Customer relationship management system control method and device, and storage medium

Abstract

The invention discloses a control method and device of a customer relationship management system and a storage medium, and belongs to the technical field of management systems. The method comprises the steps of generating an association problem set according to service scene information of at least one predefined service scene, constructing graph structure data based on the association problem set, forming a domain knowledge graph, distributing priority weights for conflict problems in the domain knowledge graph, and integrating the domain knowledge graph into an intention recognition module, and in response to a target service scene triggered by a user, outputting at least one guide problem matched with the target service scene through an intention recognition module on the basis of the domain knowledge graph and the priority weight. According to weight distribution and a context association mechanism of the knowledge graph, the accuracy of a recommendation system in a complex service scene is improved.

Description

Technical Field

[0001] This application relates to the field of management system technology, and in particular to a control method, device and storage medium for a customer relationship management system. Background Technology

[0002] Customer Relationship Management (CRM) systems are typically used to assist banks, businesses, and other users in managing customer relationships to improve business efficiency.

[0003] In related technologies, customer relationship management systems (CRM) typically operate based on certain business functions, and they usually provide users with guiding questions to assist in user interaction with the system. Specifically, business management systems often employ static rule-based agents that recommend guiding questions for relevant operations or information to users based on pre-set instruction sets or historical behavioral data of user groups.

[0004] However, due to the different business needs of different users, the customer relationship management system lacks a deep understanding of specific business scenarios. Consequently, when interacting with users, the system cannot perceive the user's immediate intent in the current work context, resulting in a serious disconnect between the recommended guidance questions and the specific business scenario in which the user is currently located, leading to low recommendation relevance.

[0005] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0006] The main objective of this application is to provide a control method, device, and storage medium for a customer relationship management system, aiming to solve the technical problem of low accuracy in the recommendation of bootstrap questions in customer relationship management systems.

[0007] To achieve the above objectives, this application provides a control method for a customer relationship management system, the method comprising the following steps: Generate a set of related questions based on business scenario information from at least one predefined business scenario. Based on the aforementioned set of related questions, a graph-structured data is constructed to form a domain knowledge graph; Priority weights are assigned to conflict issues in the domain knowledge graph, and the domain knowledge graph is integrated into the intent recognition module; In response to a user triggering a target business scenario, the intent recognition module outputs at least one guiding question that matches the target business scenario, based on the domain knowledge graph and the priority weight.

[0008] In one embodiment, after the step of responding to a user triggering a target business scenario and outputting at least one guiding question matching the target business scenario through the intent recognition module based on the domain knowledge graph and the priority weight, the method further includes: Collect user interaction data on the guided questions and convert the interaction data into feedback signals for model training; Based on the positive and negative feedback signals in the feedback signals, a reward function for the reinforcement learning algorithm is constructed. Based on the reward function, the domain knowledge graph is trained using feedback, and the priority weights are updated.

[0009] In one embodiment, the step of assigning priority weights to conflict issues in the domain knowledge graph and integrating the domain knowledge graph into the intent recognition module includes: Retrieve a set of predefined concerns associated with user role information; Acquire historical operation data and perform semantic analysis on the set of historical questions in the historical operation data. Determine the set of high-frequency questions based on the results of the semantic analysis. Based on the predefined set of concerns and the set of high-frequency questions, the priority of the questions in the domain knowledge graph is calculated to obtain the priority weight of the conflicting questions. The priority weights are associated with the domain knowledge graph, and the domain knowledge graph is integrated into the intent recognition module.

[0010] In one embodiment, the step of generating a set of related questions based on business scenario information of at least one predefined business scenario includes: Obtain the business scenario information of the predefined business scenario, and determine the business function based on the business scenario information; Based on the aforementioned business skills, generate at least one relevant guiding question; The associated guidance questions are associated with the corresponding predefined business scenarios to form the associated question set.

[0011] In one embodiment, the step of constructing graph-structured data and forming a domain knowledge graph based on the set of related questions includes: Extract entity information from the aforementioned business scenario information; Based on the entity information and the associated guiding questions in the associated question set, construct graph nodes; Based on the business logic information in the business scenario information, and the relationship between the business scenario information and the associated guidance question, edges are established between the graph nodes to form the graph structure data; Using graph retrieval enhancement generation technology, multi-dimensional correlation analysis is performed on the graph structure data to form the domain knowledge graph.

[0012] In one embodiment, the step of responding to a user triggering a target business scenario by outputting at least one guiding question matching the target business scenario through the intent recognition module, based on the domain knowledge graph and the priority weights, includes: Obtain the user's target historical question set, and determine the historical associated question sequence corresponding to the target historical question set; The intent recognition module analyzes user behavior pattern information based on the historical associated question sequence. Based on the collaborative filtering algorithm, a target question matching the user behavior pattern is selected from the domain knowledge graph; Obtain the priority weight of the target question, and generate and output the guiding question based on the priority weight using a large model real-time recommendation engine.

[0013] In one embodiment, the control method of the customer relationship management system further includes: Based on a preset testing strategy, the users were divided into an experimental group and a control group; A first guiding question generated based on a first strategy is output to the experimental group, and a second guiding question generated based on a second strategy is output to the control group. Receive first feedback data corresponding to the first guidance question, and second feedback data corresponding to the second guidance question; The first strategy or the second strategy is adjusted by comparing the first feedback data and the second feedback data.

[0014] In one embodiment, before the step of outputting at least one guiding question matching the target business scenario through the intent recognition module based on the domain knowledge graph and the priority weight in response to a user triggering a target business scenario, the method further includes: Based on a preset time window, the user's operation data is obtained; Identify a set of high-frequency target questions from the user operation data; Based on the target high-frequency question set and the domain knowledge graph, generate and output preliminary guiding questions.

[0015] In addition, to achieve the above objectives, this application also provides a control device for a customer relationship management system, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the control method for the customer relationship management system as described above.

[0016] In addition, to achieve the above objectives, this application also provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the control method of the customer relationship management system as described above.

[0017] One or more technical solutions proposed in this application have at least the following technical effects: This application generates a set of related questions based on at least one predefined business scenario and constructs a domain knowledge graph. It assigns priority weights to conflicting questions in the graph and integrates them into an intent recognition module. When a user triggers a target business scenario, the intent recognition module dynamically outputs matching guidance questions based on the knowledge graph and weights. Addressing the core shortcomings of existing intelligent agent recommendation logic being rigid and lacking feedback-driven optimization, this application achieves continuous adaptive adjustment of recommended content based on user interaction behavior through weight allocation and the context association mechanism of the knowledge graph. This effectively improves the accuracy, relevance, and long-term usability of the recommendation system in complex business scenarios. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the first embodiment of the control method for the customer relationship management system of this application; Figure 2 This is an example diagram illustrating the application scenario of the large-model recommendation engine involved in the first embodiment of the control method of the customer relationship management system of this application; Figure 3 This is a flowchart illustrating the second embodiment of the control method for the customer relationship management system of this application; Figure 4 This is a flowchart illustrating the third embodiment of the control method for the customer relationship management system of this application; Figure 5 This is a flowchart illustrating the fourth embodiment of the control method for the customer relationship management system of this application; Figure 6 This is a schematic diagram of the customer relationship management system framework involved in the fourth embodiment of the control method of the customer relationship management system of this application; Figure 7This is a schematic diagram of the structure of the control device of the customer relationship management system in the hardware operating environment involved in the embodiments of this application.

[0021] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0023] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0024] The main solution of this application embodiment is as follows: Based on the business scenario information of at least one predefined business scenario, generate a set of related questions; based on the set of related questions, construct graph structure data to form a domain knowledge graph; assign priority weights to conflicting questions in the domain knowledge graph; and integrate the domain knowledge graph into the intent recognition module. In response to the user triggering the target business scenario, the intent recognition module outputs at least one guiding question matching the target business scenario based on the domain knowledge graph and the priority weights.

[0025] In existing technologies, Customer Relationship Management (CRM) systems typically operate based on specific business functions, often providing users with guidance questions to facilitate interaction. These systems typically employ static rule-based agents that recommend guidance questions based on pre-set instruction sets or historical user behavior data. However, due to the diverse business needs of different users, CRM systems lack a deep understanding of specific business scenarios. Consequently, during user interactions, the system fails to perceive the user's immediate intent within the current work context, resulting in a significant disconnect between the recommended guidance questions and the user's current business situation, leading to low relevance.

[0026] This application generates a set of related questions based on at least one predefined business scenario and constructs a domain knowledge graph. It assigns priority weights to conflicting questions in the graph and integrates them into an intent recognition module. When a user triggers a target business scenario, the intent recognition module dynamically outputs matching guidance questions based on the knowledge graph and weights. Addressing the core shortcomings of existing intelligent agent recommendation logic being rigid and lacking feedback-driven optimization, this application achieves continuous adaptive adjustment of recommended content based on user interaction behavior through weight allocation and the context association mechanism of the knowledge graph. This effectively improves the accuracy, relevance, and long-term usability of the recommendation system in complex business scenarios.

[0027] To better understand the above technical solutions, exemplary embodiments of this application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of this application are shown in the drawings, it should be understood that this application can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0028] It should be noted that the executing entity in this embodiment can be a customer relationship management system, or a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or control device of a customer relationship management system capable of performing the above functions. This embodiment does not specifically limit the specific implementation. The following uses a customer relationship management system as an example to describe this embodiment and the following embodiments.

[0029] Based on this, embodiments of this application provide a control method for a customer relationship management system, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the control method for the customer relationship management system of this application.

[0030] In this embodiment, the control method of the customer relationship management system includes steps S10 to S40: Step S10: Generate a set of related questions based on business scenario information of at least one predefined business scenario; In this embodiment, the Customer Relationship Management (CRM) system dynamically constructs a set of guiding questions related to each scenario by parsing predefined business scenario configuration information. The CRM system first reads the metadata definition of the business scenario from the scenario configuration library, then generates corresponding related questions based on business skill dependencies, and stores these questions according to scenario categories to form a structured question set. The predefined business scenario refers to a typical work situation pre-configured in the CRM system, such as team morning meeting performance analysis or supply chain financing review. Business scenario information includes the scenario name, the relevant business skills and / or process rules, etc. The related question set is a set of guiding questions automatically generated by the system based on the business scenario information to assist users in completing related operations.

[0031] For example, a customer relationship management system (CRM) starts with key user scenarios, such as team leader morning meeting performance analysis, and initially selects 10 scenarios to design a set of related questions. Each deployed skill provides three follow-up questions, such as an account manager checking whether customer XX has acquired customers with high quality, guiding the questions to: query the customer's marketing path / query the customer's contact information / query the customer's basic business registration information.

[0032] Optionally, the CRM system integrates an intelligent agent that guides intent based on recommendation algorithms. Its core lies in building a dynamic reasoning engine that integrates business knowledge graphs, user behavior data, and real-time context. For example, it enhances semantic understanding through Graph RAG technology, uses a hybrid algorithm of collaborative filtering and reinforcement learning for personalized recommendations, and establishes a complete closed-loop learning mechanism from question generation and user feedback to model fine-tuning. This enables it to proactively predict the potential needs of account managers in corporate banking scenarios and accurately push guiding questions or operational suggestions. The intelligent agent can call various modules within the CRM system to automate and unify the overall CRM functionality.

[0033] As an optional implementation, the customer relationship management system can generate related questions based on business skills. The system obtains business scenario information of predefined business scenarios, determines business functions based on the business scenario information, and generates at least one related guiding question based on the business skill. The related guiding question is then associated with the corresponding predefined business scenario to form a set of related questions.

[0034] Specifically, the system first obtains a list of business skills included in predefined business scenarios, and then configures multiple question templates for each business skill. For example, for the customer credit assessment skill, the system configures templates such as "Analyze the credit history of [customer name]" and "Assess the debt ratio of [customer name]". When a user triggers a relevant scenario, the system obtains specific customer information through real-time data queries and uses a template engine to populate variables and generate complete related questions. Through the mapping relationship between skills and scenarios, the system automatically categorizes these related questions into the corresponding business scenarios, forming a complete set of related questions.

[0035] As an alternative implementation, the system can combine time window analysis to generate related questions. The system first sets a configurable time window, such as the last 7 days or 30 days. Then, the system collects user operation data within this time window and uses natural language processing technology to parse key intents and entities in the operation logs. The system uses a frequent pattern mining algorithm to identify high-frequency business patterns and related question templates. Finally, the system dynamically adds these high-frequency questions to a set of related questions and establishes associations with corresponding business scenarios. For example, if the system finds that many users have recently frequently searched for "supply chain finance eligibility," it will automatically generate similar questions and associate them with supply chain financing scenarios.

[0036] Step S20: Based on the set of related questions, construct graph-structured data to form a domain knowledge graph; In this embodiment, the customer relationship management system transforms a set of related questions into a graph-structured data model. By establishing semantic relationships between questions, entities, and business concepts, it constructs a domain knowledge graph with reasoning capabilities. The system first extracts key entities and concepts from the question set as graph nodes. Then, based on business logic, it establishes edges between nodes to form a knowledge graph that supports complex queries. The graph-structured data model represents entities and their relationships using nodes and edges. Nodes represent questions, users, or business entities, and edges represent logical relationships between entities. The domain knowledge graph is a semantic network constructed from the graph-structured data, used for storing and reasoning about business knowledge.

[0037] Specifically, the customer relationship management system extracts elements such as question text, business entities, and user roles from a set of related questions as nodes in a graph. It uses named entity recognition and relation extraction algorithms from natural language processing to identify key entities and relationships from the question text. For example, from the question "Query customer transaction history," it extracts "customer" and "transaction history" as entity nodes. Next, the system establishes edges between nodes based on a business rule base and semantic relationships, including the association between questions and entities, the business logic relationships between entities, and the semantic similarity relationships between questions. It then uses graph database technology to store these nodes and edges and establishes a corresponding index structure. Finally, the system applies graph retrieval enhancement generation technology to perform multi-dimensional analysis of the graph structure, including using graph neural networks for node embedding representation learning, identifying question clusters through community detection algorithms, and using path analysis algorithms to calculate entity association strength. The optimized graph structure is then persistently stored, forming a domain knowledge graph that supports real-time querying and reasoning.

[0038] It should be noted that, based on the above set of all questions, a question network is formed, which is the knowledge graph of this domain. The customer relationship management system may choose to use Graph Retrieval Augmented Generation (Graph RAG) to represent data such as questions, users, roles, and branches as interconnected entities, and support multi-dimensional association filtering to achieve user-guided question recommendation.

[0039] As an optional implementation, a customer relationship management system (CRM) can construct a knowledge graph through entity extraction and relationship building. Entity information is extracted from business scenario information. Based on this entity information and related guiding questions in a set of related questions, graph nodes are constructed. Edges are established between these graph nodes based on business logic information within the business scenario information and the relationships between the business scenario information and the related guiding questions, forming graph structure data. Graph retrieval enhancement generation technology is then used to perform multi-dimensional correlation analysis on the graph structure data, resulting in a domain knowledge graph.

[0040] Specifically, the system first extracts entity information from business scenario information, including identifying entities such as business skills, user roles, and business objects using named entity recognition technology. Based on this entity information and questions in the associated question set, the system constructs graph nodes. Then, based on the business logic information in the business scenario information, it uses semantic role labeling technology to analyze the semantic relationships between questions and entities, establishing edges between nodes. Finally, it uses graph retrieval enhancement generation technology to perform multi-dimensional association analysis on the graph structure data, and encodes semantic information into the graph structure through graph embedding algorithms, forming a high-quality domain knowledge graph.

[0041] As an alternative implementation, the customer relationship management system (CRM) can enhance the knowledge graph by incorporating user behavior data. When constructing the knowledge graph, the system integrates historical user operation data and identifies typical user workflows using sequence pattern mining algorithms. The system then dynamically adds business logic edges to the knowledge graph using this workflow information. Simultaneously, the system performs semantic analysis on a historical question set, calculates semantic similarity between questions using a word vector model, and establishes semantic association edges between questions based on similarity thresholds.

[0042] Step S30: Assign priority weights to conflicting issues in the domain knowledge graph and integrate the domain knowledge graph into the intent recognition module; In this embodiment, the customer relationship management system (CRM) handles semantic conflicts in the knowledge graph through an intelligent weight allocation mechanism and integrates the optimized knowledge graph into the intent recognition module. The system detects semantic conflicts and logical contradictions in the knowledge graph, calculates priority weights based on multi-dimensional factors, and seamlessly integrates the knowledge graph with weighted information into the intent recognition process. Here, conflict issues refer to semantic overlaps or logical contradictions in the knowledge graph, priority weights are numerical parameters used to quantify the importance of the issues, and the intent recognition module is the algorithm component of the system responsible for parsing user input and matching intents.

[0043] Specifically, the customer relationship management system (CRM) identifies conflicting or overlapping question nodes in the knowledge graph through semantic similarity calculation and business rule checks using natural language processing (NLP) technology. The system uses word vector models to calculate the semantic similarity of question texts and combines this with constraints from the business rule base to detect logical contradictions. For example, semantically similar questions like "approving loan applications" and "approving credit limits" are identified as potential conflicts. Based on multiple dimensions such as user historical behavior data, business importance, and timeliness requirements, a machine learning weighted algorithm is used to assign priority weights to each question. The system considers parameters such as user click-through rate, business rule priority, and time decay factors, and uses a logistic regression model to comprehensively calculate the final weights. Then, the knowledge graph with weighted information is integrated into the intent recognition module through an API interface, establishing a real-time data access channel. A microservice architecture ensures that the intent recognition module can access knowledge graph data with low latency. The parameters and strategies of the intent recognition module are configured, and the intent matching algorithm is dynamically updated through a configuration management service, ensuring that the module can correctly use the weighted information in the knowledge graph for intent matching and recommendation ranking.

[0044] Optionally, the customer relationship management system can assign weights to conflict issues within the network, connect to the intent recognition control center, provide guidance questions recommendations within the scenario, and update the weight design based on user click data analysis.

[0045] As an optional implementation, the customer relationship management system (CRM) can assign weights based on user roles and historical data. Specifically, the system first obtains a predefined set of concerns associated with user role information. Then, it acquires historical operation data and performs semantic analysis on the historical question set, using a frequent itemset mining algorithm to determine the high-frequency question set. Next, based on the predefined concern set and the high-frequency question set, the system uses a multi-objective optimization algorithm to assign priority weights to the questions in the knowledge graph. The system balances the objectives of business importance and user preference, using the Pareto optimality method to derive the weight allocation scheme, i.e., the priority weights. The system then associates the priority weights with the knowledge graph and integrates the complete knowledge graph into the intent recognition module.

[0046] As an alternative implementation, the customer relationship management system can dynamically adjust weights using reinforcement learning. After assigning initial weights, the system continuously collects user interaction data and dynamically adjusts the weights through deep reinforcement learning algorithms. The system transforms user feedback signals into a reinforcement learning reward function, trains the knowledge graph based on a proximal policy optimization algorithm, and updates priority weights in real time. The system uses an experience replay mechanism to store historical interaction data, improving training efficiency and stability, and enabling the system to continuously learn and optimize.

[0047] Optionally, the customer relationship management system obtains a predefined set of concerns associated with user role information, acquires historical operation data, performs semantic analysis on the historical question set in the historical operation data, determines a high-frequency question set based on the semantic analysis results, calculates the priority of questions in the domain knowledge graph based on the predefined set of concerns and the high-frequency question set, obtains the priority weight of conflicting questions, associates the priority weight with the domain knowledge graph, and integrates the domain knowledge graph into the intent recognition module.

[0048] Step S40: In response to the user triggering the target business scenario, the intent recognition module outputs at least one guiding question that matches the target business scenario based on the domain knowledge graph and priority weights.

[0049] In this embodiment, the customer relationship management system monitors user operations in real time. Upon detecting a business scenario trigger event, it immediately initiates the intent recognition process, generating accurate guidance question recommendations based on knowledge graphs and weight information. Through multi-dimensional contextual analysis, real-time intent reasoning, and personalized recommendation algorithms, the system ensures that the most relevant guidance questions are recommended to the user at the appropriate time.

[0050] Specifically, the Customer Relationship Management System (CRM) continuously monitors user actions on the interface through an event listening mechanism. When an event matching predefined business scenario trigger conditions is detected, such as a user entering a specific function page or performing a key operation, the system immediately activates the intent recognition module. The intent recognition module first collects current context information, including user identity, operation history, and page state, using context-aware technology to construct a complete scenario profile. The intent recognition module accesses the domain knowledge graph, performs multi-dimensional reasoning analysis based on priority weights, and uses graph neural network algorithms to filter out the question nodes most relevant to the current scenario and context. It comprehensively considers factors such as question weight, semantic relevance, and business logic matching, and uses an attention mechanism for comprehensive scoring and ranking. Next, the CRM uses a recommendation engine to further optimize the selected questions, including personalized filtering based on collaborative filtering, sampling control based on diversity constraints, and real-time updates based on timeliness verification. The system uses a Transformer architecture sequence model to process user historical behavior sequences and predict potential user intents. Finally, the system presents the optimized list of guiding questions to the user through user interface components, using dynamic rendering technology to ensure the accuracy and usability of the recommended content.

[0051] Optionally, the customer relationship management system can leverage knowledge graph-based contextual semantic recommendation, combined with user history-related question analysis, enhanced collaborative filtering, and weighted design to generate in-scenario guided question recommendations. Simultaneously, the large-scale real-time recommendation engine employs a Graph RAG scheme and model fine-tuning to achieve real-time recommendations, representing questions, users, roles, and line breaks as interconnected entities, supporting multi-dimensional correlation filtering to achieve guided question recommendations for users.

[0052] Furthermore, the customer relationship management system (CRM) can generate guiding questions through behavioral pattern analysis. The system first obtains the user's set of target historical questions, uses a sequence-to-sequence model to analyze the sequence patterns and relationships corresponding to the target historical question set, and determines the historical associated question sequence. Based on the historical associated question sequence, the intent recognition module analyzes user behavioral pattern information, and uses a clustering algorithm to identify the user's work habits and preference characteristics. A collaborative filtering (CF) algorithm is used to select target questions matching the user's behavioral patterns from the domain knowledge graph. Matrix factorization is used to calculate the user-question matching degree, and based on this matching degree, target questions matching the user's behavioral patterns are selected from the domain knowledge graph. The system then uses a large-scale model real-time recommendation engine, combining question priority weights to generate and output the final guiding question recommendation.

[0053] It's important to note that collaborative filtering is one of the most classic methods in recommender systems. It comes in two forms: user-to-user collaborative filtering and item-to-item collaborative filtering. The algorithm analyzes users' historical behavior to identify similarities between users or relationships between items, recommending items liked by other users with similar historical behavior or items similar to the user's known preferences. Collaborative filtering algorithms are implemented using techniques such as matrix factorization, neighborhood methods, or deep learning to improve the accuracy and coverage of recommendations.

[0054] As an optional implementation, the system can directly output preset guidance questions to the user. Since the number of guidance questions required for business functions in predefined business scenarios is limited, different guidance questions can be set according to different business functions. After the user triggers the target business scenario, the system calculates the matching degree between the user and the guidance questions in that target business scenario, selects a guidance question for the user, and outputs it.

[0055] As another optional implementation, after calculating the matching degree between the user and the guidance question, the system can combine user information, user behavior patterns and / or scenario information of the target business scenario to construct contextual information and form prompt words for the big model. The big model real-time recommendation engine can generate guidance questions for the user that conform to the current business scenario and / or context based on the prompt words and output them.

[0056] For example, such as Figure 2 As shown, Figure 2This diagram illustrates an application scenario of the large-scale model recommendation engine involved in this application. For example, in a bank CRM system, when account manager user A is viewing customer X1's annual average deposit data, the system uses a Graph RAG-enhanced large-scale model recommendation engine to generate accurate scenario guidance questions in real time. When user A views customer X1's annual average deposit data in the CRM system, the large-scale model real-time recommendation engine immediately starts. The system first uses Graph RAG technology to map the entities in the current context—user A (role: account manager), customer X1, and annual average deposit (deposit metric)—to corresponding nodes in the knowledge graph. The knowledge graph pre-constructs rich relationships: user A belongs to a business team, customer X1 is marked as a high-value customer, annual average deposit is an important component of the deposit metric, and the deposit metric is closely related to the team's deposit performance requirements. Based on these relationships, the Graph RAG engine performs multi-hop queries, discovering that user A's team is currently under pressure to meet deposit performance targets, and simultaneously identifying customer X1 as a high-value customer with significant deposit contribution potential. This deep relationship information is injected into the large-scale model context in real time. At this point, the finely tuned main model generates recommendations based on key factors, including the current operational context (viewing deposit data), extended business information retrieved through Graph RAG (team performance pressure, customer value level), and user A's historical behavior patterns. Specifically, this user frequently compares the overall team performance with individual customer deposit data. Therefore, the system generates a guiding question: "View your team's deposit performance this month." When user A clicks on this recommended question, the system records this positive feedback and uses it for subsequent model fine-tuning. During model training, the system uses reinforcement learning to positively reward such business-related recommendations, enabling the model to gradually learn a recommendation strategy that proactively associates important customer deposit data with team performance.

[0057] This application's embodiments generate a set of related questions based on at least one predefined business scenario and construct a domain knowledge graph. Conflicting questions in the graph are assigned priority weights and integrated into an intent recognition module. When a user triggers a target business scenario, the intent recognition module dynamically outputs matching guiding questions based on the knowledge graph and weights. Addressing the core shortcomings of existing intelligent agent recommendation logic being rigid and lacking feedback-driven optimization, this application utilizes weight allocation and the contextual association mechanism of the knowledge graph to achieve continuous adaptive adjustment of recommended content based on user interaction behavior. This effectively improves the accuracy, relevance, and long-term usability of the recommendation system in complex business scenarios.

[0058] Based on the same inventive concept, this application also provides a second embodiment, referring to... Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the control method for the customer relationship management system of this application.

[0059] In this embodiment, the control method of the customer relationship management system includes steps S50-S70: Step S50: Collect user interaction data on the guided questions and convert the interaction data into feedback signals for model training; Step S60: Construct the reward function of the reinforcement learning algorithm based on the positive and negative feedback signals in the feedback signals; Step S70: Based on the reward function, perform feedback training on the domain knowledge graph and update the priority weights.

[0060] In this embodiment, after performing the guidance question recommendation, the customer relationship management system initiates a feedback learning mechanism. The system first collects user interaction behavior with the guidance questions in real time through a data acquisition interface, including raw data such as clicks, ignores, and dwell time. It then uses data cleaning and feature extraction techniques to transform this behavioral data into structured feedback signals. The core principle is to make the implicit user preference information explicit through behavioral logs, thus establishing a data foundation for model training.

[0061] Furthermore, the system constructs a reinforcement learning environment based on feedback signals. User clicks and subsequent actions are marked as positive feedback signals and assigned positive reward values, while user actions such as ignoring or quickly closing recommendations are marked as negative feedback signals and assigned negative reward values. A reasonable reward function is designed to establish a mapping mechanism from behavior to evaluation, providing a clear direction for policy optimization. Deep reinforcement learning algorithms are used to train the domain knowledge graph, using node weights as trainable parameters. The policy gradient method is used to adjust the weight distribution based on the reward signals. An experience replay mechanism is introduced during training to improve data utilization efficiency. Through continuous environmental interaction, the weight configuration of the knowledge graph continuously approximates real user needs.

[0062] This application embodiment establishes a complete feedback learning closed loop, enabling the system to continuously optimize the weight configuration of the knowledge graph based on real-time user interaction data. This effectively improves the adaptive ability and long-term stability of the recommendation system and solves the problem of recommendation effect decaying over time due to the lack of a continuous learning mechanism in traditional recommendation systems.

[0063] Since the system described in Embodiment 2 of this application is a system used to implement the method of Embodiment 1 of this application, those skilled in the art can understand the specific structure and variations of the system based on the method described in Embodiment 1 of this application, and therefore will not be described again here. All systems used in the method of Embodiment 1 of this application fall within the scope of protection of this application.

[0064] Based on the same inventive concept, this application also provides a third embodiment, referring to... Figure 4 , Figure 4 This is a flowchart illustrating the third embodiment of the control method for the customer relationship management system of this application.

[0065] In this embodiment, the control method of the customer relationship management system includes steps S80~S110: Step S80: Based on the preset testing strategy, the users are divided into an experimental group and a control group; Step S90: Output the first guiding question generated based on the first strategy to the experimental group, and output the second guiding question generated based on the second strategy to the control group; Step S100: Receive the first feedback data corresponding to the first guidance question and the second feedback data corresponding to the second guidance question; Step S110: Adjust the first strategy or the second strategy by comparing the first feedback data and the second feedback data.

[0066] It's important to note that traditional strategy updates often rely on offline evaluation or small-scale subjective judgment, which makes it difficult to accurately predict the strategy's performance in real-world environments. A / B testing frameworks, through scientific experimental design, randomly divide users into experimental and control groups. While keeping other conditions constant, only the recommended strategy is changed, thus accurately assessing the actual effects of strategy changes. This method originates from the controlled experiment principle in statistics and has now become a standard effectiveness verification method in the internet industry.

[0067] In this embodiment, the first strategy is a new recommendation strategy to be verified, the first guiding question refers to the experimental group guiding question generated based on the first strategy, the second strategy is a control group strategy generated based on the baseline recommendation strategy, and the second guiding question is the control group guiding question generated based on the second strategy. The first feedback data is the set of interaction behavior data of experimental group users to the first guiding question, and the second feedback data is the set of interaction behavior data of control group users to the second guiding question.

[0068] Specifically, the customer relationship management system (CRM) employs a grouped experiment mechanism to optimize recommendation strategies. First, the system divides user traffic into experimental and control groups based on preset experimental configurations. Stratified sampling ensures a balanced distribution of key features between the two groups, while a hash algorithm guarantees the randomness and repeatability of the grouping process. This scientific grouping eliminates external interference, ensuring the reliability of subsequent performance comparisons. The system then implements differentiated recommendation strategies for different user groups. Experimental group users are provided with a first guiding question generated based on the new algorithm, while the control group users are provided with a second guiding question generated based on a baseline algorithm. A traffic control mechanism ensures the accuracy of strategy execution, and a parallel experimental environment is constructed for objective comparison of strategy effectiveness. The system receives feedback data from both groups in real-time through a data pipeline, including core metrics such as click-through rate and conversion rate. Stream processing technology is used for real-time cleaning and aggregation, establishing a complete experimental monitoring system. Statistical hypothesis testing methods are used to analyze the significant differences between the two groups. The strategy management module dynamically adjusts the parameter configuration of the recommendation algorithm, extending the validated strategy to all users and achieving data-driven strategy iteration.

[0069] In one example, during the actual system operation phase, when fluctuations in recommendation performance are detected, the system automatically initiates A / B testing to optimize the strategy. At this point, the first guiding question is generated using a new strategy dynamically adjusted based on real-time feedback data. This strategy optimizes the weight parameters in the knowledge graph in real-time by analyzing recent user interaction records. The second guiding question is generated using the stable strategy before adjustment, maintaining the original weight configuration. For example, in the "supply chain finance recommendation" scenario of a customer relationship management system, the first guiding question received by the experimental group users will dynamically adjust the recommendation order based on the latest customer behavior data, while the control group users will still receive recommendations according to the original fixed rules. By comparing the click-through rate, conversion rate, and other indicators of the two groups of users over 7 days, the system can accurately evaluate the actual effect of the parameter adjustment. If the new strategy shows a significant advantage in statistical testing, such as a p-value less than 0.05, the system will automatically promote the new strategy to all users, completing the online self-adjustment process. Simultaneously, the system can also further adjust the strategy parameters of the first or second strategy in real-time based on the comparison results of feedback data.

[0070] In another example, during the model training phase before system deployment, the development team validated the effectiveness of different algorithm variants through A / B testing. The first bootstrapping question was generated using the new algorithm to be validated, such as a deep recommendation model based on graph neural networks, while the second bootstrapping question was generated using a benchmark algorithm currently running stably online, such as a traditional model based on collaborative filtering. During the two-week testing period, the system simultaneously provided services to both experimental and control groups of users, recording user interactions in detail. For example, in the "customer risk assessment" scenario, experimental group users received risk warning suggestions generated based on a deep learning model, while control group users received regular suggestions generated based on a rule engine. By comparing the performance of the two groups of users on professional indicators such as risk assessment accuracy and false positive rate, the development team could make a scientific decision on whether to deploy the new algorithm to the production environment. This validation method effectively avoided subjective judgment, ensuring that each algorithm upgrade was based on sufficient data support.

[0071] This application's embodiments establish a scientific A / B testing framework, enabling the system to optimize recommendation strategies based on empirical data. This improves the scientific rigor and reliability of the recommendation system's iterations and solves the problem of blind strategy optimization caused by the lack of an effective verification mechanism in traditional systems.

[0072] Since the system described in Embodiment 3 of this application is a system used to implement the method of Embodiment 1 of this application, those skilled in the art can understand the specific structure and variations of the system based on the method described in Embodiment 1 of this application, and therefore will not be described again here. All systems used in the method of Embodiment 1 of this application fall within the scope of protection of this application.

[0073] Based on the same inventive concept, this application also provides a fourth embodiment, referring to... Figure 5 , Figure 5 This is a flowchart illustrating the fourth embodiment of the control method for the customer relationship management system of this application.

[0074] In this embodiment, the control method of the customer relationship management system includes steps S41-S43: Step S41: Obtain user operation data based on a preset time window; Step S42: Identify the target set of high-frequency questions from the user operation data; Step S43: Generate and output preliminary guiding questions based on the target high-frequency question set and domain knowledge graph.

[0075] In this embodiment, since users need to interact with the system to trigger specific business scenarios, it is difficult to directly output corresponding prompts or guidance questions to the user before the system interacts with the user. Based on this, the system can determine whether the user is using the system or has logged into the system based on the user's account or camera recognition, and can also output pre-guidance questions based on the user's operation data and / or the user's historical operation records.

[0076] Specifically, the customer relationship management system implements a pre-guided recommendation function. First, the system extracts recent operation data from user behavior logs based on a preset time period. It processes massive historical records using distributed query technology and ensures the timeliness and representativeness of the data through a time window mechanism. This process captures the latest changes in user behavior patterns through periodic data collection. Next, the system uses a frequent pattern mining algorithm to identify high-frequency question sets from the operation data. Natural language processing technology transforms the raw operation records into structured question statements. The FP-Growth algorithm discovers frequently occurring question patterns, and high-frequency questions with statistical significance are selected based on support metrics. This process aims to uncover common needs from group behavior, providing content sources for pre-guided recommendations. Finally, the system associates and matches the identified high-frequency questions with a domain knowledge graph. Graph query technology finds semantically relevant and high-weight graph nodes, and a template-based method generates pre-guided questions in natural language. Recommended content is proactively presented before the user enters a specific business scenario, achieving truly predictive recommendation.

[0077] Optionally, the customer relationship management system uses a cold start design, based on three key questions generated from user role surveys, and combines these with weighted design to create daily pre-launch guidance questions for users, obtained from operational data analysis of historically high-frequency user role questions.

[0078] For example, such as Figure 6 As shown, Figure 6This is a schematic diagram of the customer relationship management system framework involved in this embodiment. Based on a layered technical architecture, the customer relationship management system realizes a full-link pre-guided function from data collection to front-end recommendation. The entire process runs through three core layers: the data input layer, the recommendation strategy layer, and the user interaction layer, forming a complete technical closed loop. In the data input layer, the system continuously collects and aggregates real-time user operation data based on a preset 24-hour time window through the user behavior data operation analysis module. This data includes raw behavior logs such as page browsing, function clicks, and query records, providing rich raw materials for subsequent analysis. The system adopts distributed stream processing technology to ensure that massive behavioral data can be processed and analyzed in real time. After entering the recommendation strategy layer, the system first uses the frequent pattern mining algorithm in the cold start strategy design to identify the target high-frequency question set from the aggregated user behavior data. Specifically, the system analyzes the user operation sequence through the FP-Growth algorithm, counts the frequency of occurrence of each type of question, and filters out high-frequency questions with support exceeding a preset threshold. At the same time, the personalized hybrid recommendation algorithm begins to play its role, performing deep association matching between the identified high-frequency questions and the domain knowledge graph. The system uses a large-scale model recommendation engine to perform semantic understanding and weight calculation on the matching results, ensuring that the generated pre-launch guidance questions not only conform to group behavioral characteristics but also possess personalized recommendation value. Finally, at the user interaction layer, the system presents these multi-layered optimized pre-launch guidance questions to users through a dedicated recommendation module, forming "Guess What You Want to Ask" recommendations. These pre-launch guidance questions are proactively displayed on the initial interface after users log in to the system, providing users with predictive operational suggestions. Throughout the process, the system also incorporates an A / B testing mechanism, continuously optimizing the accuracy and usability of the pre-launch guidance by comparing the effects of different recommendation strategies. For example, in the specific application of a bank CRM system, the data input layer collects all operation records of account managers within the past 24 hours. The recommendation strategy layer, through analysis, identifies "checking customer credit limits" and "checking loan approval progress" as the two most frequent operational needs. The user interaction layer then immediately displays these two pre-launch guidance questions in a prominent position on the homepage when account managers log in to the system the following day, greatly improving business processing efficiency.

[0079] This application embodiment achieves proactive guidance before users explicitly express their needs by mining user group behavior patterns and combining them with the semantic reasoning capabilities of knowledge graphs. This enhances the predictability and proactivity of the system and solves the problem of delayed recommendation timing caused by the passive response of traditional systems.

[0080] Since the system described in Embodiment 4 of this application is a system used to implement the method of Embodiment 1 of this application, those skilled in the art can understand the specific structure and variations of the system based on the method described in Embodiment 1 of this application, and therefore will not be described again here. All systems used in the method of Embodiment 1 of this application fall within the scope of protection of this application.

[0081] This application provides a control device for a customer relationship management system. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform the control method of the customer relationship management system in Embodiment 1 described above.

[0082] The following is for reference. Figure 7 The diagram illustrates a structural schematic of a control device suitable for implementing a customer relationship management system according to embodiments of this application. The control device for the customer relationship management system in these embodiments may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 7 The control device of the customer relationship management system shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0083] like Figure 7As shown, the control device of the customer relationship management system may include a processing unit 1001 (e.g., a core processor, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 1002 or a program loaded from storage device 1003 into random access memory (RAM) 1004. The random access memory 1004 also stores various programs and data required for the operation of the control device of the customer relationship management system. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the control device of the customer relationship management system to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows a control device for a customer relationship management system with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented or possessed alternatively.

[0084] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0085] The control device for the customer relationship management system provided in this application, employing the control method of the customer relationship management system in the above embodiments, can solve the technical problem of low accuracy in guiding question recommendations in the customer relationship management system. Compared with the prior art, the beneficial effects of the control device for the customer relationship management system provided in this application are the same as those of the control method for the customer relationship management system provided in the above embodiments, and other technical features in the control device for the customer relationship management system are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0086] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0087] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0088] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the control method of the customer relationship management system in the above embodiments.

[0089] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination thereof.

[0090] The aforementioned computer-readable storage medium may be included in the control device of the customer relationship management system; or it may exist independently and not be assembled into the control device of the customer relationship management system.

[0091] The aforementioned computer-readable storage medium carries one or more programs. When the aforementioned one or more programs are executed by the control device of the customer relationship management system, the control device of the customer relationship management system causes the following: based on business scenario information of at least one predefined business scenario, it generates a set of related questions; based on the set of related questions, it constructs graph structure data to form a domain knowledge graph; it assigns priority weights to conflicting questions in the domain knowledge graph; and it integrates the domain knowledge graph into the intent recognition module. In response to a user triggering a target business scenario, the intent recognition module outputs at least one guiding question matching the target business scenario based on the domain knowledge graph and the priority weights.

[0092] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0093] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0094] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0095] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the control method of the customer relationship management system described above. This addresses the technical problem of low accuracy in the guidance question recommendations provided by the customer relationship management system. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the control method of the customer relationship management system provided in the above embodiments, and will not be elaborated upon here.

[0096] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A control method for a customer relationship management system, characterized in that, The method includes the following steps: Generate a set of related questions based on business scenario information from at least one predefined business scenario. Based on the aforementioned set of related questions, a graph-structured data is constructed to form a domain knowledge graph; Priority weights are assigned to conflict issues in the domain knowledge graph, and the domain knowledge graph is integrated into the intent recognition module; In response to a user triggering a target business scenario, the intent recognition module outputs at least one guiding question that matches the target business scenario, based on the domain knowledge graph and the priority weight.

2. The control method for a customer relationship management system as described in claim 1, characterized in that, After the step of responding to a user-triggered target business scenario by outputting at least one guiding question matching the target business scenario through the intent recognition module based on the domain knowledge graph and the priority weight, the method further includes: Collect user interaction data on the guided questions and convert the interaction data into feedback signals for model training; Based on the positive and negative feedback signals in the feedback signals, a reward function for the reinforcement learning algorithm is constructed. Based on the reward function, the domain knowledge graph is trained using feedback, and the priority weights are updated.

3. The control method for a customer relationship management system as described in claim 1, characterized in that, The step of assigning priority weights to conflict issues in the domain knowledge graph and integrating the domain knowledge graph into the intent recognition module includes: Retrieve a set of predefined concerns associated with user role information; Acquire historical operation data and perform semantic analysis on the set of historical questions in the historical operation data. Determine the set of high-frequency questions based on the results of the semantic analysis. Based on the predefined set of concerns and the set of high-frequency questions, the priority of the questions in the domain knowledge graph is calculated to obtain the priority weight of the conflicting questions. The priority weights are associated with the domain knowledge graph, and the domain knowledge graph is integrated into the intent recognition module.

4. The control method for the customer relationship management system as described in claim 1, characterized in that, The step of generating a set of related questions based on business scenario information of at least one predefined business scenario includes: Obtain the business scenario information of the predefined business scenario, and determine the business function based on the business scenario information; Based on the aforementioned business skills, generate at least one relevant guiding question; The associated guidance questions are associated with the corresponding predefined business scenarios to form the associated question set.

5. The control method for a customer relationship management system as described in claim 1, characterized in that, The steps of constructing a graph-structured data based on the set of related questions to form a domain knowledge graph include: Extract entity information from the aforementioned business scenario information; Based on the entity information and the associated guiding questions in the associated question set, construct graph nodes; Based on the business logic information in the business scenario information, and the relationship between the business scenario information and the associated guidance question, edges are established between the graph nodes to form the graph structure data; Using graph retrieval enhancement generation technology, multi-dimensional correlation analysis is performed on the graph structure data to form the domain knowledge graph.

6. The control method for a customer relationship management system as described in claim 1, characterized in that, The step of responding to a user-triggered target business scenario by outputting at least one guiding question matching the target business scenario through the intent recognition module, based on the domain knowledge graph and the priority weights, includes: Obtain the user's target historical question set, and determine the historical associated question sequence corresponding to the target historical question set; The intent recognition module analyzes user behavior pattern information based on the historical associated question sequence. Based on the collaborative filtering algorithm, a target question matching the user behavior pattern is selected from the domain knowledge graph; Obtain the priority weight of the target question, and generate and output the guiding question based on the priority weight using a large-scale real-time recommendation engine.

7. The control method for a customer relationship management system as described in claim 1, characterized in that, The control method of the customer relationship management system further includes: Based on a preset testing strategy, the users were divided into an experimental group and a control group; A first guiding question generated based on a first strategy is output to the experimental group, and a second guiding question generated based on a second strategy is output to the control group. Receive first feedback data corresponding to the first guidance question, and second feedback data corresponding to the second guidance question; The first strategy or the second strategy is adjusted by comparing the first feedback data and the second feedback data.

8. The control method for a customer relationship management system as described in claim 1, characterized in that, Before the step of responding to a user-triggered target business scenario and, through the intent recognition module, outputting at least one guiding question matching the target business scenario based on the domain knowledge graph and the priority weights, the method further includes: Based on a preset time window, the user's operation data is obtained; Identify a set of high-frequency target questions from the user operation data; Based on the target high-frequency question set and the domain knowledge graph, generate and output preliminary guiding questions.

9. A control device for a customer relationship management system, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the control method for the customer relationship management system as described in any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the control method of the customer relationship management system as described in any one of claims 1 to 8.

Images