Online customer service method, system and device, storage medium and program product

Abstract

The embodiment of the invention provides an online customer service method, system and device, a storage medium and a program product. According to the online customer service scheme provided by the invention, when the distributed customer service is determined to be the intelligent customer service for the user entering the session interface, the customer service is provided for the user by using the preset model, and the dialogue content in the session interface is detected based on the preset rule set by using the rule engine in the process; and when at least one rule in the preset rule set is matched in the dialogue content, executing an adaptive action according to the at least one rule. According to the scheme, the rule engine is introduced to detect the whole session in real time in the stage of providing the customer service for the user by using the large model (the preset model), so that sudden problem scenes can be well coped with, and the overall service quality of the intelligent customer service is further improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an online customer service method, system, device, storage medium, and program product. Background Art

[0002] With the rapid development of large-scale model technology, its application in intelligent customer service is becoming increasingly widespread. For example, currently, intelligent customer service systems are generally built using large language models (LLMs). While these large-scale model-based intelligent customer service systems have improved intelligence capabilities, they also suffer from issues such as model hallucinations, which can lead to unrealistic answers and low service quality in emergencies. These issues seriously affect the operation of intelligent customer service systems and provide users with a poor service experience. Summary of the Invention

[0003] Multiple embodiments of the present application provide an online customer service method, system, electronic device, storage medium and program product to solve problems such as model hallucinations and low emergency service quality in intelligent customer service systems based on a single large model.

[0004] In a first embodiment, the present application provides an online customer service method. The method includes:

[0005] Determining a suitable customer service representative for the user;

[0006] When the adapted customer service is an intelligent customer service, a preset model is used to provide customer service to the user;

[0007] The conversation content in the customer service conversation interface is detected to switch customer service when the service provided by the intelligent customer service is abnormal.

[0008] In a second embodiment, the present application also provides an online customer service method. The method includes:

[0009] In response to a user entering a conversation interface, determining, using a rules engine, an assigned customer service type for the user;

[0010] Based on the customer service type, accessing a target customer service in the conversation interface to provide customer service to the user;

[0011] The rule engine is used to detect the conversation content in the conversation interface, so that when the conversation content matches the rules in the rule engine, an adaptation action is performed according to the rules.

[0012] In a third embodiment, the present application also provides an online customer service method. The method includes:

[0013] Use the rules engine to determine the type of customer service assigned to users entering the conversation interface;

[0014] When the customer service type is intelligent customer service, an intelligent agent is connected to the conversation interface to provide customer service to the user;

[0015] The rule engine is used to detect the conversation content in the conversation interface, so that when the conversation content matches the rules in the rule engine, an adaptation action is performed according to the rules.

[0016] In a fourth embodiment, the present application also provides an online customer service method. The method includes:

[0017] Leverage intelligent agents to provide customer service to users entering the conversational interface;

[0018] The rule engine and the agent are used to collaboratively control the customer service conversation status in the conversation interface.

[0019] In a fifth embodiment, the present application provides an online customer service system. The system includes a client and a server. The client is configured to respond to a user's conversation request and display a conversation interface; the conversation interface is configured to display user-related conversation content. The server is configured to implement the various method embodiments provided above.

[0020] In a sixth embodiment, the present application provides an electronic device. The electronic device includes: a memory and a processor, wherein the memory is configured to store a program; and the processor is coupled to the memory and configured to execute the program stored in the memory to implement the various method embodiments provided above.

[0021] In a seventh embodiment, the present application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a computer, the computer program can implement the various method embodiments provided in the present application.

[0022] In an eighth embodiment, the present application provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, implements the above-mentioned various method embodiments provided by the present application.

[0023] The solutions provided by various embodiments of the present application utilize a preset model to provide customer service to the user when the customer service agent is determined to be an intelligent customer service agent. During this process, the conversation content in the conversation interface is also monitored to switch customer service agents if the service provided by the agent is abnormal. Specifically, a rule engine is used to monitor the conversation content in the conversation interface based on a preset rule set. When the conversation content matches at least one rule in the preset rule set, an adaptation action is performed according to the at least one rule to switch the customer service type. This method, while using a large model (the preset model) to provide customer service to the user, introduces a rule engine to monitor the entire conversation in real time, ensuring better response to unexpected issues and improving the overall service quality of intelligent customer service. Furthermore, the rule engine and the agent (including the preset model) are used to collaboratively control the customer service conversation status in the conversation interface, so that both are jointly responsible for the current response. This ensures a consistent and coherent user experience throughout the entire customer service process (for example, when the rule engine triggers a transfer to a human customer service agent, the agent can continue to respond to the user's input in the conversation). Of course, when it is determined that the customer service type assigned to the user is manual, a manual customer service agent is connected to the conversation interface to provide customer service to the user. A rule engine can also be used to monitor the entire conversation to improve the quality of customer service. Therefore, this application uses the customer service type determined for the user to connect a target customer service agent (either a preset model (or an agent containing the preset model) or a manual customer service agent) to provide customer service to the user in the conversation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0025] Figure 1 A flowchart of an online customer service method provided by an exemplary embodiment of the present application;

[0026] Figure 2 A schematic diagram illustrating a user entering a conversation interface according to an exemplary embodiment of the present application;

[0027] Figure 3a 、 Figure 3b and Figure 3c A schematic diagram illustrating the principle of implementing online customer service provided by an exemplary embodiment of this application;

[0028] Figure 4 and Figure 5 An example diagram of response information output by an agent provided in an exemplary embodiment of the present application;

[0029] Figure 6A schematic structural diagram of an electronic device provided as an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0030] To provide better service to users, various platforms (such as e-commerce platforms and short video platforms) often offer intelligent customer service systems. Currently, with the advancement of big model technology, intelligent customer service systems are generally rebuilt and upgraded using big models such as LLM to achieve greater intelligence. However, while these customer service systems built on big LLM models have improved their intelligence capabilities, they also expose numerous issues. For example, big models are prone to "hallucinations," where the answers generated by the model to question-and-answer questions are unrealistic, even completely fictitious, and fabricated. This phenomenon is particularly prevalent when dealing with complex or unusual questions. Another example is the low quality of service during emergencies. Specifically, when dealing with sudden, large-scale events, the corresponding questions are beyond the capabilities of the big model and cannot be handled. These issues seriously affect the operation of the customer service system, resulting in a poor service experience for users.

[0031] In response to the problems existing in the current intelligent customer service system, this application provides a solution. The basic idea of ​​this solution is to solve problems such as unstable service quality of large models in the intelligent customer service system by introducing a rules engine (RulesEngine), thereby improving the service level and stability of the intelligent customer service system; in addition, the retrieval-augmented generation (RAG) technology is introduced to solve the large model hallucination problem.

[0032] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0033] It should be noted that, in the case of user information involved in the embodiments of the present application, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the embodiments of the present application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data need to comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse. In addition, the various models involved in this application (including but not limited to language models or large models) are in compliance with relevant laws and standards.

[0034] In addition, it should be noted that when the embodiments of the present application involve user interaction operations or triggering operations, the user interaction operations or triggering operations involved in the embodiments of the present application include but are not limited to: touch operations, gesture operations, voice operations, head movement operations, eye movement operations and other interactive operations in various ways; among which, touch operations include but are not limited to: click operations, double-click operations, long press operations, sliding operations, pinch operations or mouse hover operations, etc. Sliding operations include but are not limited to: straight sliding, curved sliding, etc.

[0035] Before introducing each embodiment, some technical terms appearing in this document are briefly explained. It should be understood that this explanation is for a clearer understanding of the embodiments of the present application and does not necessarily constitute a limitation of the embodiments of the present application.

[0036] A preset model refers to a machine learning model with a large number of parameters and complex computational structure, generated by training with massive amounts of data using self-supervised or unsupervised methods. It provides excellent distributed feature representation and model generalization capabilities for downstream tasks. A preset model can be called a generalized large model. The preset model can be obtained by training and fine-tuning a base model. The base model can be a large language model (LLM) or other machine learning model. The base model has learned rich feature representations and pattern recognition capabilities. Through fine-tuning, these learned feature representations can be used to quickly and effectively adapt to tasks in specific scenarios (such as tasks in e-commerce customer service scenarios). Fine-tuning is faster than training a model from scratch, which accelerates the training process.

[0037] The retrieval enhancement model is a RAG tool. Combining the preset model with the retrieval enhancement model enables the preset model to access and utilize external knowledge bases. After the preset model and the retrieval enhancement model are combined, the retrieval enhancement model is responsible for querying the external knowledge base based on the input information to obtain relevant external knowledge and combining the retrieved external knowledge with the input information to enhance the input information. The preset model is used to process the enhanced input information and generate output information. The knowledge base is a database that stores external information.

[0038] Large Language Models (LLMs): These are deep learning models with a large number of parameters, primarily used for natural language processing (NLP) tasks. Based on the Transformer architecture, LLMs can understand and generate high-quality human language text.

[0039] A rules engine generally refers to a system that executes based on a rule set. Specifically, it is a software tool for processing service logic, allowing users to define, execute, and manage a series of "if"-"then" rules (i.e., conditions and actions). These rules can be used to simulate the human decision-making process, automatically reacting and triggering corresponding operations based on input data or events. The main components of a rules engine include: rule set, working memory, inference environment, and input / output interface. A rule set is a set of predefined rules. Each rule consists of a condition (when certain conditions are met) and an action (to perform a specific action). Working memory: Stores all relevant information about the current session or transaction. The rules engine uses this information to evaluate whether the conditions in the rules are met. The inference environment: Responsible for executing the rule set, examining the data in the working memory and deciding which rules should be executed. The input / output interface: Interacts with external systems, receiving input data and transmitting processing results.

[0040] Intelligent agent: The corresponding English word is Agent. Agent is a software or hardware entity that can act autonomously. In the field of artificial intelligence, Agent is translated as "intelligent body", and is also translated as "agent", "agent", "intelligent subject", etc. Intelligent agents have the following characteristics: intelligent agents can automatically adjust their behavior and state according to changes in the external environment, and have the ability to respond to external stimuli; intelligent agents can take the initiative to act in response to changes in the external environment; intelligent agents have the ability to cooperate with other intelligent agents or people, and different intelligent agents can interact with other intelligent agents or people according to their own intentions to achieve the purpose of solving problems; intelligent agents can accumulate or learn experience and knowledge, and need to change their own behavior to adapt to the environment. With the above content, that is: intelligent agent refers to an entity that can autonomously perceive the environment, make decisions and execute actions.

[0041] Among them, the intelligent agent can be constructed based on the preset model mentioned above. In addition, during the construction, the retrieval enhancement model mentioned above can be further introduced.

[0042] The following describes the various embodiments provided in this application in conjunction with the accompanying drawings.

[0043] Figure 1 This is a flowchart of an online customer service method provided as an exemplary embodiment of the present application. This method is performed by a server in an online customer service system. The server can be a server, service cluster, virtual server, or cloud-based system, and this embodiment does not specifically limit this. The server provides corresponding functional services to the client, such as intelligent customer service. Therefore, when the server is used in a customer service scenario, it can be understood that the server is a customer service workstation.

[0044] It is important to note that this application uses the intelligent customer service scenario on an e-commerce platform as an example to introduce the technical solution of this application. In fact, this application solution can also be applied to intelligent customer service scenarios on other types of platforms, such as intelligent question-and-answer scenarios on corporate websites.

[0045] See also Figure 1 As shown, the online customer service method provided in this embodiment includes the following steps:

[0046] 101. Determine a suitable customer service representative for the user;

[0047] 102. When the adapted customer service is an intelligent customer service, provide customer service to the user using a preset model;

[0048] 103. Detect the conversation content in the conversation interface to switch customer service when the service provided by the intelligent customer service is abnormal.

[0049] In the above 101, the determination of a customer service representative for the user may be triggered in response to the user entering the conversation interface. The user may enter the conversation interface by logging in through a client device. The client device may include, but is not limited to, a smartphone, smart wearable device, tablet computer, laptop computer, desktop computer, and the like.

[0050] The term "user connection" is used to indicate the start of a user's interaction with the customer service system. This refers to the user accessing the customer service system through a browser, client application (APP), web application (HyperText Markup Language 5, the fifth generation of HTML), light application (also known as mini-program, a lightweight application), or cloud application on the customer service device. This term is often used to mark the starting point of a new customer service session.

[0051] For example, see Figure 2 As shown, an e-commerce application is installed on a client device. After entering the "My" interface provided by the e-commerce application, the user clicks the dedicated customer service control on the "My" interface to initiate a conversation request. After receiving the conversation request, the server creates a conversation and connects the user to the conversation, that is, the user enters the conversation interface 11 of the conversation. Furthermore, the user can enter conversation information (such as inquiry information) through the input box in the conversation interface.

[0052] In addition, see Figure 3a The user enters the corresponding link 1 shown in FIG. 1 . After the session is created, the user is further assigned to a customer service representative. The execution of assigning a customer service representative triggers the hosting mechanism. When the hosting mechanism is triggered, the customer service system will be based on the hosting admission policy (such as Figure 3b) to make an admission judgment to decide whether it is necessary to provide customer service to the user through intelligent customer service.

[0053] In this application, an introduced rule engine is used to perform the above-mentioned access judgment. A preset rule set is deployed in the rule engine, and each rule in the preset rule set is usually composed of a condition (when) and an action (then). When the condition is met, the corresponding action is triggered. That is: the condition (when) is the prerequisite for the specified trigger rule, and the action (Then) is the operation to be performed when the condition is met. Based on the above content, in a specific achievable technical solution, the "determining an adapted customer service for the user" in the above 101 may include:

[0054] 1011. Use the rule engine to determine suitable customer service for the user based on a preset rule set.

[0055] During specific implementation, the above-mentioned preset rule set may include customer service allocation rules. It can be understood that the customer service allocation rules are managed access policies. The rule engine can make access judgments based on the customer service allocation rules to determine whether intelligent customer service needs to be assigned to the user. Specifically, the triggered rules in the customer service allocation rules can be determined in combination with relevant information about the user (such as one or more of user attribute information, user operation information, dialogue information input by the user, etc.) and / or manual customer service resources, so as to determine the type of customer service assigned to the user according to the triggered rules. Among them, the customer service allocation rules can be defined, but not limited to, based on at least one of the following: user priority (the priority of processing can be determined based on factors such as the user type (VIP, ordinary user, etc.), historical interaction records, etc.), user operation selection, problem complexity, and customer service resource availability (such as manual customer service resources). Customer service types include intelligent customer service and manual customer service.

[0056] For example, if the user level is determined to be an ordinary user based on user attribute information, the customer service type assigned to the user can be determined to be intelligent customer service; conversely, if the user level is a VIP user, the customer service type assigned to the user can be determined to be manual customer service.

[0057] For example, when a user initiates a conversation request, the server will receive it and display a small window asking the user whether they want to talk directly to a human agent or try self-service intelligent customer service first. Based on the user's choice in response to this query, the server can determine whether the user is assigned to an intelligent or human agent.

[0058] For example, after entering the conversation interface, the user enters the first conversation information (such as Figure 2The system analyzes the conversation information to determine the complexity of the question and then determines whether the question is within the knowledge and capabilities of the intelligent customer service representative. If the question is within the intelligent customer service representative's knowledge and capabilities, the user can be assigned to the intelligent customer service representative; if it is not, the user can be assigned to a human customer service representative. For example, if the conversation information entered by the user contains specific keywords such as "refund" and "account locked", indicating an urgent or complex issue, the user can be assigned to a human customer service representative; otherwise, the user can be assigned to an intelligent customer service representative, and so on.

[0059] Another example is checking whether there are currently available human customer service resources. If all customer service personnel are busy with other tasks, the user may need to wait in line for a long time or, in this case, it can be determined that intelligent customer service is assigned to the user.

[0060] If the customer service type assigned to the user is determined to be intelligent customer service (i.e., intelligent customer service access is determined), a hosting operation will be executed to host a preset model to provide customer service to the user. Specifically, the intelligent agent built based on the preset model is used to provide customer service to the user. Therefore, the "using the preset model to provide customer service to the user" in the above 102 may include:

[0061] 1021. Determine an intelligent agent suitable for the user; wherein the intelligent agent includes the preset model;

[0062] 1022. Utilize the intelligent agent to provide customer service to the user.

[0063] The above step 1021 corresponds to Figure 3a The initialization agent in the Agent creates an Agent instance, which is implemented by the Agent Center (also known as Intelligent Customer Service (AI Customer Service)). AI Customer Service is generally a broader system concept that includes multiple intelligent customer service functions and technologies, such as natural language processing, dialogue management, and knowledge base retrieval. The Agent is a specific implementation instance of the AI ​​Customer Service system, focusing on providing customer service, including but not limited to answering user questions, processing orders, and providing product information.

[0064] In the above 1021, a suitable agent can be determined for the user and called based on one or more of the acquired user information (such as preferences, etc.), user search information, page operation information, agents used in historical conversations, and the conversation information input this time.

[0065] For example, before entering the conversation interface, a user searches for "women's summer clothing" on the e-commerce application's homepage and browses. Based on the user's search keyword "women's summer clothing," it can be determined that the user currently wants to consult about women's fashion. In this case, an agent with the role of "fashion expert" can be automatically assigned to the user. For another example, if the user views an order before entering the conversation interface, based on this order view, it can be determined that the user may want to consult about product after-sales service issues. In this case, an agent with the role of "product after-sales service" can be automatically assigned to the user. For another example, if the user immediately enters the conversation interface and enters the first conversation message "What eyelash curler is good for beginners?", based on this conversation message, an agent with the role of "beauty expert" can be assigned to the user. Alternatively, if the user opens the e-commerce application and enters the homepage without performing any operation, the agent assigned to the user in this conversation can be determined based on the agents used in previous conversations. For example, if the user's agent in the last conversation or multiple conversations in the past was "beauty expert," then it can be determined that the user needs to be assigned the agent with the role of "beauty expert" this time.

[0066] Of course, in other embodiments, an agent may be randomly determined for the user. This embodiment does not limit the specific implementation method of determining an agent for the user.

[0067] The determined agent is one of multiple agents in an agent algorithm factory. The agent algorithm factory is an automated system for generating, managing, and optimizing agents for specific customer service scenarios (e.g., e-commerce customer service scenarios).

[0068] For example, in an agent algorithm factory, a base model is fine-tuned using pre-constructed training samples using fine-tuning methods (such as the LoRA method) to enhance the logic and accuracy of the base model's output in e-commerce platform customer service scenarios. This fine-tuned base model (the default model) is then used to construct an agent for e-commerce customer service. LoRA is a model fine-tuning method that freezes the model parameters of the base model and injects a trainable low-rank matrix into the attention layer of the Transformer architecture (a model that uses an attention mechanism to speed up model training). This significantly reduces the number of trainable parameters for downstream tasks and improves training speed. The base model can be a large language model (LLM) or other machine learning model. The base model has learned rich feature representations and pattern recognition capabilities. Fine-tuning can leverage these learned feature representations to quickly and effectively adapt to specific tasks (such as those in e-commerce customer service). Fine-tuning is faster than training the model from scratch, accelerating the training process.

[0069] In the above 1022, after the agent accesses the current session, it begins to provide customer service to the user. Among them, the agent will clear the memory before providing customer service to the user. Clearing the memory means clearing the data related to the previously stored historical sessions, which can avoid data confusion, response errors, etc. For example, when an agent service has other historical users, if the session-related data of the previous user is not cleared in time, it may cause information leakage or response errors. In addition, the use of this timely clearing of memory can also effectively protect user privacy to avoid leakage of user personal data, and can also free up storage space. In addition, the agent will also execute an update memory for the current session to store the acquired data related to the current session in the memory.

[0070] Figure 3c The following is a sequence diagram of the online customer service solution provided by this application. Figure 3c The contents of steps 1 to 5 are shown in the figure, where step 4 is a given session process in which when the hosting end condition is met (such as transferring to manual customer service), the hosting end will be triggered, the agent will exit the session, and the memory update will fail.

[0071] It should be noted that the data related to a session covers all kinds of information collected and generated from the time the customer initiates a session request to the resolution of the problem, such as user information (such as user ID, account information, preferences, contact information, etc.), context information (such as the source channel of the session initiation (such as through a website, application, etc.), device information (such as the type of device used by the user, the operating system version, etc.), user geographic location), conversation content records (such as all communication records between the agent and the user (including text, pictures, videos, etc.), conversation interaction details (such as the start and end time of the session, transfer records, service satisfaction ratings, etc.), problem description and solution (such as the classification of questions raised by the user, the specific response provided by the agent, whether the problem is solved, knowledge base references, etc.), user sentiment analysis results (such as the change trajectory of user emotions), anomaly detection and logs (such as abnormal situations such as tool call failures and knowledge base connection interruptions, key performance indicators such as agent response speed and accuracy). In addition, continue to refer to Figure 3c In step 6 shown in the figure, the server (the customer service workstation) generates a universally unique call UUID (universally unique identifier) ​​for the agent for the current session. This ensures that each session has a unique identifier through the UUID, so that no duplication can be guaranteed even in a high-concurrency environment. In addition, the universally unique call UUID can also facilitate tracking of all operations and events in the session, facilitating subsequent auditing and problem troubleshooting.

[0072] After the agent joins the conversation, if the user has not yet input the conversation information, the agent can use the collected user-related information (such as product order information, user information (such as preferences), etc.) as the input of its preset model to output a response information first, such as Figure 4 Alternatively, if the user has entered a dialogue message, the agent can use the preset model within it to output a response message based on the dialogue message entered by the user, such as Figure 5 The response information given by the customer service robot or Xiaomi is shown in FIG.

[0073] In this application, when the agent performs the answering task, as shown in Figure 3a Shown (or Figure 3b Step 2 shown in, or Figure 3c 7), the preset model is to explore a series of possible action plans (for planning decisions) by calling the built-in planning model, and use the built-in critic mechanism to evaluate each action plan. The evaluation usually involves the following: predicting the possible results after taking a specific action; evaluating the quality of each result based on predefined standards or learned value functions; assigning a score to each action plan to reflect its expected effect. Furthermore, the agent makes decisions based on the scores of each action plan provided by the critic mechanism. Specifically, for example, it selects the action plan with the highest score, and then executes the selected action plan (i.e., the corresponding Figure 3a The planning model is an action decision-making model that generates possible action plans based on the information collected by the agent. Each action plan aims to achieve a specific goal, such as solving a user problem or providing useful information. The critic mechanism can be implemented as, but is not limited to, a neural network to evaluate action plans.

[0074] For example, in a customer service scenario, suppose a user enters a conversation message asking, "How do I reset my password?" The agent, recognizing the conversation message as an account management question, invokes its built-in Planning model to generate several possible response options, such as providing a reset link, asking for other related questions, or providing a window to enter an account number, contact information, or email address. Furthermore, the agent uses the built-in Critic mechanism to evaluate each option output by the Planning model, taking into account factors such as speed of resolution and user satisfaction. The agent then makes a decision based on the ratings of each option provided by the Critic mechanism, selecting an appropriate course of action, such as providing a window to enter an account number, contact information, or email address. The agent then executes the selected course of action and outputs a response message generated based on this course of action.

[0075] Furthermore, in order to solve the problem of model hallucination that may occur when the intelligent agent provides customer service to users, the present application scheme introduces a retrieval enhancement model (i.e., RAG, also called a retrieval enhancement generation model or a retrieval enhancement generation tool) for the intelligent agent. That is, the intelligent agent in 1022 above is constructed based on the preset model and the retrieval enhancement model. For example, a retrieval enhancement model layer can be added outside the preset model to realize the construction of the intelligent agent, so that the functions of the original preset model can be retained, and the retrieval resources can be flexibly configured as needed to ensure the accuracy and reliability of the response information output by the intelligent agent. The above-mentioned retrieval enhancement model is used to query the external knowledge base based on the input of the intelligent agent to obtain relevant external knowledge, and combine the external knowledge with the input to obtain enhanced information. Among them, the retrieval enhancement model can call the knowledge recall tool to query the external knowledge base (such as a knowledge base exclusive to the e-commerce field) to recall relevant external knowledge. The preset model is used to generate the output response information based on the enhanced information. The content described here corresponds to Figure 3b Steps 2 to 3 in the Figure 3c Follow steps 7 to 9 in the .

[0076] The following points need to be supplemented for the response information output by the agent:

[0077] 1) For the response information output by the agent, see Figure 3a As shown, it needs to be processed by the Action Executor in the Customer Service Assistant before being sent. Figure 3c The server shown in the figure is a component of the customer service workbench. The action executor can realize the core function of playing the standard operating procedure (SOP). By calling the SOP to play the response information output by the agent, a standardized and personalized response information can be generated to reply to the user (i.e., the corresponding Figure 3c Steps 16 to 17 shown in ). Calling sop to play includes: playing knowledge according to the knowledge identifier and slot filling information; calling sop to execute the sending action of the corresponding node (sop nod). The above-mentioned knowledge identifier is a unique identifier of the knowledge item (such as knowledge ID), which can be used to quickly locate the specific knowledge content that needs to be played. Slot filling information refers to specific variables or parameters that are dynamically filled into the template, such as the user's name, order number, product model, etc., which can be dynamically inserted into the fixed template according to the context to generate personalized response information. The above-mentioned sop node is a node in sop. SOP may be composed of multiple nodes (node), each node represents a specific action or logical branch, such as performing a specific sending action.

[0078] For example, a user enters the message "When will my order be delivered?" The agent recognizes the user's intent to inquire about the order status and outputs information such as the order number 117####456 and a delivery time of 3 days. The action executor, based on the agent's output, can perform the following operations: It matches a knowledge ID with knowledge_001 and retrieves a corresponding knowledge template from the knowledge base based on this knowledge ID: "Hello, your order number {order_id} has been shipped and is expected to arrive within {delivery_days} days." It then replaces the variables {order_id} in this knowledge template with 117####456 and {delivery_days} with 3 days, generating a standardized response: "Hello, your order number 17####456 has been shipped and is expected to arrive within 3 days." Furthermore, the send_message node in the SOP executes, sending this standardized response to the user through the conversation interface window, completing the response.

[0079] 2) See Figure 3a Before executing and sending the output of the action executor (such as the output standardized response information) to the user-side client, the following processing operations may be performed, but are not limited to: quality inspection and risk control are performed through the Critic evaluation model. The reasons for doing this include, but are not limited to: ensuring the accuracy of the response information to prevent the spread of false information and verifying the factual basis (this can ensure that the information provided is up-to-date and correct), ensuring the language fluency of the response information, and ensuring that the response information does not involve personal information, sensitive content, or inappropriate content (such as discriminatory remarks).

[0080] Furthermore, considering the limitations of the capabilities of the preset models in the intelligent agent, the intelligent agent can often directly give relatively accurate answers to general questions (such as long-tail generalization problems), but in some sudden problem scenarios, it is difficult to guarantee the accuracy of the answers. Sudden problem scenarios refer to problem situations that are difficult to fully predict, occur less frequently, but do occur, such as unforeseen product problems (this problem has not occurred before or is not in the known problem database), cross-domain complex consultations, etc. These sudden problem scenarios often bring challenges to intelligent customer service and are prone to service anomalies. In response to this scenario, the present application uses a rule engine to deal with it, and the rule engine has a built-in preset rule set. Based on this, that is: in a specific implementable technical solution, the above step 103 "detecting the conversation content in the conversation interface" may include:

[0081] 1031. Detect the conversation content using the rule engine.

[0082] 1032. When the rule engine detects that the conversation content matches at least one rule in a preset rule set, an adaptation action is performed according to the at least one rule to switch the customer service type.

[0083] Among them, the above-mentioned conversation content is the customer service conversation content, which includes: the conversation information input by the user and the response information output by the intelligent agent. When the rule engine detects that the conversation content (mainly the conversation information input by the user) matches at least one rule in the preset rule set (that is, it meets the conditions in at least one rule), it can select a rule with the highest matching degree from at least one rule to execute the action defined in this selected rule, such as transferring to manual customer service, having the rule engine act as customer service to provide specific solutions, or switching to another intelligent agent to provide customer service, etc. Among them, the role of the other intelligent agent that is switched is generally different from the role of the intelligent agent currently providing customer service. For example, assuming that the current intelligent agent providing customer service to the user is a "beauty expert" role, when it is detected that the conversation information input by the user is asking about dressing issues, it can switch to an intelligent agent with a "dressing expert" role to continue providing customer service to the user.

[0084] It can be seen here that in this embodiment, when using an intelligent agent to provide customer service to users, a rule engine will be used to monitor the entire conversation in real time. If a rule in the preset rule set is hit in the conversation content, the action in the corresponding rule will be executed. This approach can enhance the customer service response capability and quality, thereby effectively improving the overall service quality of intelligent customer service.

[0085] Furthermore, to ensure the consistency and continuity of the user experience throughout the entire customer service process, the present application solution will also monitor the customer service conversation status simultaneously by the intelligent agent (specifically, the preset model therein) and the rule engine, so that both are jointly responsible for the current conversation. Therefore, the provided method may also include the following steps:

[0086] 104. Utilize the preset model and the rule engine to collaboratively control the conversation state (i.e., customer service conversation state) in the conversation interface.

[0087] In specific implementation, the above step 104 includes but is not limited to at least one of the following steps:

[0088] 1041. When it is determined that the preset model cannot respond to the user's conversation message, the rule engine is triggered, and the rule engine performs one or more of the following adapted actions according to the relevant rules in the rule set: transferring the call to a human customer service representative, or having the rule engine act as a customer service representative to provide a response to the conversation message (e.g., providing a relevant resource link, providing a compensation plan, scheduling a call back, or scheduling an email reply, etc.);

[0089] 1042. During the transfer to manual customer service, the preset model continues to output response information for the dialogue information.

[0090] For example, the basic process of implementing steps 1041 to 1042 may include the following steps:

[0091] S1. Identify the question type: The preset model first interprets and analyzes the user's input and determines the question type. If the question is determined to be beyond the system's capabilities, it may be marked as requiring further processing. Conversely, if the question is determined to be within the system's capabilities, the preset model may directly output a corresponding response based on the user's input.

[0092] S2. Triggering the rule engine: Once it is determined that the problem in the current dialogue information cannot be directly solved by the preset model, the rule engine can be automatically triggered. For example, whether to trigger the rule engine can be determined based on some predefined trigger criteria. Predefined trigger criteria include but are not limited to: problem complexity (for example, if the preset model marks the dialogue information as requiring further processing, it can be determined that the problem is relatively complex and needs to trigger the rule engine for processing), user sentiment analysis (for example, if the user continuously enters the same dialogue information, it means that the user is dissatisfied with the response information output by the preset model, and the rule engine needs to be triggered for processing), etc.

[0093] S3. The rule engine processes user-entered conversations based on a built-in set of preset rules. For example, a rule in the set might be "Automatically transfer difficult conversations to human customer service." When this rule is triggered, the rule engine initiates the transfer to human customer service. During the transfer (before successful access to human customer service), the preset model can continue to communicate with the user through natural language to ensure a consistent and continuous user experience. Once human customer service is successfully accessed, the preset model (i.e., the agent) terminates its hosting and exits the conversation. For another example, after the rule engine is triggered, it might respond to questions in the conversation based on the corresponding rules in the preset rules. For example, it might provide links to relevant self-help resources (e.g., links to product assembly tutorial videos). Alternatively, if the issue cannot be resolved immediately, it might prompt a call-back or email response for the user to fill in their contact information, facilitating a subsequent call-back or detailed response. Furthermore, for special circumstances (e.g., if the product was not delivered as agreed), it might offer special discounts or compensation plans.

[0094] It should be noted that when the hosting agent provides customer service to the user, once it is determined that the hosting end conditions are met (such as successful transfer to manual customer service), the hosting end will be triggered and the agent will exit the current session (see Figure 3c 15). Further, as shown in FIG. Figure 3a As shown in link 3, after the hosting is completed, this application supports status tracking, which helps to ensure the quality of customer service, optimize customer service operation efficiency, and promptly discover and solve potential problems. Among them, status tracking may include but is not limited to anomaly detection and agent shutdown status management. Anomaly detection may include, for example, the reason why the agent exits the current session and determines whether the agent exits abnormally. Agent shutdown status management, for example, refers to tracking and managing the status of the agent customer service after the service is completed, ensuring that the agent resources are correctly released and subsequent operations are carried out in an orderly manner (such as arranging return visits to obtain customer service quality evaluations, etc.).

[0095] In addition to using the rule engine to detect the entire conversation content and conversation status when the intelligent agent provides customer service to the user, there is another scenario where the rule engine can still be used to detect the entire conversation content and / or conversation status when the call is transferred to manual customer service. To this end, this application also provides another online customer service method, the execution subject of which is the server. In addition, the online customer service method includes the following steps:

[0096] 201. In response to a user entering a conversation interface, determining a customer service type assigned to the user using a rule engine;

[0097] 202. Based on the customer service type, access a target customer service in the conversation interface to provide customer service to the user;

[0098] 203. Utilize the rule engine to detect the conversation content in the conversation interface, and when the conversation content matches a rule in the rule engine, perform an adaptation action according to the rule.

[0099] For the specific implementation of the above 201, please refer to the relevant content in other embodiments.

[0100] In the above 202-203, if the customer service type is intelligent customer service, an intelligent agent will be connected to provide customer service for the user. In the case of intelligent customer service, the specific implementation of the above 202-203 can be found in other embodiments, and will not be described in detail here.

[0101] If the customer service type is manual customer service, a manual customer service (commonly known as "customer service waiter") is connected to provide customer service to users. During the stage when the manual customer service provides customer service to users, the rule engine can also be used to monitor the content of the conversation, so that once the content of the conversation triggers a rule, the corresponding action in this rule can be executed.

[0102] For example, it is possible to monitor whether the content of the conversation complies with relevant laws and regulations and whether there are sensitive words. If potential violations and / or sensitive words are identified in the conversation content, the administrator can be immediately notified to intervene.

[0103] For example, if the rule engine detects that customer service personnel deviate from established communication strategies or fail to effectively resolve user issues, the rule engine can provide guidance or suggestions, such as recommending relevant knowledge base articles, to help customer service personnel find relevant answers faster.

[0104] Furthermore, the provided method may further comprise the following steps:

[0105] 204. Utilize the rule engine to monitor the conversation status in the conversation interface.

[0106] When the customer service type is intelligent customer service, for the specific implementation of the above 204, please refer to the relevant content in other embodiments, and will not be described in detail here.

[0107] In the case where the customer service type is artificial intelligence customer service, for example, during manual customer service, if the manual customer service fails to handle the user's problem in a timely manner due to certain factors, the rule engine can be used to determine the response information and send this response information to the user in the tone of the manual customer service, or it can also trigger the switch to the intelligent agent to temporarily provide the user with a response, etc., so as to ensure the consistency and coherence of the user experience.

[0108] The method provided in this embodiment may include other steps in addition to the steps given above. For details on the specific steps that may be included in this embodiment and the implementation of each step, please refer to the relevant content in other embodiments.

[0109] In addition to the above, this application also provides the following online customer service methods, each of which is performed by the server. Among them, this application also provides an online customer service method including the following steps:

[0110] 301. Using a rules engine, determine the customer service type assigned to the user entering the conversation interface;

[0111] 302. When the customer service type is intelligent customer service, an intelligent agent is connected to the conversation interface to provide customer service to the user;

[0112] 303. Utilize the rule engine to detect the conversation content in the conversation interface, so as to perform an adaptation action according to the rule when the conversation content matches the rule in the rule engine.

[0113] The method provided in this embodiment may include other steps in addition to the steps given above. For details on the specific steps that may be included in this embodiment and the implementation of each step, please refer to the relevant content in other embodiments.

[0114] Furthermore, another online customer service method provided by this application includes the following steps:

[0115] 401. Using an agent to provide customer service to users entering the conversation interface;

[0116] 402. Utilize the rule engine and the agent to collaboratively control the customer service conversation status in the conversation interface.

[0117] The method provided in this embodiment may include other steps in addition to the steps given above. For details on the specific steps that may be included in this embodiment and the implementation of each step, please refer to the relevant content in other embodiments.

[0118] It should be noted that in some of the processes described in the above embodiments and the accompanying drawings, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0119] This application also provides apparatuses corresponding to various method embodiments. Specifically, they are as follows:

[0120] An exemplary embodiment of the present application provides an online customer service device, comprising: a determination module, a provision module, and a detection module. The determination module is configured to determine a suitable customer service representative for a user. The provision module is configured to provide customer service to the user using a preset model when the suitable customer service representative is an intelligent customer service representative. The detection module is configured to detect the content of conversations in a conversation interface and to switch customer service representatives if the service provided by the intelligent customer service representative is abnormal.

[0121] Furthermore, when used to detect conversation content in a conversation interface, the detection module may be specifically configured to: utilize a rules engine to detect the conversation content; and when the rules engine detects that the conversation content matches at least one rule in a preset rule set, perform an adaptation action based on the at least one rule to switch the customer service type. The preset rule set is pre-installed in the rules engine, and the conversation content includes the conversation information input by the user and the response information output by the preset model.

[0122] Furthermore, the device also includes: a collaborative module, which is used to use the preset model and the rule engine to collaboratively control the customer service dialogue state in the conversation interface.

[0123] Furthermore, the collaborative module, when used to collaboratively control the customer service conversation state in the conversation interface using the preset model and the rule engine, includes at least one of the following: upon recognizing that the preset model is unable to respond to the conversation message input by the user, triggering the rule engine to execute one or more of the following actions according to relevant rules in the preset rule set: transferring the call to a human customer service representative, or having the rule engine act as a customer service representative and provide a response to the conversation message. During the transfer to the human customer service representative, the preset model continues to output a response to the conversation message.

[0124] Furthermore, the above-mentioned determination module, when used to determine the customer service adapted for the user, is specifically used to: utilize the rule engine to determine the customer service adapted for the user based on the preset rule set.

[0125] Furthermore, the above-mentioned providing module, when used to provide customer service to the user using a preset model, includes: determining an intelligent agent suitable for the user; using the intelligent agent to provide customer service to the user; wherein, the intelligent agent is constructed based on the preset model and the retrieval enhancement model; the retrieval enhancement model is used to query an external knowledge base based on the input of the intelligent agent to obtain relevant external knowledge, and combine the external knowledge with the input to obtain enhanced information; the preset model is used to generate output based on the enhanced information.

[0126] Another exemplary embodiment of the present application further provides an online customer service device, comprising: a determination module, a provision module, and a detection module. The determination module is configured to, in response to a user entering a conversation interface, determine a customer service type to assign to the user using a rules engine. The provision module is configured to, based on the customer service type, connect a target customer service representative to the conversation interface to provide customer service to the user. The detection module is configured to utilize the rules engine to detect the conversation content in the conversation interface, and when the conversation content matches a rule in the rules engine, perform an adaptation action based on the rule.

[0127] Another exemplary embodiment of the present application further provides an online customer service device, comprising: a determination module, a provision module, and a detection module. The determination module is configured to utilize a rules engine to determine the customer service type assigned to a user entering a conversation interface; the provision module is configured to, when the customer service type is intelligent customer service, connect an intelligent agent to the conversation interface to provide customer service to the user. The detection module is configured to utilize the rules engine to detect the conversation content in the conversation interface, and when the conversation content matches a rule in the rules engine, perform an adaptation action based on the rule.

[0128] Another exemplary embodiment of the present application further provides an online customer service device, comprising: a provisioning module and a collaboration module. The provisioning module is configured to utilize an agent to provide customer service to a user entering a conversation interface. The collaboration module is configured to utilize a rules engine and the agent to collaboratively control the customer service conversation state within the conversation interface.

[0129] What needs to be explained here about the above-mentioned devices is that the above-mentioned devices can implement the technical solutions described in the above-mentioned corresponding method embodiments. The specific implementation principles of the above-mentioned modules or units can be found in the relevant contents of the above-mentioned corresponding method embodiments, and will not be described in detail here.

[0130] This application also provides an online customer service system, comprising a client and a server. The client is configured to display a conversation interface in response to a user's conversation request; the conversation interface is configured to display user-related conversation content. The server is configured to implement the method embodiments provided above.

[0131] The client can be, but is not limited to, various terminal devices such as smartphones, laptops, tablets, smart wearable devices, etc.

[0132] The server can be a server, a service cluster, a virtual server or a cloud, etc., and this embodiment does not specifically limit this. The server provides corresponding functional services to the client, such as intelligent customer service. Specifically, the server is a customer service workstation under an e-commerce platform or website. The customer service workstation includes but is not limited to Figure 3a The hosting center, intelligent agent center, customer service assistant, manual customer service system, etc. shown in the figure.

[0133] It should be noted that the online customer service system (intelligent customer service system) provided by this application is optimized to adopt a concurrent decision-making system solution that integrates a large model (such as an LLM large model) and a rule engine. Of course, in other embodiments, in addition to adopting a large model (such as an LLM large model) for the construction of an intelligent customer service system, in certain specific scenarios, it is also possible to optimize for certain specific scenarios by combining different small models (such as deep learning models such as CNN). In addition, the large model can also cooperate with other systems such as expert systems and can also be used to build an intelligent customer service system. These solutions can also avoid the problems caused by the use of a single large model and improve the service effect of intelligent customer service.

[0134] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 6 As shown, the electronic device includes a memory 61 and a processor 62. The memory 61 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on the electronic device, data structures, contact data, phone book data, messages, images, videos, etc.

[0135] The processor 62 is coupled to the memory 61 and is used to execute the computer program in the memory 61 to implement the various method embodiments provided in this application.

[0136] Further, if Figure 6 As shown, the electronic device also includes: a communication component 63, a display 64, a power component 65, an audio component 66 and other components. Figure 6 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 6 In addition, Figure 6The components in the dotted box are optional components, not mandatory components, and the specific components may depend on the product form of the electronic device. The electronic device of this embodiment can be implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone or an IOT device, or a server device such as a conventional server, a cloud server or a server array. If the working node of this embodiment is implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, etc., it can include Figure 6 If the electronic device of this embodiment is implemented as a conventional server, cloud server or server array and other server-side devices, it may not include Figure 6 Components within the dotted box.

[0137] The above-mentioned memory can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0138] The communication component is configured to facilitate wired or wireless communication between the device in which the communication component resides and other devices. The device in which the communication component resides can access a wireless network based on a communication standard, such as a 2G, 3G, 4G / LTE, 5G, or other mobile communication network, or a combination thereof. In an exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel.

[0139] The above-mentioned display includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor can not only sense the boundary of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0140] The power supply assembly provides power to various components of the device in which the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which the power supply assembly is located.

[0141] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC), and when the device where the audio component is located is in an operating mode, such as call mode, recording mode, and voice recognition mode, the microphone is configured to receive external audio signals. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0142] Accordingly, an embodiment of the present application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to implement the steps in the above method embodiment. The computer-readable storage medium includes volatile or non-volatile or a combination thereof, and may be removable or non-removable. Examples of computer-readable storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, magnetic cassette, tape disk storage or other magnetic storage device or any other non-transmission medium.

[0143] Accordingly, an embodiment of the present application further provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, the processor is enabled to implement the steps in the above-mentioned method embodiment. It should be understood that each process or a combination of multiple processes in the above-mentioned method flow can be implemented by a computer program or instruction. In addition, these computer programs or instructions can be applied to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device, so that the processor of the general-purpose computer, the special-purpose computer, the embedded processor or other programmable data processing device can be implemented as a device for implementing the corresponding functions in the above-mentioned method embodiment.

[0144] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0145] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. An online customer service method, characterized in that: include: Determine suitable customer service for users; When the adapted customer service is an intelligent customer service, a preset model is used to provide customer service to the user; The conversation content in the conversation interface is detected to switch customer service when the service provided by the intelligent customer service is abnormal.

2. The method according to claim 1, characterized in that Detect the conversation content in the conversation interface, including: Using a rule engine to detect the content of the conversation; When the rule engine detects that the conversation content matches at least one rule in a preset rule set, an adaptation action is performed according to the at least one rule to switch the customer service type; The conversation content includes the conversation information input by the user and the response information output by the preset model.

3. The method according to claim 2, characterized in that Also includes: The preset model and the rule engine are used to collaboratively control the dialog state in the conversation interface.

4. The method according to claim 3, characterized in that Utilizing the preset model and the rule engine to collaboratively control the customer service conversation state in the conversation interface includes at least one of the following: When it is determined that the preset model cannot respond to the dialogue message input by the user, the rule engine is triggered, and the rule engine performs one or more of the following actions according to the corresponding rules in the preset rule set: transferring the call to a human customer service representative, or having the rule engine act as a customer service representative to provide a response to the dialogue message; During the transfer to manual customer service, the preset model continues to output response information for the dialogue information.

5. The method according to any one of claims 2 to 4, characterized in that Determining a suitable customer service representative for the user includes: Using the rule engine, determining customer service suitable for the user is performed based on the preset rule set.

6. The method according to any one of claims 1 to 4, characterized in that Providing customer service to the user using a preset model, including: determining an agent suitable for the user; Using the agent to provide customer service to the user; Among them, the intelligent agent is constructed based on the preset model and the retrieval enhancement model; the retrieval enhancement model is used to query the external knowledge base according to the input of the intelligent agent to obtain relevant external knowledge, and combine the external knowledge with the input to obtain enhanced information; the preset model is used to generate output based on the enhanced information.

7. An online customer service method, characterized in that: include: In response to a user entering a conversation interface, determining, using a rules engine, an assigned customer service type for the user; Based on the customer service type, accessing a target customer service in the conversation interface to provide customer service to the user; The rule engine is used to detect the conversation content in the conversation interface, so that when the conversation content matches the rules in the rule engine, an adaptation action is performed according to the rules.

8. An online customer service method, characterized in that: include: Use the rules engine to determine the type of customer service assigned to users entering the conversation interface; When the customer service type is intelligent customer service, an intelligent agent is connected to the conversation interface to provide customer service to the user; The rule engine is used to detect the conversation content in the conversation interface, so that when the conversation content matches the rules in the rule engine, an adaptation action is performed according to the rules.

9. An online customer service method, characterized in that: include: Leverage intelligent agents to provide customer service to users entering the conversational interface; The rule engine and the agent are used to collaboratively control the customer service conversation status in the conversation interface.

10. An online customer service system, characterized in that: include: The client is configured to display a conversation interface in response to a user's conversation request; the conversation interface is configured to display conversation content related to the user; The server is used to implement the method described in any one of claims 1 to 9 above.

11. An electronic device, characterized in that: include: memory and a processor, wherein The memory is used to store programs; The processor is coupled to the memory, and is configured to execute the program stored in the memory to implement the method according to any one of claims 1 to 9.

12. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a computer, the method according to any one of claims 1 to 9 can be implemented.

13. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 9 is implemented.

Images