency threshold, and the second entity type discrimination information is less than or equal to the set ratio threshold, determine that the problem description type entity type to which the group of problem-type entity phrases belong is the system version; when the first entity type discrimination information is less than or equal to the set frequency threshold, and the problem-type entity phrase appears in the error feature rule table, determine that the problem expression type entity type to which the group of problem-type entity phrases belong is the error item; when the first entity type discrimination information is less than or equal to the set frequency threshold, and the problem-type entity phrase does not appear in the error feature rule table, determine that the problem-type entity type to which the group of problem-type entity phrases belong is the detail item.

[0132] In an optional embodiment, when selecting at least one answer-type entity phrase belonging to an answer-type entity type, the processor 1002 is specifically used to: combine question-type entity phrases belonging to different question-type entity types to obtain multiple entity phrase combinations, and count the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base; select at least one standard solution from the standard solutions in the after-sales knowledge base according to the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base, as at least one answer-type entity phrase.

[0133] Further optionally, when the processor 1002 selects at least one standard solution from the standard solutions in the after-sales knowledge base as at least one answer-type entity phrase, it is specifically used to: for each entity phrase combination, if the co-occurrence frequency of the entity phrase combination in the original after-sales question is greater than a first co-occurrence frequency threshold and the co-occurrence frequency in the after-sales knowledge base is greater than a second co-occurrence frequency threshold, then the standard solution in the after-sales knowledge base that has co-occurred with the entity phrase combination is used as the answer-type entity phrase corresponding to the entity phrase combination.

[0134] In an optional embodiment, when the processor 1002 obtains the knowledge graph for cloud computing after-sales service, it is specifically used to: for each entity phrase combination corresponding to an answer-type entity phrase, add edges between each question-type entity phrase in the entity phrase combination according to the edge relationship between question-type entity types defined in the knowledge graph relationship graph; add edges between each question-type entity phrase in the entity phrase combination and the answer-type entity phrase corresponding to the entity phrase combination according to the edge relationship between question-type entity types and answer-type entity types defined in the knowledge graph relationship graph, so as to obtain the knowledge graph for cloud computing after-sales service.

[0135] Further, if Fig.10 As shown, the electronic device also includes other components such as a display 1004 , a power component 1005 , and an audio component 1006 . Fig.10 Only some components are shown schematically, which does not mean that the electronic device only includes Fig.10 In addition, Fig.10 The components in the dashed box are optional components, not mandatory components, and the specific components depend on the product form of the electronic device. The electronic device of this embodiment can be implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, or an IOT device, or a server device such as a conventional server, a cloud server, or a server array. If the electronic device of this embodiment is implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, etc., it can include Fig.10 If the electronic device of this embodiment is implemented as a server device such as a conventional server, a cloud server or a server array, it may not include Fig.10 Components within the dashed box.

[0136] It should be noted that the embodiment of the present application also provides another electronic device, which is Fig.10 The structures of the electronic devices shown are the same or similar, so they are not shown in the figure again. The electronic device of this embodiment includes at least a memory and a processor, and the processor executes a computer program stored in the processor to: generate a plurality of question-type entity types according to the analysis results of after-sales problems of cloud computing services, and different question-type entity types are used to describe different types of after-sales problems or describe after-sales problems from different dimensions; and generate an answer-type entity type, and the answer-type entity type is used to describe solutions to after-sales problems; determine the edge relationships between the plurality of question-type entity types and between the plurality of question-type entity types and the answer-type entity types according to the association relationship between the after-sales problems described by the plurality of question-type entity types and the demand information for solutions of the after-sales problems; generate a knowledge graph relationship diagram according to the plurality of question-type entity types, the answer-type entity types and the edge relationships, and the knowledge graph relationship diagram is used to generate the knowledge graph required for cloud computing after-sales services.

[0137] In an optional embodiment, the processor is further used to: obtain cloud computing services related to after-sales problems described by the multiple question-type entity types, and use the cloud computing services as attribute information corresponding to the question-type entity type and / or the answer-type entity type.

[0138] In an optional embodiment, the processor is specifically used to: determine the existence of edge relationships between problem phenomena and diagnosis items, between error items and problem phenomena, between fuzzy problems and problem phenomena, and between fuzzy problems and error items based on the association relationships between after-sales problems described by the multiple problem-type entity types; determine the existence of edge relationships between problem phenomena and solutions, between detail items and solutions, between system versions and solutions, between diagnosis items and solutions, and between error items and solutions based on the demand information for solutions for after-sales problems described by the multiple problem-type entity types; wherein the solution is the name of the answer-type entity type.

[0139] Accordingly, the embodiment of the present application further provides a computer-readable storage medium storing a computer program, which can achieve the above-mentioned Figure 1 The steps in the method embodiment are shown.

[0140] Accordingly, the embodiment of the present application further provides a computer-readable storage medium storing a computer program, which can achieve the above-mentioned Figure 4-Figure 7 The steps in the method embodiment are shown.

[0141] Above Fig.10 The communication component in is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0142] Above Fig.10The display in the embodiment includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0143] Above Fig.10 The power supply component in the device provides power to various components of the device where the power supply component is located. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply component is located.

[0144] Above Fig.10 The audio component in can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC), and when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting an audio signal.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0147] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0149] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0150] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0151] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0152] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0153] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for constructing a knowledge graph for cloud computing, characterized in that: include: Mining entity phrases for original after-sales questions corresponding to the cloud computing service and standard after-sales questions in the after-sales knowledge base to obtain a plurality of question-type entity phrases, wherein attribute information of the plurality of question-type entity phrases includes the cloud computing service; Based on the multiple question-type entity types defined in the pre-generated knowledge graph relationship diagram, combined with the attribute information of the multiple question-type entity phrases, determine the question-type entity types to which the multiple question-type entity phrases belong, different question-type entity types are used to describe different types of after-sales problems or describe after-sales problems from different dimensions, and the attribute information of the question-type entity type includes the cloud computing services involved in the after-sales problems described by the question-type entity type; According to the answer-type entity type defined in the knowledge graph relationship diagram, selecting at least one answer-type entity phrase belonging to the answer-type entity type from the standard solutions in the after-sales knowledge base; According to the edge relationships between question-type entity types and between question-type entity types and answer-type entity types defined in the knowledge graph relationship graph, edges are added between the multiple question-type entity phrases and between the multiple question-type entity phrases and at least one answer-type entity phrase to obtain a knowledge graph for cloud computing after-sales service.

2. The method according to claim 1, characterized in that Based on the multiple question-type entity types defined in the pre-generated knowledge graph relationship graph, combined with the attribute information of the multiple question-type entity phrases, the question-type entity types to which the multiple question-type entity phrases belong are determined, including: According to the meaning expressions of the multiple question-type entity phrases, the question-type entity phrases expressing the same or similar meanings are divided into a group to obtain multiple groups of question-type entity phrases; Determine entity type discrimination information corresponding to each group of question-type entity phrases according to the frequency of occurrence of each group of question-type entity phrases in standard after-sales questions and standard solutions in the after-sales knowledge base; According to the entity type discrimination information corresponding to each group of question-type entity phrases, the question-type entity type to which each group of question-type entity phrases belongs is determined.

3. The method according to claim 2, characterized in that According to the meaning expressions of the multiple question-type entity phrases, question-type entity phrases expressing the same or similar meanings are divided into a group to obtain multiple groups of question-type entity phrases, including: Inputting the multiple question-type entity phrases into a language representation model for vectorization processing to obtain vector expressions of the multiple question-type entity phrases; According to the vector expressions of the multiple question-type entity phrases, multiple rounds of density clustering are performed on the multiple question-type entity phrases to obtain multiple groups of question-type entity phrases.

4. The method according to claim 2, characterized in that: According to the frequency of occurrence of each group of question-type entity phrases in standard after-sales questions and standard solutions in the after-sales knowledge base, the entity type discrimination information corresponding to each group of question-type entity phrases is determined, including: For each group of problem-type entity phrases, the frequency of occurrence of this group of problem-type entity phrases in standard after-sales questions is used as the first entity type discrimination information, and the ratio of the frequency of occurrence of this group of problem-type entity phrases in standard after-sales questions and the frequency of occurrence in standard solutions is calculated as the second entity type discrimination information.

5. The method according to claim 4, characterized in that The multiple question-type entity types include: question-expression-type entity types and question-description-type entity types; determining the question-type entity type to which each group of question-type entity phrases belongs according to the entity type discrimination information corresponding to each group of question-type entity phrases, including: For each group of question-type entity phrases, when the first entity type discrimination information is greater than a set frequency threshold, and the second entity type discrimination information is greater than a set ratio threshold, it is determined that the group of question-type entity phrases belongs to a question expression type entity type; When the first entity type discrimination information is less than or equal to a set frequency threshold, or the second entity type discrimination information is less than or equal to a set ratio threshold, it is determined that the group of question-type entity phrases belongs to a question-description-type entity type.

6. The method according to claim 2, characterized in that According to the answer-type entity type defined in the knowledge graph relationship diagram, at least one answer-type entity phrase belonging to the answer-type entity type is selected from the standard solutions in the after-sales knowledge base, including: Combine question-type entity phrases belonging to different question-type entity types to obtain multiple entity phrase combinations, and count the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base; According to the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base, at least one standard solution is selected from the standard solutions in the after-sales knowledge base as the at least one answer-type entity phrase.

7. The method according to claim 6, characterized in that According to the co-occurrence frequency of each entity phrase combination in the original after-sales question and the after-sales knowledge base, at least one standard solution is selected from the standard solutions in the after-sales knowledge base as the at least one answer-type entity phrase, including: For each entity phrase combination, if the co-occurrence frequency of the entity phrase combination in the original after-sales question is greater than the first co-occurrence frequency threshold and the co-occurrence frequency in the after-sales knowledge base is greater than the second co-occurrence frequency threshold, the standard solution in the after-sales knowledge base that has co-occurred with the entity phrase combination is used as the answer-type entity phrase corresponding to the entity phrase combination.

8. The method according to claim 7, characterized in that According to the edge relationships between question-type entity types and between question-type entity types and answer-type entity types defined in the knowledge graph relationship graph, edges are added between the multiple question-type entity phrases and between the multiple question-type entity phrases and the at least one answer-type entity phrase, to obtain a knowledge graph for cloud computing after-sales service, including: For each entity phrase combination corresponding to an answer-type entity phrase, edges are added between each question-type entity phrase in the entity phrase combination according to the edge relationship between question-type entity types defined in the knowledge graph relationship graph; edges are added between each question-type entity phrase in the entity phrase combination and the answer-type entity phrase corresponding to the entity phrase combination according to the edge relationship between question-type entity types and answer-type entity types defined in the knowledge graph relationship graph, so as to obtain a knowledge graph for cloud computing after-sales service.

9. A method for generating a knowledge graph relationship diagram for cloud computing, characterized in that: include: Based on the analysis results of after-sales problems of cloud computing services, multiple problem-type entity types are generated. Different problem-type entity types are used to describe different types of after-sales problems or describe after-sales problems from different dimensions. Generate an answer-type entity type, where the answer-type entity type is used to describe a solution to an after-sales problem; Determine the edge relationships between the multiple question-type entity types and between the multiple question-type entity types and the answer-type entity types according to the association relationships between the after-sales problems described by the multiple question-type entity types and the demand information for solutions of the after-sales problems; A knowledge graph relationship graph is generated based on the multiple question-type entity types, the answer-type entity types, and the edge relationships, and the knowledge graph relationship graph is used to generate a knowledge graph required for cloud computing after-sales service.

10. The method according to claim 9, characterized in that When generating a knowledge graph relationship diagram, it also includes: The cloud computing services involved in the after-sales problems described by the multiple question-type entity types are obtained, and the cloud computing services are used as attribute information corresponding to the question-type entity type and / or the answer-type entity type.

11. The method according to claim 9 or 10, characterized in that: The multiple problem-type entity types include: at least one problem expression type entity type and at least one problem description type entity type; each problem expression type entity type is used to describe a type of after-sales problem, and each problem description type entity type is used to describe detailed information of the after-sales problem from one dimension; The at least one problem expression type entity type includes: problem phenomenon and fuzzy problem; the problem phenomenon is an entity type used to describe the key phenomenon exhibited by the after-sales problem, and the fuzzy problem is an entity type used to describe an incomplete or ambiguous after-sales problem; The at least one problem description-type entity type includes: detail item, error item, diagnosis item and system version; the detail item is an entity type used to describe the detail information of the after-sales problem, the error item is an entity type used to describe the error information occurring in the after-sales problem, the diagnosis item is an entity type used to describe the diagnosis of the after-sales problem and obtain the diagnostic information, and the system version is an entity type used to describe the operating system or database information involved in the after-sales problem.

12. The method according to claim 11, characterized in that According to the association relationship between the after-sales problems described by the multiple question-type entity types and the demand information of the after-sales problems for solutions, the edge relationships between the multiple question-type entity types and between the multiple question-type entity types and the answer-type entity types are determined, including: According to the association relationship between the after-sales problems described by the multiple problem-type entity types, determine the edge relationship between the problem phenomenon and the diagnosis item, between the error item and the problem phenomenon, between the fuzzy problem and the problem phenomenon, and between the fuzzy problem and the error item; According to the demand information for solutions to after-sales problems described by the multiple problem-type entity types, it is determined that there are edge relationships between problem phenomena and solutions, between detail items and solutions, between system versions and solutions, between diagnostic items and solutions, and between error items and solutions; wherein the solution is the name of the answer-type entity type.

13. An electronic device, characterized in that: include: Memory and processor; The memory is used for a computer program, and the processor is coupled to the memory and is used to execute the computer program to implement the steps in the method according to any one of claims 1 to 12.

14. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the processor is enabled to implement the steps in the method according to any one of claims 1 to 12.

Images