WO2021189956A1 - 基于知识图谱的智能客服方法、装置、设备及存储介质 - Google Patents

基于知识图谱的智能客服方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2021189956A1
WO2021189956A1 PCT/CN2020/135260 CN2020135260W WO2021189956A1 WO 2021189956 A1 WO2021189956 A1 WO 2021189956A1 CN 2020135260 W CN2020135260 W CN 2020135260W WO 2021189956 A1 WO2021189956 A1 WO 2021189956A1
Authority
WO
WIPO (PCT)
Prior art keywords
node
intention
layer
actual
execution
Prior art date
Application number
PCT/CN2020/135260
Other languages
English (en)
French (fr)
Inventor
褚秋实
张岳江
陈宗阳
王应
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021189956A1 publication Critical patent/WO2021189956A1/zh

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367Ontology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • G06F16/332Query formulation
    • G06F16/3329Natural language query formulation or dialogue systems

Definitions

  • This application relates to the technical field of intelligent customer service, and in particular to an intelligent customer service method, device, equipment and storage medium based on a knowledge graph.
  • Medical insurance provides basic protection for the health of Chinese residents. Buying medical insurance, such as accidental medical insurance, has become a very common way to protect life and property. A large number of residents need to be reimbursed for medical insurance every day, or have medical insurance issues that need to be consulted.
  • the medical insurance customer service is one of the services closest to customers in the medical insurance industry. It is not only related to the company’s business promotion, business expansion, pre-sales and after-sales consulting, etc., but also directly affects the company’s business volume and performance. Therefore, excellent The medical insurance customer service will continue to attract more customers and improve customer service experience. With the development of artificial intelligence technology and natural language technology, more and more companies use robots to provide customer service Q&A.
  • Business-level changes need to be updated to complete the adjustment, and each service function is independent of each other.
  • the same business areas are not integrated and cannot be specific.
  • the embodiments of the present disclosure provide an intelligent customer service method based on a knowledge graph, including:
  • Hierarchical knowledge graph corresponding to the customer service scenario, where the hierarchical knowledge graph includes an intention layer, a service layer, and an entity layer;
  • the embodiment of the present disclosure provides an intelligent customer service device based on a knowledge graph, including:
  • the creation module is used to create a hierarchical knowledge graph corresponding to the customer service scenario, where the hierarchical knowledge graph includes an intention layer, a service layer, and an entity layer;
  • the acquisition module is used to acquire the actual intention of the customer according to the node data in the intention layer;
  • the execution module is used to generate the execution path according to the actual intention and the node data in the service layer;
  • the dialogue module is used to complete the man-machine dialogue according to the execution path and the objects in the entity layer.
  • the embodiments of the present disclosure provide a computer device, including a memory and a processor.
  • the memory stores computer-readable instructions.
  • the processor executes the intelligent customer service based on the knowledge graph as described below. method:
  • Hierarchical knowledge graph corresponding to the customer service scenario, where the hierarchical knowledge graph includes an intention layer, a service layer, and an entity layer;
  • the embodiments of the present disclosure provide a storage medium storing computer-readable instructions.
  • the computer-readable instructions When the computer-readable instructions are executed by one or more processors, the one or more processors execute the intelligent customer service based on the knowledge graph as described below. method:
  • Hierarchical knowledge graph corresponding to the customer service scenario, where the hierarchical knowledge graph includes an intention layer, a service layer, and an entity layer;
  • Fig. 1 is an implementation environment diagram of an intelligent customer service method based on a knowledge graph according to an exemplary embodiment
  • Fig. 2 is a diagram showing the internal structure of a computer device according to an exemplary embodiment
  • Fig. 3 is a schematic flow chart showing a method for intelligent customer service based on a knowledge graph according to an exemplary embodiment
  • Fig. 4 is a schematic structural diagram showing an intention layer according to an exemplary embodiment
  • Fig. 5 is a schematic structural diagram showing a physical layer according to an exemplary embodiment
  • Fig. 6 is a schematic structural diagram showing a knowledge graph according to an exemplary embodiment
  • Fig. 7 is a schematic diagram showing an execution path according to an exemplary embodiment
  • Fig. 8 is a schematic diagram showing an execution path according to an exemplary embodiment
  • Fig. 9 is a schematic structural diagram of an intelligent customer service device based on a knowledge graph according to an exemplary embodiment.
  • first, second, etc. used in this application can be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from another element.
  • first field and algorithm determination module can be made into the second field and algorithm determination module, and similarly, the second field and algorithm determination module can be made the first field. And algorithm determination module.
  • Fig. 1 is an implementation environment diagram of an intelligent customer service method based on a knowledge graph according to an exemplary embodiment. As shown in Fig. 1, the implementation environment includes a server 110 and a terminal 120.
  • the server 110 is an intelligent customer service device based on a knowledge graph, for example, a computer equipment such as a computer used by a technician, and an intelligent customer service tool is installed on the server 110.
  • the terminal 120 is installed with an application that requires smart customer service.
  • the technician can send a request to provide smart customer service on the computer device 110.
  • the request carries a request identifier.
  • the computer device 110 receives the request and obtains the computer.
  • the knowledge graph stored in the device 110. Then use the knowledge graph to drive the dialogue management engine platform to complete the man-machine dialogue.
  • the terminal 120 and the computer device 110 may be smart phones, tablet computers, notebook computers, desktop computers, etc., but are not limited thereto.
  • the computer device 110 and the terminal 120 can be connected via Bluetooth, USB (Universal Serial Bus) or other communication connection methods, which is not limited in this application.
  • USB Universal Serial Bus
  • Fig. 2 is a diagram showing the internal structure of a computer device according to an exemplary embodiment.
  • the computer device includes a processor, a non-volatile storage medium, a memory, and a network interface connected through a system bus.
  • the non-volatile storage medium of the computer device stores an operating system, a database, and computer-readable instructions.
  • the database may store control information sequences.
  • the processor can realize a An intelligent customer service method based on knowledge graph.
  • the processor of the computer equipment is used to provide calculation and control capabilities, and supports the operation of the entire computer equipment.
  • Computer readable instructions may be stored in the memory of the computer device, and when the computer readable instructions are executed by the processor, the processor may execute an intelligent customer service method based on the knowledge graph.
  • the network interface of the computer device is used to connect and communicate with the terminal.
  • the following describes in detail the intelligent customer service method based on the knowledge graph provided by the embodiments of the present application in conjunction with accompanying drawings 3 to 6.
  • the method can be implemented by relying on a computer program, and can be run on a data transmission device based on the von Neumann system.
  • the computer program can be integrated in the application or run as an independent tool application.
  • FIG. 3 provides a schematic flow chart of an intelligent customer service method based on a knowledge graph for this embodiment of the application.
  • the method of the embodiment of this application may include the following steps:
  • the knowledge graph is a series of different graphs that show the development process and structural relations of knowledge. It uses visualization technology to describe knowledge resources and their carriers, analyzes, draws and displays knowledge and their interrelationships.
  • a layered modeling method is used to create a knowledge graph corresponding to the customer service scenario, including the intention layer, the service layer, and the entity layer.
  • the intention layer adopts a tree structure.
  • the intention layer can adopt a binary tree structure, a quad tree structure, a red-black tree structure, a B-tree structure, an AVL tree structure, etc., and the embodiments of the present disclosure do not specifically limit it.
  • the personnel can choose by themselves according to the actual situation.
  • the intent layer includes abstract intent nodes and actual intent nodes.
  • the upper nodes are abstract intent nodes
  • the bottom nodes are discrete actual intent nodes
  • the relationship is a containment relationship from top to bottom.
  • the abstract intent nodes can include one or more actual intent nodes.
  • one of the abstract intent nodes is "check the validity period of the insurance policy”
  • the actual intent nodes included are "check the finite period of the auto insurance policy” and "check the validity period of the personal insurance policy”.
  • the discrete actual intent nodes at the bottom layer are configured with path identifiers.
  • the actual intent node "check the validity period of auto insurance policy” is configured with an independent path identifier routerKey: autopolicydatequery, and is connected to the nodes of the service layer.
  • Fig. 4 is a schematic diagram showing the structure of an intention layer according to an exemplary embodiment.
  • the intention layer is a tree structure, from top to bottom is the containment relationship, the first layer is abstract intent nodes, and the second The layer is an abstract intent node, and the bottom is an actual intent node.
  • Some abstract intent nodes include one actual intent node, and some abstract intent nodes include two actual intent nodes.
  • the service layer includes a conversation collection node, a business data interface call node, and a business processing node. Each node completes a specific function.
  • the nodes are interconnected, and the execution sequence is marked between the nodes to form a directed graph structure.
  • a path identifier is configured on the path between the node and the node, which is used to determine the execution path according to the path identifier and the execution sequence.
  • Some nodes in the service layer are connected with nodes in the entity layer to obtain objects in the business domain.
  • the entity layer includes the business entity ontology node and the business entity attribute node. According to the connection relationship between the nodes, the attributes owned by the entity are determined to form the entity attribute triple structure, which expresses the business attribute value between the entity layer and each functional node of the service layer Access, business entity value access.
  • Fig. 5 is a schematic diagram showing the structure of an entity layer according to an exemplary embodiment. As shown in Fig. 5, an entity node is connected to multiple attribute nodes, which indicates that an entity can have multiple attributes.
  • Fig. 6 is a schematic diagram showing a knowledge graph according to an exemplary embodiment.
  • the knowledge graph includes an intention layer, a service layer, and an entity layer.
  • the network nodes of the layer are connected and marked with the order execution relationship of Next.
  • the network nodes of the service layer are marked with the order execution relationship of Next and are configured with path identification.
  • the network node of the service layer is connected with the network node of the entity layer, and the network node of the entity layer is connected.
  • Nodes include entity nodes and attribute nodes, forming a triple structure of entity attributes.
  • S302 Acquire the actual intention of the customer according to the node data in the intention layer.
  • the intent classification model is trained using the node data of the intent layer
  • the dialogue management system obtains the user’s voice information
  • the node data in the intent layer is used as the intent classification target through NLU (NaturalLanguage Understanding)
  • NLU NaturalLanguage Understanding
  • the semantic understanding algorithm trains the intent classification model. Input the obtained user's voice information into the trained intention classification model to obtain the intention classification result.
  • the dialog engine performs a query based on the input intention classification results, and generates a dialog according to the tree structure of the intention layer, from top to bottom, from abstract to concrete, to guide the customer to confirm the actual intention. For example, firstly, a dialogue is generated based on the abstract intent node to obtain the customer's abstract intent, and the dialogue is generated based on the actual intent node contained in the customer's abstract intent node to obtain the actual intent of the customer.
  • the actual intention of the customer can be obtained, and the execution path can be determined according to the actual intention of the customer.
  • S303 Generate an execution path according to actual intentions and node data in the service layer.
  • the corresponding path identifier and the node in the service layer connected to it are determined according to the actual intention, and the execution path is generated according to the path identifier and the execution order of the nodes in the service layer.
  • each actual intent node is connected to the service layer network node, and each actual intent node is configured with an independent path identifier.
  • the intent node to query the validity period of a car insurance policy is configured with routerKey: autopolicydatequery.
  • the specific actual intent is in The sequence relationship between the business service graph nodes is marked with the corresponding routerKey logo, and the engine navigates according to the determined actual intent node, the service layer network node connected to it, the corresponding path identifier, and the connection sequence between the service layer network nodes , Get the execution path.
  • Fig. 7 is a schematic diagram showing an execution path according to an exemplary embodiment. As shown in Fig. 7, the execution path has no branches, and the engine executes in order according to the connection relationship between the service layer network nodes, and the obtained execution path It is ABCD.
  • the execution order of the nodes in the service layer when it also includes determining how many branches to execute according to the branch execution strategy preset at the branch, and calculating the branch aggregation node through the search algorithm based on the graph.
  • the execution reaches the branch aggregation node At the time, according to the branch execution strategy to determine whether it is necessary to continue to execute other branches.
  • the branch execution strategy includes: M branches select N executions, N is greater than or equal to 1 and less than or equal to M, M is the actual number of branches, and the value of N is configured in the relationship. For example, query the validity period of the auto insurance policy. Inquire about policy number query and query through license plate number, this branch can be executed by choosing one of them, so configure routerautopolicydatequery: 1 on the branch path. By configuring branch execution strategies on the path, it can be used to handle the situation where there are branches.
  • the engine calculates the branch and aggregation node according to the graph-based search algorithm, which mainly includes two steps:
  • the convergent node of the branch is obtained.
  • the engine executes the branch to the convergent node, it determines whether other branches need to be executed according to the branch execution strategy. When other branches need to be executed, continue to execute other branches, and when no other branches need to be executed , Continue to execute until there is no follow-up service node.
  • Fig. 8 is a schematic diagram showing an execution path according to an exemplary embodiment.
  • the nodes of the service layer have an execution sequence, and there are branch paths.
  • the branch paths are marked with branch execution strategies, based on the graph It can be seen from the search algorithm that F node and G node are potential sink nodes, and G node is sink node. From the labeled branch execution strategy, it can be seen that there are two branches to be executed from K node to G node. In a possible implementation, first Execute the branch from node K, node A to node G, where there are two branches from node A to node F, and two branches need to be executed according to the branch execution strategy.
  • the branch execution strategy there is a branch path from node K to node G
  • a branch needs to be executed, for example, the KB branch is executed. Therefore, the execution path from node K to node G is K-A-D-E-F-B-G, and there is also a branch path from node G to node J.
  • the branch execution strategy only one branch path needs to be executed. Therefore, the complete branch execution path can be K-A-D-E-F-B-G-H-J.
  • the execution path can be determined according to the actual intention of the customer, and the network nodes in the service layer can be executed sequentially according to the execution path.
  • S304 Complete the man-machine dialogue according to the execution path and the objects in the entity layer.
  • the dialog management engine platform navigates down according to the execution path, path nodes, execution nodes, and the main process calls the processor corresponding to the specific node to check whether there is a dialog output, and if there is a dialog output, the dialog is output, human-machine communication is performed, and the current The execution position of the node, after the dialogue is over, continue to execute, if there is no dialogue output, continue to execute.
  • the service layer network node is connected with the network node in the entity layer, and is used to obtain objects in the business domain.
  • the execution fails and the manual customer service call is transferred.
  • the sequence of nodes successfully executed in the current dialogue is recorded in real time for subsequent multiple-intent multiplexing of nodes successfully completed. According to this step, when a new function needs to be added, only a new intention needs to be added, and then a path that can complete the new function can be found in the existing function node network and marked, and new function nodes are added only when necessary.
  • all the node sequences executed after the completion of the conversation are recorded. Through this step, the execution process of the conversation can be traced, and the business function graph modeling can be optimized.
  • the intent layer includes abstract intent nodes and actual intent nodes, and the abstract intent nodes include one or more actual intent nodes.
  • the actual intention node is configured with a path identifier, and the actual intention node is connected to a node in the service layer.
  • obtaining the actual intention of the customer according to the node data in the intention layer may also include: generating a dialogue according to the abstract intent node, obtaining the abstract intent of the customer, and generating a dialogue according to the actual intent node contained in the abstract intent node of the customer, Get the actual intention of the customer.
  • generating the execution path according to the actual intention and the node data in the service layer may also include: determining the corresponding path identifier and the node in the service layer connected to it according to the actual intention, and according to the path identifier and the node in the service layer The order of execution generates an execution path.
  • the execution order of the nodes in the service layer may also include: determining how many branches to execute according to the branch execution strategy preset at the branch, and calculating the branch aggregation nodes through a graph-based search algorithm.
  • determining how many branches to execute according to the branch execution strategy preset at the branch and calculating the branch aggregation nodes through a graph-based search algorithm.
  • a dialogue management engine platform is created for the knowledge graph modeling method.
  • the dialogue management engine platform uses the knowledge graph as a guide to complete the human-machine dialogue.
  • the engine and the knowledge graph are independent. During the application process, only the specific knowledge graph needs to be modified without modification. The engine will take effect.
  • the embodiments of this application can be applied to the field of digital medical insurance, which helps to improve customer experience.
  • FIG. 9 shows a schematic structural diagram of an intelligent customer service device based on a knowledge graph provided by an exemplary embodiment of the present application.
  • the intelligent customer service apparatus based on the knowledge graph may be integrated into the above-mentioned computer device 110, and specifically may include a creation module 901, an acquisition module 902, an execution module 903, and a dialogue module 904.
  • the creation module 901 is used to create a hierarchical knowledge graph corresponding to a customer service scenario, where the hierarchical knowledge graph includes an intention layer, a service layer, and an entity layer.
  • the acquiring module 902 is used to acquire the actual intention of the customer according to the node data in the intention layer.
  • the execution module 903 is used to generate an execution path according to actual intentions and node data in the service layer.
  • the dialogue module 904 is used to complete the man-machine dialogue according to the execution path and the objects in the entity layer.
  • the intent layer includes abstract intent nodes and actual intent nodes, and the abstract intent nodes include one or more actual intent nodes.
  • the actual intention node is configured with a path identifier, and the actual intention node is connected to a node in the service layer.
  • the obtaining module 902 includes:
  • the abstract intent acquisition unit is used to generate a dialogue according to the abstract intent node to obtain the abstract intent of the customer.
  • the actual intention acquisition unit is used to generate a dialogue based on the actual intention node included in the customer's abstract intention node to obtain the actual intention of the customer.
  • the execution module 903 includes:
  • the first execution unit is used to determine the corresponding path identifier and the node in the service layer connected to it according to the actual intention;
  • the second execution unit is configured to generate an execution path according to the path identifier and the execution order of the nodes in the service layer.
  • the execution module 903 may further include:
  • the branch execution unit is used to determine the number of branches to execute according to the branch execution strategy preset at the branch, calculate the branch aggregation node through the search algorithm based on the graph, and when the execution reaches the branch aggregation node, judge whether it needs to continue to execute other branches according to the branch execution strategy Branch.
  • the smart customer service device based on the knowledge graph may further include:
  • the recording module is used to record the sequence of nodes successfully executed in the current dialogue, and to record all the node sequences executed after the dialogue is completed.
  • a dialogue management engine platform is created for the knowledge graph modeling method.
  • the dialogue management engine platform uses the knowledge graph as a guide to complete the human-machine dialogue.
  • the engine and the knowledge graph are independent. During the application process, only the specific knowledge graph needs to be modified without modification. The engine will take effect.
  • the smart customer service device based on the knowledge graph provided in the above embodiment executes the smart customer service method based on the knowledge graph
  • only the division of the above functional modules is used as an example for illustration.
  • the above-mentioned Function allocation is completed by different functional modules, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.
  • the intelligent customer service device based on the knowledge graph provided in the above embodiment and the embodiment of the intelligent customer service method based on the knowledge graph belong to the same concept, and the implementation process is detailed in the method embodiment, which will not be repeated here.
  • a computer device in one embodiment, includes a memory, a processor, and a computer program stored in the memory and running on the processor.
  • the processor executes the computer program, the following steps are implemented: creating and customer service scenarios
  • the corresponding hierarchical knowledge graph where the hierarchical knowledge graph includes the intention layer, the service layer, and the entity layer.
  • the actual intention of the customer is obtained according to the node data in the intention layer, and the execution path is generated according to the actual intention and the node data in the service layer.
  • Complete the man-machine dialogue according to the execution path and the objects in the entity layer.
  • the intent layer includes abstract intent nodes and actual intent nodes, and the abstract intent nodes include one or more actual intent nodes.
  • the actual intention node is configured with a path identifier, and the actual intention node is connected to a node in the service layer.
  • when acquiring the actual intention of the customer based on the node data in the intention layer it includes: generating a dialog according to the abstract intention node, acquiring the abstract intention of the customer, generating a dialog according to the actual intention node contained in the abstract intention node of the customer, Get the actual intention of the customer.
  • the execution path when the execution path is generated according to the actual intention and the node data in the service layer, it includes: determining the corresponding path identifier and the node in the service layer connected to it according to the actual intention, and according to the path identifier and the node in the service layer The order of execution generates an execution path.
  • the execution order of the nodes in the service layer when the execution order of the nodes in the service layer has branches, it further includes: determining how many branches to execute according to the branch execution strategy preset at the branch, and calculating the branch aggregation nodes through a graph-based search algorithm, and when the execution reaches When branching and converging nodes, it is determined whether it is necessary to continue to execute other branches according to the branch execution strategy.
  • the processor further executes the following steps when executing the computer-readable instructions: recording the sequence of nodes successfully executed in the current dialogue, and recording all the node sequences executed after the dialogue is completed.
  • a storage medium storing computer-readable instructions.
  • the storage medium may be volatile or non-volatile.
  • the one or more processors perform the following steps: create a hierarchical knowledge graph corresponding to the customer service scene, where the hierarchical knowledge graph includes an intention layer, a service layer, and The entity layer obtains the actual intention of the customer according to the node data in the intention layer, generates the execution path according to the actual intention and the node data in the service layer, and completes the man-machine dialogue according to the execution path and the objects in the entity layer.
  • the intent layer includes abstract intent nodes and actual intent nodes, and the abstract intent nodes include one or more actual intent nodes.
  • the actual intention node is configured with a path identifier, and the actual intention node is connected to a node in the service layer.
  • when acquiring the actual intention of the customer based on the node data in the intention layer it includes: generating a dialog according to the abstract intention node, acquiring the abstract intention of the customer, generating a dialog according to the actual intention node contained in the abstract intention node of the customer, Get the actual intention of the customer.
  • the execution path when the execution path is generated according to the actual intention and the node data in the service layer, it includes: determining the corresponding path identifier and the node in the service layer connected to it according to the actual intention, and according to the path identifier and the node in the service layer The order of execution generates an execution path.
  • the execution order of the nodes in the service layer when the execution order of the nodes in the service layer has branches, it further includes: determining how many branches to execute according to the branch execution strategy preset at the branch, and calculating the branch aggregation nodes through a graph-based search algorithm, and when the execution reaches When branching and converging nodes, it is determined whether it is necessary to continue to execute other branches according to the branch execution strategy.
  • the processor further executes the following steps when executing the computer-readable instructions: recording the sequence of nodes successfully executed in the current dialogue, and recording all the node sequences executed after the dialogue is completed.
  • the computer program can be stored in a computer readable storage medium. When executed, it may include the procedures of the above-mentioned method embodiments.
  • the aforementioned storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disc, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Animal Behavior & Ethology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种基于知识图谱的智能客服方法、装置、设备及存储介质,所述方法包括:创建与客服场景对应的层次化知识图谱,其中,所述层次化知识图谱包括意图层、服务层以及实体层(S301);根据所述意图层中的节点数据获取客户的实际意图(S302);根据所述实际意图以及服务层中的节点数据生成执行路径(S303);根据所述执行路径以及实体层中的对象完成人机对话(S304)。该方法不仅有利于知识沉淀,而且便于进行直观动态的调整,引擎和知识图谱独立,在应用过程中,只需修改具体的知识图谱,不需要修改引擎就能生效,可应用于数字医疗保险领域,有助于提升客户体验。

Description

基于知识图谱的智能客服方法、装置、设备及存储介质
本申请要求于2020年9月18日提交中国专利局、申请号为202010988375.9,发明名称为“基于知识图谱的智能客服方法、装置、设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及智能客服技术领域,特别涉及一种基于知识图谱的智能客服方法、装置、设备及存储介质。
背景技术
医疗保险为我国居民的健康提供基本的保障,买医疗保险比如意外医疗险等已经成为非常普遍的保障生命财产的方式之一。每天有大量的居民需要报销医疗保险,或者有医疗保险方面的问题需要咨询。而医疗保险客服是医疗保险行业中最接近客户的服务之一,不但关系到公司业务的宣传、业务的拓展、售前售后的咨询等,更直接牵动着公司的业务量和业绩,因此,优秀的医疗保险客服将会不断吸引更多的客户,提升客户的服务体验度。随着人工智能技术和自然语言技术的发展,越来越多的公司采用机器人提供客服问答服务。
发明人意识到,目前智能客服系统存在流程固化,无法可视化的进行动态调整的问题,业务层面变更需要更新版本才能完成调整,而且各个服务功能彼此独立,相同业务领域没有整合在一起,无法进行具体业务领域的知识沉淀,因此,现有技术无法为智能客服系统提供更好的解决方案。
技术解决方案
本公开实施例提供了一种基于知识图谱的智能客服方法,包括:
创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层;
根据意图层中的节点数据获取客户的实际意图;
根据实际意图以及服务层中的节点数据生成执行路径;
根据执行路径以及实体层中的对象完成人机对话。
本公开实施例提供了一种基于知识图谱的智能客服装置,包括:
创建模块,用于创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层;
获取模块,用于根据意图层中的节点数据获取客户的实际意图;
执行模块,用于根据实际意图以及服务层中的节点数据生成执行路径;
对话模块,用于根据执行路径以及实体层中的对象完成人机对话。
本公开实施例提供了一种计算机设备,包括存储器和处理器,存储器中存储有计算机可读指令,计算机可读指令被处理器执行时,使得处理器执行如下所述的基于知识图谱的智能客服方法:
创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层;
根据意图层中的节点数据获取客户的实际意图;
根据实际意图以及服务层中的节点数据生成执行路径;
根据执行路径以及实体层中的对象完成人机对话。
本公开实施例提供了一种存储有计算机可读指令的存储介质,计算机可读指令被一个或多个处理器执行时,使得一个或多个处理器执行如下所述的基于知识图谱的智能客服方法:
创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层;
根据意图层中的节点数据获取客户的实际意图;
根据实际意图以及服务层中的节点数据生成执行路径;
根据执行路径以及实体层中的对象完成人机对话。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分,示出了符合本申请的实施例,并与说明书一起用于解释本申请的原理。
图1是根据一示例性实施例示出的一种基于知识图谱的智能客服方法的实施环境图;
图2是根据一示例性实施例示出的一种计算机设备的内部结构图;
图3是根据一示例性实施例示出的一种基于知识图谱的智能客服方法的流程示意图;
图4是根据一示例性实施例示出的一种意图层的结构示意图;
图5是根据一示例性实施例示出的一种实体层的结构示意图;
图6是根据一示例性实施例示出的一种知识图谱的结构示意图;
图7是根据一示例性实施例示出的一种执行路径的示意图;
图8是根据一示例性实施例示出的一种执行路径的示意图;
图9是根据一示例性实施例示出的一种基于知识图谱的智能客服装置的结构示意图。
本发明的实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处所描述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
可以理解,本申请所使用的术语“第一”、“第二”等可在本文中用于描述各种元件,但这些元件不受这些术语限制。这些术语仅用于将第一个元件与另一个元件区分。举例来说,在不脱离本申请的范围的情况下,可以将第一字段及算法确定模块成为第二字段及算法确定模块,且类似地,可将第二字段及算法确定模块成为第一字段及算法确定模块。
图1是根据一示例性实施例示出的一种基于知识图谱的智能客服方法的实施环境图,如图1所示,在该实施环境中,包括服务器110以及终端120。
服务器110为基于知识图谱的智能客服设备,例如为技术人员使用的电脑等计算机设备,服务器110上安装有智能客服工具。终端120上安装有需要进行智能客服的应用,当需要提供智能客服时,技术人员可以在计算机设备110发出提供智能客服的请求,该请求中携带有请求标识,计算机设备110接收该请求,获取计算机设备110中存储的知识图谱。然后利用知识图谱驱动对话管理引擎平台完成人机对话。
需要说明的是,终端120以及计算机设备110可为智能手机、平板电脑、笔记本电脑、台式计算机等,但并不局限于此。计算机设备110以及终端120可以通过蓝牙、USB(UniversalSerialBus,通用串行总线)或者其他通讯连接方式进行连接,本申请在此不做限制。
图2是根据一示例性实施例示出的一种计算机设备的内部结构图。如图2所示,该计算机设备包括通过系统总线连接的处理器、非易失性存储介质、存储器和网络接口。其中,该计算机设备的非易失性存储介质存储有操作系统、数据库和计算机可读指令,数据库中可存储有控件信息序列,该计算机可读指令被处理器执行时,可使得处理器实现一种基于知识图谱的智能客服方法。该计算机设备的处理器用于提供计算和控制能力,支撑整个计算机设备的运行。该计算机设备的存储器中可存储有计算机可读指令,该计算机可读指令被处理器执行时,可使得处理器执行一种基于知识图谱的智能客服方法。该计算机设备的网络接口用于与终端连接通信。本领域技术人员可以理解,图2中示出的结构,仅仅是与本申请方案相关的部分结构的框图,并不构成对本申请方案所应用于其上的计算机设备的限定,具体的计算机设备可以包括比图中所示更多或更少的部件,或者组合某些部件,或者具有不同的部件布置。
下面将结合附图3-附图6,对本申请实施例提供的基于知识图谱的智能客服方法进行详细介绍。该方法可依赖于计算机程序实现,可运行于基于冯诺依曼体系的数据传输装置上。该计算机程序可集成在应用中,也可作为独立的工具类应用运行。
请参见图3,为本申请实施例提供了一种基于知识图谱的智能客服方法的流程示意图,如图3所示,本申请实施例的方法可以包括以下步骤:
S301,创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层。
其中,知识图谱是显示知识发展进程与结构关系的一系列各种不同的图形,用可视化技术描述知识资源及其载体,分析、绘制和显示知识以及它们之间的相互联系。
针对客服领域的特点,采用分层建模的方式创建与客服场景对应的知识图谱,包括意图层、服务层以及实体层。
其中,意图层采用树形结构,例如,意图层可以采用二叉树结构、四叉树结构、红黑树结构、B树结构、AVL树结构等等,本公开实施例不做具体限制,本领域技术人员可根据实际情况自行选定。
意图层包括抽象意图节点以及实际意图节点,上层节点为抽象意图节点,底层节点为离散的实际意图节点,关系为从上到下的包含关系,抽象意图节点可以包含一个或多个实际意图节点。在一些示例性场景中,其中一个抽象意图节点为“查询保单的有效期”,包含的实际意图节点有“查询车险保单的有限期”以及“查询人身保单的有效期”。
底层离散的实际意图节点上配置有路径标识,例如实际意图节点“查询车险保单的有效期”上配置了独立的路径标识routerKey:autopolicydatequery,且与服务层的节点相连。
图4是根据一示例性实施例示出的一种意图层的结构示意图,如图4所示,意图层为树形结构,从上到下为包含关系,第一层为抽象意图节点,第二层为抽象意图节点,底层为实际意图节点,其中,有的抽象意图节点包含一个实际意图节点,有的抽象意图节点包含两个实际意图节点。
服务层包括对话采集节点、业务数据接口调用节点、业务处理节点,每个节点单一完成一个具体的功能,节点之间互联互通,节点之间标注有执行顺序,形成有向图的结构,且节点与节点之间的路径上配置有路径标识,用于根据路径标识以及执行顺序,确定执行路径。
服务层中的部分节点与实体层中的节点相连,用于获取业务领域的对象。
实体层包括业务实体本体节点以及业务实体属性节点,根据节点之间的连接关系,确定实体拥有的属性,形成实体属性三元组结构,表述实体层与服务层各个功能节点之间的业务属性值存取,业务实体值存取。
图5是根据一示例性实施例示出的一种实体层的结构示意图,如图5所示,一个实体节点与多个属性节点相连,表述一个实体可以拥有多个属性。
图6是根据一示例性实施例示出的一种知识图谱的示意图,如图6所示,知识图谱包括意图层、服务层以及实体层,意图层为树形结构,底部的实际意图节点与服务层的网络节点相连,且标注有Next的顺序执行关系,服务层的网络节点之间标注有Next的顺序执行关系,且配置有路径标识,服务层网络节点与实体层网络节点相连,实体层网络节点包含实体节点和属性节点,组成实体属性三元组结构。
通过建立层次化的知识图谱,不仅有利于沉淀业务知识,还便于进行直观动态的调整。
S302,根据意图层中的节点数据获取客户的实际意图。
具体地,知识图谱构建完成后,利用意图层的节点数据训练意图分类模型,对话管理系统获取用户的语音信息,将意图层中的节点数据作为意图分类目标,通过NLU(NaturalLanguageUnderstanding,自然语义理解)语义理解算法训练意图分类模型。将获取的用户语音信息输入训练好的意图分类模型,得到意图分类结果。
在一种可能的实现方式中,对话引擎根据输入的意图分类结果进行查询,根据意图层的树形结构,从上到下,从抽象到具体,产生对话引导客户确认准确的实际意图。例如,首先根据抽象意图节点产生对话,获取客户的抽象意图,根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
根据该步骤,可以获取客户的实际意图,根据客户的实际意图确定执行路径。
S303,根据实际意图以及服务层中的节点数据生成执行路径。
在一种可能的实现方式中,根据实际意图确定对应的路径标识以及与其相连的服务层中的节点,根据路径标识以及服务层中节点的执行顺序生成执行路径。
具体地,每个实际意图节点都接入到服务层网络节点上,而且每个实际意图节点上都配置独立的路径标识,比如查询车险保单有效期意图节点配置了routerKey: autopolicydatequery,具体一个实际意图在业务服务图谱节点之间的顺序关系上都标注有对应的routerKey标志,引擎根据确定的实际意图节点、与其相连的服务层网络节点、对应的路径标识以及服务层网络节点之间的连接顺序进行导航,得到执行路径。
图7是根据一示例性实施例示出的一种执行路径的示意图,如图7所示,该执行路径没有分支,引擎根据服务层网络节点之间的连接关系顺序执行即可,得到的执行路径为A-B-C-D。
可选地,当服务层中节点的执行顺序有分支时,还包括根据分支处预设的分支执行策略确定执行几条分支,通过基于图的搜索算法计算分支汇聚节点,当执行到分支汇聚节点时,根据分支执行策略判断是否需要继续执行其他分支。
其中,分支执行策略包括:M个分支选择N个执行,N大于等于1且小于等于M,M为实际分支数,在关系上配置N值,例如查询车险保单有效期意图节点有两个分支,通过询问保单号查询和通过车牌号查询,这个分支任选其一就可以往下执行,所以在分支的路径上配置routerautopolicydatequery: 1。通过在路径上配置分支执行策略可以用来处理有分支的情况。
遇到分支时,引擎根据基于图的搜索算法计算分支汇聚节点,主要包括两个步骤:
(1)搜索潜在汇聚节点,基于neo4j图数据库的cypher查询实现,MATCH (a)-[r1:Next*1..]->(x)<-[r2:Next*1..]-(b) returndistinctx,其中a和b表示一个节点下级所有分支节点集合,a,b往下搜索到至少有一个以上Next关系汇聚的x,通过这个查询出分支节点两两的汇聚节点。
(2)剔除局部汇聚节点,根据步骤一筛选的潜在分支节点集合,逐个判断是否每一个分支节点都有路径可以到达该潜在节点,剔除局部汇聚节点,(各个分支节点..)-[r:NEXT{路径标识}*1..]->(待验证汇聚节点x..) returncount(r) ,r为0则只是局部汇聚节点需要剔除。
根据该步骤,得到分支的汇聚节点,当引擎执行分支到汇聚节点时,根据分支执行策略判断是否还需执行其他分支,当需要执行其他分支时,继续执行其他分支,当不需要执行其他分支时,继续往下执行,直到没有后续服务节点。
通过计算分支汇聚节点,可以确保所有分支都执行完成,避免在服务节点分支复杂时漏掉需要执行的分支。
图8是根据一示例性实施例示出的一种执行路径的示意图,如图8所示,服务层的节点之间具有执行顺序,而且存在分支路径,分支路径处标注有分支执行策略,基于图的搜索算法可知,F节点和G节点为潜在汇聚节点,G节点为汇聚节点,由标注的分支执行策略可知,K节点到G节点要执行两条分支,在一种可能的实现方式中,首先执行K节点、A节点到G节点的分支,其中,A节点到F节点存在两条分支,根据分支执行策略需要执行两条分支,因此,根据分支执行策略,K节点到G节点的一条分支路径为K-A-D-E-F,执行到G分支汇聚节点时,根据分支执行策略可知,还需执行一条分支,例如,执行K-B分支。因此K节点到G节点的执行路径为K-A-D-E-F-B-G,G节点到J节点同样存在分支路径,由分支执行策略可知,只需执行一条分支路径即可。因此,完整的分支执行路径可为K-A-D-E-F-B-G-H-J。
根据该步骤,可以根据客户的实际意图确定执行路径,根据执行路径依次执行服务层中的网络节点。
S304,根据执行路径以及实体层中的对象完成人机对话。
具体地,对话管理引擎平台根据执行路径往下导航,途径节点,执行节点,主流程调用具体节点对应的处理器,检查是否有对话输出,若有就输出对话,进行人机交流,并且保留当前节点的执行位置,待对话结束后,继续往下执行,若没有对话输出,则继续往下执行。
其中,服务层网络节点与实体层中的网络节点相连,用于获取业务领域的对象。
在一个实施例中,若遇到节点无法成功往下执行时,则执行失败,转接人工客服电话。
在一个实施例中,在进行人机对话的过程中,实时记录当前对话成功执行的节点序列,用于后续多意图复用成功完成的节点。根据该步骤,当需要增加新功能时,只需要增加新的意图,然后可以在现有功能节点网络中找到能完成新功能的路径标注出来即可,有必要才添加新的功能节点。
在一个实施例中,记录对话完成后执行的所有节点序列,通过该步骤,可以追溯会话执行过程,优化业务功能图谱建模。
在一个实施例中,意图层包括抽象意图节点和实际意图节点,抽象意图节点包含一个或多个实际意图节点。
在一个实施例中,实际意图节点配置有路径标识,实际意图节点与服务层中的节点相连。
在一个实施例中,根据意图层中的节点数据获取客户的实际意图,还可以包括:根据抽象意图节点产生对话,获取客户的抽象意图,根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
在一个实施例中,根据实际意图以及服务层中的节点数据生成执行路径,还可以包括:根据实际意图确定对应的路径标识以及与其相连的服务层中的节点,根据路径标识以及服务层中节点的执行顺序生成执行路径。
在一个实施例中,当服务层中节点的执行顺序有分支时,还可以包括:根据分支处预设的分支执行策略确定执行几条分支,通过基于图的搜索算法计算分支汇聚节点,当执行到分支汇聚节点时,根据分支执行策略判断是否需要继续执行其他分支。
基于本公开实施例提供的基于知识图谱的智能客服方法,使用图谱的形式建模具体业务规则,不仅有利于知识沉淀,而且便于进行直观动态的调整。针对知识图谱建模方式创建了对话管理引擎平台,对话管理引擎平台利用知识图谱作为指导驱动完成人机对话,引擎和知识图谱独立,在应用过程中,只需修改具体的知识图谱,不需要修改引擎就能生效。本申请实施例可应用于数字医疗保险领域,有助于提升客户体验。
下述为本申请装置实施例,可以用于执行本申请方法实施例。对于本申请装置实施例中未披露的细节,请参照本申请方法实施例。
请参见图9,其示出了本申请一个示例性实施例提供的基于知识图谱的智能客服装置的结构示意图。如图9所示,该基于知识图谱的智能客服装置可以集成于上述的计算机设备110中,具体可以包括创建模块901、获取模块902、执行模块903以及对话模块904。
创建模块901,用于创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层。
获取模块902,用于根据意图层中的节点数据获取客户的实际意图。
执行模块903,用于根据实际意图以及服务层中的节点数据生成执行路径。
对话模块904,用于根据执行路径以及实体层中的对象完成人机对话。
在一个实施例中,意图层包括抽象意图节点和实际意图节点,抽象意图节点包含一个或多个实际意图节点。
在一个实施例中,实际意图节点配置有路径标识,实际意图节点与服务层中的节点相连。
在一个实施例中,获取模块902,包括:
抽象意图获取单元,用于根据抽象意图节点产生对话,获取客户的抽象意图。
实际意图获取单元,用于根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
在一个实施例中,执行模块903,包括:
第一执行单元,用于根据实际意图确定对应的路径标识以及与其相连的服务层中的节点;
第二执行单元,用于根据路径标识以及服务层中节点的执行顺序生成执行路径。
在一个实施例中,执行模块903还可以包括:
分支执行单元,用于根据分支处预设的分支执行策略确定执行几条分支,通过基于图的搜索算法计算分支汇聚节点,当执行到分支汇聚节点时,根据分支执行策略判断是否需要继续执行其他分支。
在一个实施例中,基于知识图谱的智能客服装置还可以包括:
记录模块,用于记录当前对话成功执行的节点序列,记录对话完成后执行的所有节点序列。
基于本公开实施例提供的基于知识图谱的智能客服装置,使用图谱的形式建模具体业务规则,不仅有利于知识沉淀,而且便于进行直观动态的调整。针对知识图谱建模方式创建了对话管理引擎平台,对话管理引擎平台利用知识图谱作为指导驱动完成人机对话,引擎和知识图谱独立,在应用过程中,只需修改具体的知识图谱,不需要修改引擎就能生效。
需要说明的是,上述实施例提供的基于知识图谱的智能客服装置在执行基于知识图谱的智能客服方法时,仅以上述各功能模块的划分进行举例说明,实际应用中,可以根据需要而将上述功能分配由不同的功能模块完成,即将设备的内部结构划分成不同的功能模块,以完成以上描述的全部或者部分功能。另外,上述实施例提供的基于知识图谱的智能客服装置与基于知识图谱的智能客服方法实施例属于同一构思,其体现实现过程详见方法实施例,这里不再赘述。
在一个实施例中,提出了一种计算机设备,计算机设备包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序,处理器执行计算机程序时实现以下步骤:创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层,根据意图层中的节点数据获取客户的实际意图,根据实际意图以及服务层中的节点数据生成执行路径,根据执行路径以及实体层中的对象完成人机对话。
在一个实施例中,意图层包括抽象意图节点和实际意图节点,抽象意图节点包含一个或多个实际意图节点。
在一个实施例中,实际意图节点配置有路径标识,实际意图节点与服务层中的节点相连。
在一个实施例中,在根据意图层中的节点数据获取客户的实际意图时,包括:根据抽象意图节点产生对话,获取客户的抽象意图,根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
在一个实施例中,在根据实际意图以及服务层中的节点数据生成执行路径时,包括:根据实际意图确定对应的路径标识以及与其相连的服务层中的节点,根据路径标识以及服务层中节点的执行顺序生成执行路径。
在一个实施例中,当服务层中节点的执行顺序有分支时,还包括:根据分支处预设的分支执行策略确定执行几条分支,通过基于图的搜索算法计算分支汇聚节点,当执行到分支汇聚节点时,根据分支执行策略判断是否需要继续执行其他分支。
在一个实施例中,处理器执行计算机可读指令时还执行以下步骤:记录当前对话成功执行的节点序列,记录对话完成后执行的所有节点序列。
在一个实施例中,提出了一种存储有计算机可读指令的存储介质,所述存储介质可以是易失性的,也可以是非易失性的。该计算机可读指令被一个或多个处理器执行时,使得一个或多个处理器执行以下步骤:创建与客服场景对应的层次化知识图谱,其中,层次化知识图谱包括意图层、服务层以及实体层,根据意图层中的节点数据获取客户的实际意图,根据实际意图以及服务层中的节点数据生成执行路径,根据执行路径以及实体层中的对象完成人机对话。
在一个实施例中,意图层包括抽象意图节点和实际意图节点,抽象意图节点包含一个或多个实际意图节点。
在一个实施例中,实际意图节点配置有路径标识,实际意图节点与服务层中的节点相连。
在一个实施例中,在根据意图层中的节点数据获取客户的实际意图时,包括:根据抽象意图节点产生对话,获取客户的抽象意图,根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
在一个实施例中,在根据实际意图以及服务层中的节点数据生成执行路径时,包括:根据实际意图确定对应的路径标识以及与其相连的服务层中的节点,根据路径标识以及服务层中节点的执行顺序生成执行路径。
在一个实施例中,当服务层中节点的执行顺序有分支时,还包括:根据分支处预设的分支执行策略确定执行几条分支,通过基于图的搜索算法计算分支汇聚节点,当执行到分支汇聚节点时,根据分支执行策略判断是否需要继续执行其他分支。
在一个实施例中,处理器执行计算机可读指令时还执行以下步骤:记录当前对话成功执行的节点序列,记录对话完成后执行的所有节点序列。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,该计算机程序可存储于一计算机可读取存储介质中,该程序在执行时,可包括如上述各方法的实施例的流程。其中,前述的存储介质可为磁碟、光盘、只读存储记忆体(Read-OnlyMemory,ROM)等非易失性存储介质,或随机存储记忆体(RandomAccessMemory,RAM)等。
以上实施例的各技术特征可以进行任意的组合,为使描述简洁,未对上述实施例中的各个技术特征所有可能的组合都进行描述,然而,只要这些技术特征的组合不存在矛盾,都应当认为是本说明书记载的范围。
以上实施例仅表达了本申请的几种实施方式,其描述较为具体和详细,但并不能因此而理解为对本申请专利范围的限制。应当指出的是,对于本领域的普通技术人员来说,在不脱离本申请构思的前提下,还可以做出若干变形和改进,这些都属于本申请的保护范围。因此,本申请专利的保护范围应以所附权利要求为准。

Claims (20)

  1. 一种基于知识图谱的智能客服方法,其中,包括:
    创建与客服场景对应的层次化知识图谱,其中,所述层次化知识图谱包括意图层、服务层以及实体层;
    根据所述意图层中的节点数据获取客户的实际意图;
    根据所述实际意图以及服务层中的节点数据生成执行路径;
    根据所述执行路径以及实体层中的对象完成人机对话。
  2. 根据权利要求1所述的方法,其中,所述意图层包括抽象意图节点和实际意图节点,所述抽象意图节点包含一个或多个实际意图节点。
  3. 根据权利要求2所述的方法,其中,所述实际意图节点配置有路径标识,所述实际意图节点与服务层中的节点相连。
  4. 根据权利要求1所述的方法,其中,根据所述意图层中的节点数据获取客户的实际意图,包括:
    根据抽象意图节点产生对话,获取客户的抽象意图;
    根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
  5. 根据权利要求1所述的方法,其中,根据所述实际意图以及服务层中的节点数据生成执行路径,包括:
    根据所述实际意图确定对应的路径标识以及与其相连的服务层中的节点;
    根据所述路径标识以及所述服务层中节点的执行顺序生成执行路径。
  6. 根据权利要求5所述的方法,其中,当所述服务层中节点的执行顺序有分支时,还包括:
    根据分支处预设的分支执行策略确定执行几条分支;
    通过基于图的搜索算法计算分支汇聚节点;
    当执行到分支汇聚节点时,根据所述分支执行策略判断是否需要继续执行其他分支。
  7. 根据权利要求1所述的方法,其中,还包括:
    记录当前对话成功执行的节点序列;
    记录对话完成后执行的所有节点序列。
  8. 一种基于知识图谱的智能客服装置,其中,包括:
    创建模块,用于创建与客服场景对应的层次化知识图谱,其中,所述层次化知识图谱包括意图层、服务层以及实体层;
    获取模块,用于根据所述意图层中的节点数据获取客户的实际意图;
    执行模块,用于根据所述实际意图以及服务层中的节点数据生成执行路径;
    对话模块,用于根据所述执行路径以及实体层中的对象完成人机对话。
  9. 一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机可读指令,所述计算机可读指令被所述处理器执行时,使得所述处理器执行如下所述的基于知识图谱的智能客服方法:
    创建与客服场景对应的层次化知识图谱,其中,所述层次化知识图谱包括意图层、服务层以及实体层;
    根据所述意图层中的节点数据获取客户的实际意图;
    根据所述实际意图以及服务层中的节点数据生成执行路径;
    根据所述执行路径以及实体层中的对象完成人机对话。
  10. 根据权利要求9所述的计算机设备,其中,所述意图层包括抽象意图节点和实际意图节点,所述抽象意图节点包含一个或多个实际意图节点。
  11. 根据权利要求10所述的计算机设备,其中,所述实际意图节点配置有路径标识,所述实际意图节点与服务层中的节点相连。
  12. 根据权利要求9所述的计算机设备,其中,根据所述意图层中的节点数据获取客户的实际意图,包括:
    根据抽象意图节点产生对话,获取客户的抽象意图;
    根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
  13. 根据权利要求9所述的计算机设备,其中,根据所述实际意图以及服务层中的节点数据生成执行路径,包括:
    根据所述实际意图确定对应的路径标识以及与其相连的服务层中的节点;
    根据所述路径标识以及所述服务层中节点的执行顺序生成执行路径。
  14. 根据权利要求13所述的计算机设备,其中,当所述服务层中节点的执行顺序有分支时,还包括:
    根据分支处预设的分支执行策略确定执行几条分支;
    通过基于图的搜索算法计算分支汇聚节点;
    当执行到分支汇聚节点时,根据所述分支执行策略判断是否需要继续执行其他分支。
  15. 一种存储有计算机可读指令的存储介质,所述计算机可读指令被一个或多个处理器执行时,使得一个或多个处理器执行如下所述的基于知识图谱的智能客服方法:
    创建与客服场景对应的层次化知识图谱,其中,所述层次化知识图谱包括意图层、服务层以及实体层;
    根据所述意图层中的节点数据获取客户的实际意图;
    根据所述实际意图以及服务层中的节点数据生成执行路径;
    根据所述执行路径以及实体层中的对象完成人机对话。
  16. 根据权利要求15所述的存储介质,其中,所述意图层包括抽象意图节点和实际意图节点,所述抽象意图节点包含一个或多个实际意图节点。
  17. 根据权利要求16所述的存储介质,其中,所述实际意图节点配置有路径标识,所述实际意图节点与服务层中的节点相连。
  18. 根据权利要求15所述的存储介质,其中,根据所述意图层中的节点数据获取客户的实际意图,包括:
    根据抽象意图节点产生对话,获取客户的抽象意图;
    根据客户的抽象意图节点包含的实际意图节点产生对话,获取客户的实际意图。
  19. 根据权利要求15所述的存储介质,其中,根据所述实际意图以及服务层中的节点数据生成执行路径,包括:
    根据所述实际意图确定对应的路径标识以及与其相连的服务层中的节点;
    根据所述路径标识以及所述服务层中节点的执行顺序生成执行路径。
  20. 根据权利要求19所述的存储介质,其中,当所述服务层中节点的执行顺序有分支时,还包括:
    根据分支处预设的分支执行策略确定执行几条分支;
    通过基于图的搜索算法计算分支汇聚节点;
    当执行到分支汇聚节点时,根据所述分支执行策略判断是否需要继续执行其他分支。
PCT/CN2020/135260 2020-09-18 2020-12-10 基于知识图谱的智能客服方法、装置、设备及存储介质 WO2021189956A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010988375.9 2020-09-18
CN202010988375.9A CN112115276B (zh) 2020-09-18 2020-09-18 基于知识图谱的智能客服方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2021189956A1 true WO2021189956A1 (zh) 2021-09-30

Family

ID=73800741

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/135260 WO2021189956A1 (zh) 2020-09-18 2020-12-10 基于知识图谱的智能客服方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN112115276B (zh)
WO (1) WO2021189956A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114510582A (zh) * 2022-04-18 2022-05-17 支付宝(杭州)信息技术有限公司 基于知识图谱的信息处理方法及装置
CN114841519A (zh) * 2022-04-08 2022-08-02 北京中电普华信息技术有限公司 电力工作票构建方法和装置
CN115544263A (zh) * 2022-08-30 2022-12-30 中国人民财产保险股份有限公司 基于车辆知识图谱的智能问答方法、系统及设备
CN116662578A (zh) * 2023-08-02 2023-08-29 中国标准化研究院 一种基于端到端的大型知识图谱构建和存储方法和系统

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115018473A (zh) * 2022-08-03 2022-09-06 平安银行股份有限公司 一种业务处理方法、装置、存储介质及设备
CN115982391B (zh) * 2023-03-17 2023-07-25 恒生电子股份有限公司 信息处理方法及装置

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107688614A (zh) * 2017-08-04 2018-02-13 平安科技(深圳)有限公司 意图获取方法、电子装置及计算机可读存储介质
US20180246952A1 (en) * 2015-08-19 2018-08-30 Alibaba Group Holding Limited Method and device for generating online question paths from existing question banks using a knowledge graph
CN110019844A (zh) * 2019-02-20 2019-07-16 众安信息技术服务有限公司 一种保险行业知识图谱问答系统构建方法及装置
CN110825862A (zh) * 2019-11-06 2020-02-21 北京诺道认知医学科技有限公司 基于药学知识图谱的智能问答方法及装置
CN110895561A (zh) * 2019-11-13 2020-03-20 中国科学院自动化研究所 基于多模态知识感知的医疗问答检索方法、系统、装置
CN111414465A (zh) * 2020-03-16 2020-07-14 北京明略软件系统有限公司 基于知识图谱的问答系统中的处理方法和装置
CN111488402A (zh) * 2020-03-26 2020-08-04 天津大学 一种带有层次关系结构知识图谱的表示学习方法
CN111666399A (zh) * 2020-06-23 2020-09-15 中国平安人寿保险股份有限公司 基于知识图谱的智能问答方法、装置和计算机设备

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9183279B2 (en) * 2011-09-22 2015-11-10 International Business Machines Corporation Semantic questioning mechanism to enable analysis of information architectures
CN106649878A (zh) * 2017-01-07 2017-05-10 陈翔宇 基于人工智能的物联网实体搜索方法及系统
CN110019725A (zh) * 2017-12-22 2019-07-16 科沃斯商用机器人有限公司 人机交互方法、系统及其电子设备
US11954613B2 (en) * 2018-02-01 2024-04-09 International Business Machines Corporation Establishing a logical connection between an indirect utterance and a transaction
CN110347810B (zh) * 2019-05-30 2022-08-19 重庆金融资产交易所有限责任公司 对话式检索回答方法、装置、计算机设备及存储介质
CN110941693A (zh) * 2019-10-09 2020-03-31 深圳软通动力信息技术有限公司 基于任务的人机对话方法、系统、电子设备及存储介质
CN111309879A (zh) * 2020-01-20 2020-06-19 北京文思海辉金信软件有限公司 基于知识图谱的人机训练场景构建方法和装置
CN111611407A (zh) * 2020-05-27 2020-09-01 北京一维大成科技有限公司 客服服务交互方法、装置、存储介质及设备
CN111639082B (zh) * 2020-06-08 2022-12-23 成都信息工程大学 基于Ceph的十亿级节点规模知识图谱的对象存储管理方法及系统

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180246952A1 (en) * 2015-08-19 2018-08-30 Alibaba Group Holding Limited Method and device for generating online question paths from existing question banks using a knowledge graph
CN107688614A (zh) * 2017-08-04 2018-02-13 平安科技(深圳)有限公司 意图获取方法、电子装置及计算机可读存储介质
CN110019844A (zh) * 2019-02-20 2019-07-16 众安信息技术服务有限公司 一种保险行业知识图谱问答系统构建方法及装置
CN110825862A (zh) * 2019-11-06 2020-02-21 北京诺道认知医学科技有限公司 基于药学知识图谱的智能问答方法及装置
CN110895561A (zh) * 2019-11-13 2020-03-20 中国科学院自动化研究所 基于多模态知识感知的医疗问答检索方法、系统、装置
CN111414465A (zh) * 2020-03-16 2020-07-14 北京明略软件系统有限公司 基于知识图谱的问答系统中的处理方法和装置
CN111488402A (zh) * 2020-03-26 2020-08-04 天津大学 一种带有层次关系结构知识图谱的表示学习方法
CN111666399A (zh) * 2020-06-23 2020-09-15 中国平安人寿保险股份有限公司 基于知识图谱的智能问答方法、装置和计算机设备

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114841519A (zh) * 2022-04-08 2022-08-02 北京中电普华信息技术有限公司 电力工作票构建方法和装置
CN114510582A (zh) * 2022-04-18 2022-05-17 支付宝(杭州)信息技术有限公司 基于知识图谱的信息处理方法及装置
CN115544263A (zh) * 2022-08-30 2022-12-30 中国人民财产保险股份有限公司 基于车辆知识图谱的智能问答方法、系统及设备
CN116662578A (zh) * 2023-08-02 2023-08-29 中国标准化研究院 一种基于端到端的大型知识图谱构建和存储方法和系统
CN116662578B (zh) * 2023-08-02 2023-10-31 中国标准化研究院 一种基于端到端的大型知识图谱构建和存储方法和系统

Also Published As

Publication number Publication date
CN112115276B (zh) 2024-05-24
CN112115276A (zh) 2020-12-22

Similar Documents

Publication Publication Date Title
WO2021189956A1 (zh) 基于知识图谱的智能客服方法、装置、设备及存储介质
AU2018260855B2 (en) Hybrid cloud migration delay risk prediction engine
US20210004711A1 (en) Cognitive robotic process automation
US10453165B1 (en) Computer vision machine learning model execution service
JP2012234536A (ja) セマンティックモデルおよびルールからコードを生成するための方法
WO2021228264A1 (zh) 一种应用机器学习的方法、装置、电子设备及存储介质
US10839329B2 (en) Process execution using rules framework flexibly incorporating predictive modeling
JPWO2020003585A1 (ja) 仮説推論装置、仮説推論方法及び制御プログラム
CN115269683A (zh) 一种基于多方协作的数据查询方法及装置
CN112965710A (zh) 计算图的处理方法、装置和系统
US20140026112A1 (en) Task management for changes to shared artifacts
CN109840074B (zh) 一种业务生成方法、装置及网络设备
KR20220031522A (ko) 디지털 트윈과 그래프 기반 구조를 가진 컴퓨터 구현 시스템 및 방법
CN104901998B (zh) 一体化云服务监控方法
CN116560626A (zh) 基于自定义规则的数据处理方法、系统、设备和存储介质
CN105786941A (zh) 一种信息挖掘方法和装置
CN110442753A (zh) 一种基于opc ua的图数据库自动建立方法及装置
CN101751332A (zh) 工作流中间件的流程仿真方法、装置及系统
EP3084675B1 (en) Data access control for business processes
US20210208993A1 (en) Cognitive and deep learning-based software component distribution
JP5244750B2 (ja) テストケース生成装置およびその方法
CN112686580A (zh) 一种可自定义流程的工作流定义方法及系统
US11868349B2 (en) Row secure table plan generation
US20210311942A1 (en) Dynamically altering a query access plan
US20130054301A1 (en) Process refinement

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20926485

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20926485

Country of ref document: EP

Kind code of ref document: A1