Consultation request processing method and device and storage medium

Abstract

The invention provides a consultation request processing method and device and a storage medium, and the method comprises the steps: obtaining a session context of a target consultation request in response to a manual processing instruction for the target consultation request; according to the session context, identifying whether a current to-be-processed problem in the session of the target consultation request is a preset type of problem; if the current to-be-processed problem is the preset type of problem, using a preset large model to generate reply information corresponding to the current to-be-processed problem; and if the current to-be-processed problem is not the preset type of problem, accessing the session of the target consultation request to a manual customer service link. According to the method and the device, in the manual service stage of the consultation request, the known preset type of consultation is processed by fully utilizing the capability of the AI large model, and the problem that the large model cannot be processed is automatically switched to manual processing, so that the service experience of a user is ensured, and the resource utilization rate is improved.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a consultation request processing method, device and storage medium. Background Art

[0002] With the development of the Internet, users are increasingly inclined to purchase goods online through e-commerce platforms. Traditional e-commerce platforms generally provide users with a search function so that users can search for information about goods of interest through the e-commerce platform.

[0003] In e-commerce scenarios, e-commerce customer service is an important bridge for communication with users. Although the problem-solving ability of robot customer service is already very strong, the service of human customer service for high-quality and high-value users cannot be completely replaced. A large number of human customer service still plays a very important role in the task of ensuring member experience. In the traditional human customer service service model, the customer service assistant can serve member users who come in for consultation from the customer service robot from the online workbench. During the service process, due to the different personal experience, years of work, and complexity of the task scenarios handled by the customer service assistant, usually one customer service assistant can serve 1 to 6 member users at the same time, and the service capacity is limited. Due to the limitations of service capacity and upper limit, when the number of consultations increases, a large number of customer service assistants are usually required to support it. Not only is the labor cost huge, but the efficiency is also low, and it can no longer meet the growing user consultation needs. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to provide a consultation request processing method, device and storage medium, which can realize the screening and diversion of specific user consultation issues in the manual service stage of the consultation request, and make full use of the capabilities of the AI ​​big model to handle known preset types of consultations, reduce the amount of manual customer service involved, and improve the efficiency of consultation request processing. For problems that cannot be handled by the big model, they are automatically transferred to manual processing, which not only ensures the user's service experience but also improves resource utilization.

[0005] In a first aspect, an embodiment of the present application provides a method for processing consultation requests, comprising: in response to a manual processing instruction for a target consultation request, obtaining a session context of the target consultation request; based on the session context, identifying whether a current problem to be processed in the session of the target consultation request is a preset type of problem; if the current problem to be processed is a preset type of problem, using a preset large model to generate reply information corresponding to the current problem to be processed; if the current problem to be processed is not a preset type of problem, connecting the session of the target consultation request to a manual customer service link.

[0006] In one embodiment, the step of triggering a manual processing instruction for a target consultation request includes: in response to a user's consultation request, obtaining user characteristics of the user; determining whether the user is a target user based on the user characteristics; if the user is a target user, triggering a manual processing instruction for the consultation request.

[0007] In one embodiment, the step of triggering a manual processing instruction for a target consultation request includes: in response to a user's consultation request, obtaining a session context corresponding to the consultation request; identifying, based on the session context, the consultation scenario category to which the consultation request currently belongs; if the consultation scenario category is a target scenario category that requires manual customer service intervention, triggering a manual processing instruction for the consultation request.

[0008] In one embodiment, in response to a consultation request from a user, obtaining a session context corresponding to the consultation request includes: in response to the consultation request from the user, connecting the consultation request to a robot customer service link; and obtaining the session context between the robot customer service and the user.

[0009] In one embodiment, identifying the consultation scenario category to which the consultation request currently belongs based on the conversation context includes: inputting the conversation context into a preset large model, the preset large model is used to analyze the conversation context, and outputting the consultation scenario category to which the consultation request currently belongs.

[0010] In one embodiment, after connecting the conversation of the target consultation request to the manual customer service link, the method further includes: generating reply information of the current pending issue according to the input information of the manual customer service.

[0011] In one embodiment, the generation of reply information for the current pending issue based on the input information of the manual customer service includes: responding to the manual customer service's auxiliary request for the current pending issue, generating auxiliary information corresponding to the auxiliary request through a preset large model; obtaining the input information determined by the manual customer service based on the auxiliary information; and responding to the manual customer service's reply instruction to the current pending issue, issuing the input information as reply information for the current pending issue.

[0012] In one embodiment, the use of a preset big model to generate reply information corresponding to the current problem to be processed includes: identifying the user's current intention based on the conversation context; routing the current intention to the corresponding big model proxy module, so that the big model proxy module generates matching reply information according to the current problem to be processed.

[0013] In one embodiment, identifying the user's current intention based on the conversation context includes: inputting the conversation context into a pre-trained intention recognition model, the intention recognition model is used to analyze the conversation context and identify the user's current intention, wherein the intention recognition model is trained based on a preset conversation sample set, and the preset conversation samples are marked with corresponding user intentions.

[0014] In one embodiment, after using the preset large model to generate reply information corresponding to the current pending problem, it also includes: detecting the current session context of the target consultation request, and using the preset large model to determine whether the current pending problem is solved according to the current session context; if the current pending problem is not solved, connecting the session of the target consultation request to the manual customer service link, and generating reply information for the current pending problem based on the input information of the manual customer service.

[0015] In one embodiment, after connecting the conversation of the target consultation request to the manual customer service link and generating the reply information of the current problem to be processed according to the input information of the manual customer service, it also includes: responding to the hosting instruction of the manual customer service, using the preset large model to process the target consultation request.

[0016] In a second aspect, an embodiment of the present application provides a method for processing consultation requests regarding product information, comprising: in response to a manual processing instruction regarding a consultation request for product information, obtaining a session context of the consultation request for product information; based on the session context, identifying whether a current pending issue in the session of the consultation request for product information is a preset type of issue; if the current pending issue is a preset type of issue, using a preset large model to generate reply information corresponding to the current pending issue; if the current pending issue is not a preset type of issue, connecting the session of the consultation request for product information to a manual customer service link, and generating reply information for the current pending issue based on the input information of the manual customer service.

[0017] In a third aspect, an embodiment of the present application provides a consultation request processing device, including:

[0018] An acquisition module, configured to acquire a session context of a target consultation request in response to a manual processing instruction for the target consultation request;

[0019] An identification module, configured to identify, based on the session context, whether a current question to be processed in the session of the target consultation request is a question of a preset type;

[0020] A first processing module is used to generate reply information corresponding to the current problem to be processed using a preset large model if the current problem to be processed is a preset type of problem;

[0021] The second processing module is used to connect the conversation of the target consultation request to a manual customer service link if the current problem to be processed is not a problem of the preset type.

[0022] In one embodiment, it also includes: a trigger module, which is used to obtain user characteristics of the user in response to the user's consultation request; determine whether the user is a target user based on the user characteristics; if the user is a target user, trigger a manual processing instruction for the consultation request.

[0023] In one embodiment, it also includes: a trigger module, which is used to respond to the user's consultation request and obtain the session context corresponding to the consultation request; identify the consultation scenario category to which the consultation request currently belongs based on the session context; if the consultation scenario category is a target scenario category that requires manual customer service intervention, trigger a manual processing instruction for the consultation request.

[0024] In one embodiment, the trigger module is further used to respond to the consultation request of the user, connect the consultation request to the robot customer service link; and obtain the conversation context between the robot customer service and the user.

[0025] In one embodiment, the identification module is used to input the conversation context into a preset large model, and the preset large model is used to analyze the conversation context and output the consultation scenario category to which the consultation request currently belongs.

[0026] In one embodiment, the second processing module is used to generate reply information of the current problem to be processed according to the input information of the manual customer service after the session of the target consultation request is connected to the manual customer service link.

[0027] In one embodiment, the second processing module is used to respond to the auxiliary request of the manual customer service for the current problem to be processed, generate auxiliary information corresponding to the auxiliary request through a preset large model; obtain the input information determined by the manual customer service based on the auxiliary information; and respond to the reply instruction of the manual customer service to the current problem to be processed, send the input information as reply information of the current problem to be processed.

[0028] In one embodiment, the first processing module is used to identify the user's current intention based on the conversation context; and route the current intention to the corresponding large model proxy module so that the large model proxy module generates matching reply information based on the current problem to be processed.

[0029] In one embodiment, the first processing module is used to input the conversation context into a pre-trained intent recognition model, and the intent recognition model is used to analyze the conversation context and identify the user's current intent, wherein the intent recognition model is trained based on a preset conversation sample set, and the preset conversation samples are marked with corresponding user intents.

[0030] In one embodiment, it also includes: a switching module, which is used to detect the current session context of the target consultation request after the reply information corresponding to the current pending problem is generated using the preset large model, and use the preset large model to determine whether the current pending problem is solved according to the current session context; if the current pending problem is not solved, connect the session of the target consultation request to the manual customer service link, and generate the reply information of the current pending problem according to the input information of the manual customer service.

[0031] In one embodiment, the transfer module is also used to respond to the hosting instructions of the manual customer service, and use the preset large model to process the target consultation request after connecting the conversation of the target consultation request to the manual customer service link and generating the reply information of the current problem to be processed according to the input information of the manual customer service.

[0032] In a fourth aspect, an embodiment of the present application provides an electronic device, including:

[0033] at least one processor; and

[0034] a memory communicatively coupled to the at least one processor;

[0035] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the electronic device to execute the method described in any one of the above aspects.

[0036] In a fifth aspect, an embodiment of the present application provides a cloud device, including:

[0037] at least one processor; and

[0038] a memory communicatively coupled to the at least one processor;

[0039] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the cloud device to execute the method described in any one of the above aspects.

[0040] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions. When a processor executes the computer-executable instructions, the method described in any one of the above aspects is implemented.

[0041] In a seventh aspect, an embodiment of the present application provides a computer program product, including a computer program, which implements the method described in any of the above aspects when executed by a processor.

[0042] The consultation request processing method, device and storage medium provided in the embodiment of the present application, for consultation requests that require manual processing, when entering the manual service stage, identify and analyze the session context of the target consultation request, identify whether the current pending issue in the session of the target consultation request is a known preset type of consultation issue, and if it is a preset type of consultation issue, the pending issue can be handed over to the preset large model for processing, so that the preset large model generates the reply information corresponding to the current pending issue. If the pending issue is not a preset type of consultation issue, the real manual customer service will take over the session, so that the specific issues consulted by the user will be screened and diverted in the manual service stage of the consultation request, and the capabilities of the AI ​​large model will be fully utilized to handle the known preset type of consultation, thereby reducing the amount of participation of manual customer service and improving the efficiency of consultation request processing. For issues that cannot be handled by the large model, they will be automatically transferred to manual processing, which not only guarantees the user's service experience, but also improves resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application. Obviously, the drawings described below are some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.

[0044] Figure 1 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0045] Figure 2 A schematic diagram of an application scenario of a consultation request processing system provided in an embodiment of the present application;

[0046] Figure 3 A flowchart of a method for processing a consultation request provided in an embodiment of the present application;

[0047] Figure 4 A schematic diagram of a typical consultation request processing service process provided in an embodiment of the present application;

[0048] Figure 5A A schematic diagram of a consultation request service link reshaped by AI provided in an embodiment of the present application;

[0049] Figure 5B A schematic diagram of an AI workbench interface provided for an application embodiment;

[0050] Figure 6 A schematic diagram of a scenario link for AI processing consultation requests provided in an embodiment of the present application;

[0051] Figure 7 A full-link schematic diagram of a consultation request information processing method provided for this application;

[0052] Figure 8 A schematic diagram of an interface of manual customer service provided for this application;

[0053] Fig. 9 A schematic diagram of an interface of an AI workbench provided in an embodiment of the present application;

[0054] Fig.10 A schematic diagram of the distribution ratio of a consultation request provided in an embodiment of the present application;

[0055] Fig.11 A flowchart of a method for processing an inquiry request for product information provided in an embodiment of the present application;

[0056] Fig.12 A schematic diagram of the structure of a consultation request processing device provided in an embodiment of the present application;

[0057] Fig.13 A schematic diagram of the structure of a cloud device provided in an embodiment of the present application.

[0058] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0059] Here, exemplary embodiments will be described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0060] The term "and / or" in this article is used to describe the association relationship of associated objects, specifically indicating that there may be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0061] It should be noted that 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 this 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 must 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.

[0062] In order to clearly describe the technical solution of the embodiment of the present application, the terms involved in the present application are first defined:

[0063] AI: Artificial Intelligence.

[0064] LLM: Large Language Model, a natural language processing model based on deep learning, which can learn the grammar and semantics of natural language and generate human-readable text. Large language models can handle a variety of natural language tasks, such as text classification, question answering, dialogue, etc., and are an important path to artificial intelligence.

[0065] API: Application Programming Interface, application programming interface.

[0066] Prompt: A prompt is a natural language input, similar to a command or instruction, that lets the AI ​​model know what it needs to do.

[0067] IM: Instant Messaging.

[0068] Online customer service: generally refers to customer service interacting with members through text online IM to resolve members' demands.

[0069] Copilot: An intelligent assistance product in the customer service workbench that can provide customer service staff with intelligent speech, solution recommendations, intelligent summaries and other auxiliary capabilities.

[0070] Agent: A large model agent program with advanced capabilities such as perception, reasoning, and memory.

[0071] Small model: Compared with the general large model, the small model is a specific model designed and trained by the algorithm to handle special tasks.

[0072] HumanCopilot: An intelligent artificial service design with AI as the service subject and humans as the auxiliary.

[0073] ID: Identity document, ID number, account number, unique code.

[0074] ChatGLM: Chat Global Language Model, chat global model.

[0075] SOP: Standard Operating Procedure, standard operating procedure.

[0076] GPT: Generative Pre-Trained Transformer, a deep learning model for text generation based on Internet-based training with available data.

[0077] RAG: Retrieval-Augmented Generation, retrieval enhanced generation.

[0078] ISO: Digitalization of service processes.

[0079] BERT: Bidirectional Encoder Representations from Transformers, a pre-trained language model based on a multi-layer Transformer encoder.

[0080] StructBERT: StructBERT model, a pre-trained model for natural language processing, is an improved version of the BERT model.

[0081] FAQ: Frequently Asked Questions, frequently asked questions.

[0082] The consultation request processing method of the embodiment of the present application can be applied to any field that needs to process information requests.

[0083] Take the scenario where the customer service of an e-commerce platform handles user consultation requests as an example. In the e-commerce scenario, e-commerce customer service is an important bridge for communication with users. Although the problem-solving ability of AI robot customer service is already very strong, the service of human customer service for high-quality and high-value users cannot be completely replaced. The huge number of human customer service still plays a very important role in the task of ensuring member experience. In the traditional human customer service service model, the customer service assistant can serve member users who come in for consultation from the customer service robot from the online workbench. During the service process, due to the different personal experience, years of work, and complexity of the task scenario of the customer service assistant, usually one customer service assistant can serve 1 to 6 member users at the same time, and the service capacity is limited. However, due to the limitations of service capacity and upper limit, when the number of consultations increases, a large number of customer service assistants are usually required to support it. Not only is the labor cost huge, but the efficiency is also low, and it can no longer meet the growing user consultation needs.

[0084] In an optional embodiment, the customer service system can provide tools similar to SOP / ISO through product capabilities to assist customer service staff in providing standardized solutions, but factors such as individual differences in customer service staff (skills, experience, ability, emotions, personality) and the product design of the customer service system itself will still affect the final service quality. In addition, in actual scenarios, many users do not generate any valuable consultations when they come in (usually they can be solved through document questions and answers or intelligent speech), but they will occupy the manual customer service channel, causing unnecessary losses.

[0085] With the development of artificial intelligence technology, large model comprehension capabilities, logical reasoning capabilities, and multimodal capabilities have been greatly improved. AI can correctly understand member users' consulting questions in most simple service scenarios, and can self-service call system capabilities (ISO, SOP, actions) to solve member users' consulting problems.

[0086] However, in the intelligent era, manual customer service still has the following advantages:

[0087] 1. Human services are still irreplaceable because people can:

[0088] a. Bring better experience to high-value users.

[0089] b. Deal with risks.

[0090] c. Proactive service.

[0091] d. Provide emotional value.

[0092] In summary, how to create a scalable manual service model and product design based on AI technology to support the ever-increasing capabilities of large models has become a problem worthy of study.

[0093] In order to solve at least one of the above problems, the embodiment of the present application provides a consultation request processing solution. For consultation requests that require manual processing, when entering the manual service stage, the session context of the target consultation request is identified and analyzed to identify whether the current pending issue in the session of the target consultation request is a known preset type of consultation issue. If it is a preset type of consultation issue, the pending issue can be handed over to the preset large model for processing, so that the preset large model generates reply information corresponding to the current pending issue. If the pending issue is not a preset type of consultation issue, the real manual customer service will take over the session, and the reply information of the current pending issue will be generated according to the input information of the manual customer service. In this way, the specific issues consulted by the user will be screened and diverted in the manual service stage of the consultation request. For the known preset type of consultation, the ability of the AI ​​large model is fully utilized to handle it, reducing the amount of participation of manual customer service, improving the efficiency of consultation request processing, and automatically transferring the issues that cannot be handled by the large model to manual processing, which not only guarantees the user's service experience, but also improves resource utilization.

[0094] Specifically, by reconstructing service processes and manual service workbenches based on AI, simple and repetitive consulting questions can be solved by AI, and manual customer service focuses on providing better proactive services for high-value users and high-risk scenarios.

[0095] Some embodiments of the present application are described in detail below in conjunction with the accompanying drawings. In the case where there is no conflict between the embodiments, the following embodiments and the features in the embodiments can be combined with each other. In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0096] like Figure 1 As shown, this embodiment provides an electronic device 1, including: at least one processor 11 and a memory 12, Figure 1 A processor is taken as an example. The processor 11 and the memory 12 are connected via a bus 10. The memory 12 stores instructions that can be executed by the processor 11, and the instructions are executed by the processor 11 so that the electronic device 1 can execute all or part of the process of the method in the following embodiment, so as to realize the screening and diversion of the specific questions consulted by the user in the manual service stage of the consultation request, and make full use of the capabilities of the AI ​​big model to handle the known preset type of consultation, reduce the participation of manual customer service, improve the efficiency of consultation request processing, and automatically transfer the problems that cannot be handled by the big model to manual processing, which not only ensures the user's service experience, but also improves resource utilization.

[0097] In one embodiment, the electronic device 1 may be a mobile phone, a tablet computer, a laptop computer, a desktop computer, or a large computing system composed of multiple computers.

[0098] Figure 2 A schematic diagram of an application scenario 200 of a consultation request processing system provided in an embodiment of the present application. Figure 2 As shown, the system includes: a server 210 and a terminal 220, wherein:

[0099] Server 210 may be a data platform that provides consultation request processing services, such as an e-commerce customer service platform. In actual scenarios, an e-commerce customer service platform may have multiple servers 210. Figure 2 In the figure, one server 210 is taken as an example.

[0100] The terminal 220 may be a computer, mobile phone, tablet or other device used by a user to log in to the e-commerce customer service platform. There may also be multiple terminals 220. Figure 2 Two terminals 220 are taken as an example for illustration.

[0101] The terminal 220 and the server 210 can transmit information via the Internet, so that the terminal 220 can access the data on the server 210. The terminal 220 and / or the server 210 can be implemented by the electronic device 1.

[0102] The consultation request processing solution of the embodiment of the present application can be deployed on the server 210, can also be deployed on the terminal 220, or can be deployed partially on the server 210 and partially on the terminal 220. In actual scenarios, it can be selected based on actual needs, and this embodiment does not limit it.

[0103] When the consultation request processing solution is fully or partially deployed on the server 210 , a calling interface may be opened to the terminal 220 to provide algorithm support to the terminal 220 .

[0104] The method provided in the embodiment of the present application can be implemented by executing the corresponding software code by the electronic device 1, and can be implemented by exchanging data with the server. The electronic device 1 can be a local terminal device. When the method is run on the server, the method can be implemented and executed based on the cloud interaction system, wherein the cloud interaction system includes a server and a client device.

[0105] In a possible implementation, the method provided in the embodiment of the present application provides a graphical user interface through a terminal device, wherein the terminal device can be the local terminal device mentioned above, or can be the client device in the cloud interaction system mentioned above.

[0106] Please see Figure 3 , which is a consultation request processing method of an embodiment of the present application, the method can be Figure 1 The electronic device 1 shown is used to perform and can be applied to Figure 2In the application scenario of consultation request processing shown in , the specific questions of user consultation can be screened and diverted in the manual service stage of the consultation request. For known preset types of consultation, the ability of AI big model is fully utilized to handle it, reducing the amount of manual customer service involved, improving the efficiency of consultation request processing, and automatically transferring to manual processing for problems that cannot be handled by the big model, which not only ensures the user's service experience, but also improves resource utilization. This embodiment takes terminal 220 as an execution end as an example, and the method includes the following steps:

[0107] Step 301: In response to a manual processing instruction for a target consultation request, a session context of the target consultation request is obtained.

[0108] In this step, the target consultation request may be a consultation request initiated by a user for a specific question, such as a consultation request initiated by a consumer user or a merchant user in an e-commerce scenario for a specific product or specific order information. In actual scenarios, the consultation request initiated by the user can be processed by a robot or manually, and the manual processing instruction can be actively triggered by the user who initiated the consultation request, or automatically triggered by the system. When a manual processing instruction for the target consultation request is detected, in response to the instruction, the session context of the target consultation request is first obtained, where the session context may include all session content from the initiation of the target consultation request to the current moment, so as to analyze the actual needs of the user according to the session context.

[0109] In one embodiment, the step of triggering a manual processing instruction for a target consultation request includes: in response to a user's consultation request, obtaining a user feature of the user. Determining whether the user is a target user based on the user feature. If the user is a target user, triggering a manual processing instruction for the consultation request.

[0110] In this embodiment, taking the user consultation scenario of the e-commerce platform as an example, when a user has questions about a certain product information or order information, a customer service consultation request is triggered. In response to the consultation request, the user characteristics of the user are first obtained. Here, the user characteristics include but are not limited to the user's identity characteristics, personality characteristics, membership characteristics and other information. Then, based on the user characteristics, it is determined whether the user is a specific target user. The specific characteristics of the target user can be set according to actual needs. For example, the target user can be a user who has registered as a member of the platform, or a user with a specified personality characteristic (such as a bad temper), or a user marked in the historical consultation record, a user whose consumption reaches a certain amount, etc. In actual scenarios, the consultation request of the target user has a great impact on the e-commerce platform and is a key service object in the consultation service. In order to ensure the consultation experience of the target user, if it is determined that the current consulting user is the target user, a manual processing instruction for the user's consultation request can be automatically triggered.

[0111] In one embodiment, the step of triggering a manual processing instruction for a target consultation request includes: in response to a user's consultation request, obtaining a session context corresponding to the consultation request. According to the session context, identifying the consultation scenario category to which the consultation request currently belongs. If the consultation scenario category is a target scenario category requiring manual customer service intervention, triggering a manual processing instruction for the consultation request.

[0112] In this embodiment, taking the user consultation scenario of the e-commerce platform as an example, when the user triggers a customer service consultation request for a certain product information or order information, the session context of the consultation request can be detected in real time to identify the consultation scenario category to which the consultation request currently belongs. It is determined whether the consultation scenario category to which the consultation request currently belongs is the target scenario category. Here, the target scenario category refers to a high-risk scenario category that requires manual customer service intervention, such as a scenario in which the user's speech in the conversation contains insults, threats and other uncivilized words, or a scenario in which the user wants to hurt himself or others. It is possible to pre-set which scenario categories require manual customer service intervention, and to determine whether it is the target scenario category by comparing the current consultation scenario category with the preset scenario category. It is also possible to directly use the preset large model to identify whether it is the target scenario category according to the session context. If the consultation scenario category is a target scenario category that requires manual customer service intervention, the system can automatically trigger a manual processing instruction for the consultation request, so that manual customer service can be intervened in time, which not only improves the consultation processing efficiency, but also improves the user's consultation experience.

[0113] In one embodiment, in response to a user's consultation request, obtaining a conversation context corresponding to the consultation request specifically includes: in response to the user's consultation request, connecting the consultation request to a robot customer service link. Obtaining a conversation context between the robot customer service and the user.

[0114] In this embodiment, taking the user consultation scenario of the e-commerce platform as an example, when a user triggers a customer service consultation request for a certain product information or order information, the preset robot customer service can first handle the user's consultation request. The robot customer service will provide users with conversational intelligent services to save labor costs and improve efficiency. In the process of robot customer service handling user consultations, the conversation content between the robot customer service and the user is detected in real time to obtain the conversation context. The conversation context can refer to the chat corpus between the robot customer service and the user. By setting up conversation content detection in the robot service stage, it is possible to identify in real time whether manual customer service intervention is required, thereby improving the system's ability to provide proactive services.

[0115] In one embodiment, the consultation scenario category to which the consultation request currently belongs is identified based on the conversation context, including: inputting the conversation context into a preset large model, the preset large model is used to analyze the conversation context, and outputting the consultation scenario category to which the consultation request currently belongs.

[0116] In this embodiment, the AI ​​big model can be used to monitor the process of the robot customer service handling user inquiries, and the conversation context between the robot customer service and the user can be input into the preset big model, so that the preset big model can analyze the conversation context and identify the consultation scenario category to which the consultation request currently belongs, thereby making full use of the capabilities of the AI ​​big model and improving processing efficiency.

[0117] like Figure 4 As shown, it is a typical consultation request processing service process diagram provided by an embodiment of the present application. In a typical customer service process, the user who initiates the consultation request will first enter the robot customer service, and the robot customer service will provide the user with a conversational intelligent service. When the robot cannot solve the problem or the user has a strong demand for manual service, the user can apply to contact the manual customer service from the robot customer service, and then a manual attendant will establish a connection with the user from the customer service workbench, and can use tools such as customer service assistants and Copilot (an AI assistant) to provide member users with more accurate services and problem-solving capabilities. If the problem in some scenarios still cannot be solved through online chat, you can record the work order and then follow up and solve it through offline services.

[0118] However, in the robot service stage, although it has the intelligent service capabilities provided by AI+rules, due to the limitations of AI itself and the uncertainty of robot services, it is still inevitable that the member user experience will be damaged in the self-service stage. Especially for high-value and high-risk users, the system still needs to have the ability to take the initiative and cannot wait for the final distribution through the layers of funnels.

[0119] like Figure 5A As shown in the figure, it is a schematic diagram of a consultation request service link reshaped by AI provided in an embodiment of the present application, in order to solve Figure 4 In order to solve the problem of the embodiment in the embodiment, an observation panel can be provided in the robot service process to detect the user characteristics and / or conversation context of the online consultation, screen out high-value target users, or identify the target scenario categories that have high risks in the consultation session and require manual customer service intervention, so that the consultation problems of these users can be actively taken over by the system and handled by manual customer service in a timely manner, reducing the loss of service experience. At this stage, the characteristics of the AI ​​big model can be used to effectively analyze the conversation context in real time, and identify service experience, risk, value, emotion and other scenarios.

[0120] like Figure 5B As shown, it is a schematic diagram of the interface of an AI workbench provided in an embodiment of the application. During the robot service process, the scene category recognition results of multiple consulting users can be displayed in real time in the interface of the AI ​​workbench, so as to facilitate real-time observation of the actual conditions of different users. Figure 5B Take the observation of 4 users as an example, among which:

[0121] User 1's consultation type is "Exchange / Return", and the current progress is that it has been determined to be a product quality issue. The conversation content sent by the user at 00:50 is "Hello, let me urge the logistics and delivery first". It is identified that the consultation scenario category of user 1's consultation request is currently "High-risk user", which belongs to the target scenario category, and the user feature is "angry". It can automatically connect to the manual customer service channel for processing, reducing the user's experience loss.

[0122] User 2's inquiry type is "product quality issue" and the current processing progress is "The consumer wants to return the product and get a refund due to product quality issues. Can the customer service assist him in applying for after-sales service and explain the rights protection steps?" The user characteristics of User 2 are identified as "member" users. Member users are high-value users for e-commerce platforms and are therefore target users. They can be automatically connected to the manual customer service channel for processing to improve user experience.

[0123] User 3's consultation type is "repeated calls, upgraded processing", and the user's current demand is "consumers want to understand the after-sales processing flow". User 3 is not a target user, and the current consultation scenario category is not a target scenario category. It belongs to a consultation scenario that can be completed by AI, so the robot customer service can continue to process the consultation content, or the AI ​​big model can process the consultation content, saving the participation of manual customer service.

[0124] User 4's consultation type is "Exchange / Return", and the current progress is that it has been determined to be a product quality issue. It is identified that the consultation scenario category of user 4's consultation request is currently "High-risk User", which belongs to the target scenario category, and can be automatically connected to the manual customer service channel for processing. Reduce user experience loss.

[0125] Step 302: According to the conversation context, identify whether the current question to be processed in the conversation of the target consultation request is a question of a preset type. If yes, proceed to step 303; otherwise, proceed to step 304.

[0126] In this step, after entering the manual service stage, the conversation context of the target consultation request is identified and analyzed to identify whether the current pending question in the conversation of the target consultation request is a known preset type of consultation question. Preset type questions may include some questions that already have fixed answers or standard response processes, such as knowledge questions, simple repetitive questions, etc. These types of questions can directly call the relevant knowledge base or preset question and answer database to query the corresponding answers, so there is no need for manual customer service to handle them. If the current pending question is a preset type question, go to step 303 and it will be directly handled by the preset large model, otherwise go to step 304 and it will be handled by manual customer service.

[0127] Step 303: If the current problem to be processed is a problem of a preset type, a preset large model is used to generate reply information corresponding to the current problem to be processed.

[0128] In this step, the preset types of questions are generally repetitive, trivial, and worthless conversations and questions. The AI ​​hosting mode can be used. The preset large model combines the relevant knowledge base or the preset question and answer database to query the corresponding answer, and generates the reply information corresponding to the current pending question based on the query results, reducing the participation of manual customer service and improving processing efficiency. The reply information here includes but is not limited to text information, image information, audio and video information, operation processing, etc.

[0129] In one embodiment, step 303 may specifically include: identifying the user's current intention according to the conversation context, routing the current intention to the corresponding large model proxy module, so that the large model proxy module generates matching reply information according to the current problem to be processed.

[0130] In this embodiment, in the AI ​​hosting mode, AI can replace manual customer service to generate some simple words and solutions, and automatically reply without manual participation. It can identify the user's current intention based on the conversation context corpus, and then route to each corresponding large model agent module Agent according to the user's current intention. The large model agent module is used to implement specific information query functions, such as credential recognition module, word generation module, knowledge base RAG, etc. These large model agent modules can match the corresponding query results according to the user's current intention, and can generate matching reply information for the current pending problem based on the query results through the preset large model. Improve the accuracy and effectiveness of the reply information.

[0131] The great advantage of the AI ​​hosting model is "cost reduction and efficiency improvement". It can use the powerful processing power of AI to improve the parallel service capabilities of manual services, while freeing customer service staff from problems that AI can handle. On the other hand, the AI ​​model is evolvable. It can be gradually enhanced with the development of models and intelligent capabilities, and take on more and more service processes that originally required the participation of manual customer service.

[0132] In one embodiment, identifying a user's current intention based on a conversation context includes: inputting the conversation context into a pre-trained intention recognition model, the intention recognition model is used to analyze the conversation context and identify the user's current intention, wherein the intention recognition model is trained based on a preset conversation sample set, and the preset conversation samples are marked with corresponding user intentions.

[0133] In this embodiment, the intent recognition model can be a small model specially trained for consulting scenarios, such as training a StructBERT small model for intent recognition. Taking the e-commerce scenario as an example, when training the model, consulting session samples of different users in the e-commerce scenario can be collected, and the corresponding user intent is marked in each session sample to form a preset session sample set. Then, the preset session sample set is used to train the StructBERT small model to obtain an intent recognition model suitable for e-commerce scenarios. In actual use, the session context of the target consulting request is input into the intent recognition model, and the intent recognition model analyzes the session context to identify the user's current intent. The intent recognition model here can be deployed in the proxy module. When it is needed, the intent recognition model is directly called to reduce the amount of computing on the end side.

[0134] Step 304: If the current problem to be processed is not a preset type of problem, the session of the target consultation request is connected to the manual customer service link.

[0135] In this step, if the current pending issue of the consulting user is not a preset type of issue, it means that the current pending issue may not be able to be processed by relying on the preset large model. In order to ensure the user's consulting experience, it is necessary to connect to the manual customer service link for processing. In this case, the session of the target consulting request is automatically connected to the manual customer service link.

[0136] In one embodiment, after the session of the target consultation request is connected to the manual customer service link, the method may further include: generating reply information of the current pending problem according to the input information of the manual customer service. The reply information here includes but is not limited to text information, image information, audio and video information, operation processing and the like.

[0137] like Figure 6 As shown, it is a schematic diagram of a scenario link of AI processing consultation requests provided by an embodiment of the present application. In the manual service stage, by detecting and identifying the conversation context, the specific questions consulted by the user are screened and diverted. For known preset types of questions, such as repeated, trivial, and worthless conversations and questions, AI is used for processing, and the capabilities of the AI ​​big model are fully utilized to reduce the participation of manual customer service and improve the efficiency of consulting request processing. For problems that cannot be handled by the big model, the system actively transfers to manual processing. As a human, the manual customer service focuses on the service process and active service of high-value users and high-risk scenarios. It not only guarantees the user's service experience, but also improves resource utilization.

[0138] In one embodiment, generating reply information of the current pending problem according to the input information of the manual customer service may specifically include: responding to the auxiliary request of the manual customer service for the current pending problem, generating auxiliary information corresponding to the auxiliary request through a preset large model. Obtaining the input information determined by the manual customer service according to the auxiliary information. Responding to the reply instruction of the manual customer service to the current pending problem, sending the input information as the reply information of the current pending problem.

[0139] In this embodiment, the manual customer service can handle the user's consultation questions through online conversation. In the manual customer service module, an auxiliary module can be configured to assist the manual customer service to answer the user's consultation questions and improve the processing efficiency of the consultation request. For example, an intelligent agent application Copilot with common capabilities such as scene recognition, recommended first question, speech reply, and solution generation can be placed in the plug-in area of ​​the manual customer service module. The manual customer service can issue an auxiliary request for a specific consultation question in the auxiliary module, or open the automatic auxiliary function, and the system automatically triggers the auxiliary request. The auxiliary request can include the conversation material and instructions with the user. Copilot can identify its intention based on the conversation material or the instructions of the customer service waiter, and then route it to each large model agent module Agent (such as: credential recognition, speech generation, knowledge base RAG, etc.) for processing according to the intention, and return the corresponding query result as auxiliary information, such as returning the recommended reply information, and then the manual customer service can determine the information that needs to be entered according to the auxiliary information returned by the large model, or directly use the auxiliary information as the entry information, and then the manual customer service triggers the reply instruction, and the system sends the entry information as the reply information of the current pending problem to complete the conversation with the user.

[0140] like Figure 7 As shown, it is a full-link schematic diagram of a consultation request information processing method provided by the present application. Taking the consultation processing scenario of e-commerce users as an example, in order to achieve the above-mentioned stratification, HumanCopilot has designed two modes for online services: manual mode and AI hosting mode. The entire system can include: robot agent module (Agent), HumanCopilot workbench, HumanCopilot agent module, ISO solution center, among which:

[0141] Robot Agent: Equipped with robot customer service, it can start a conversation with the consulting user, ask the first question, have multiple rounds of conversations with the consulting user, call the question-and-answer knowledge base, and provide solutions (such as refund only or complain about the seller). During the conversation between the robot agent and the consulting user, the user can trigger a manual processing instruction through the robot customer service, or the system can automatically trigger a manual processing instruction by detecting and identifying the target user or the current consulting scenario as the target scenario category through the detection and recognition of the conversation. The robot agent requests the human customer service to handle it.

[0142] HumanCopilot workbench: When receiving a manual processing request from the robot agent, obtain the conversation context and / or other messages between the robot customer service and the user, and identify whether the current pending issue in the conversation is a type of issue that AI can handle based on the conversation context and / or other messages. If not, switch to manual mode, connect the user's current session to the manual customer service link, and generate the reply information of the current pending issue based on the input information of the manual customer service. If the current pending issue is a type of issue that AI can handle, you can enter the AI ​​full-hosting mode. In the AI ​​full-hosting mode, the AI ​​big model can handle the first question / conversation with the user and perform intent recognition on the user based on the conversation data. Here, the intent recognition can directly call the pre-trained StructBERT small model for intent recognition. Then, according to the intent recognition results, the routing is classified to different big model agent modules. For example, through intent routing, the user's consultation questions are divided into: pre-sales consultation, after-sales appeal, credential recognition agent and other types.

[0143] Among them, if the user's current consultation question is pre-sales consultation, such as asking about some knowledge-related content about the product, the pre-sales consultation-RAG mode of the HumanCopilot agent module can be called through the knowledge base RAG. In this mode, query (question) word rewriting, knowledge retrieval engine, and search results can be implemented according to needs. The query word rewriting and search results are merged in the prompt, and the new prompt is input into the LLM model so that the LLM model generates the answer corresponding to the specific question.

[0144] If the question the user is currently consulting belongs to an after-sales demand, the after-sales service-agent mode of the HumanCopilot agent module can be called through the solution Agent. In this mode, it is possible to recall candidate sets that match the demands through the ISO solution center based on multiple demands. Specifically, multiple demands can be matched in the ISO solution center, and the ISO solution center screens out one or more matching solutions through business policy group processing and security policy group processing, such as matching Solution 1 and Solution 2. Each solution includes access, process, execution and other stages. After obtaining the candidate set of solutions that match the demands, the after-sales service-agent mode infers and plans the candidate set of solutions, extracts influencing factors, and finally sorts the candidate set of solutions according to the factor extraction results, so that the AI ​​model can select a more appropriate solution based on the solution sorting to respond to the user's consultation questions.

[0145] If the user's current consultation question is a credential recognition request, the credential recognition agent can directly call the preset qwen-VL-agent multimodal mode to recognize the credential image uploaded by the user and provide analysis results.

[0146] On the other hand, the HumanCopilot workbench can also identify user emotions and the complexity of consulting questions, so that when it is detected that the user is in a bad mood or the consulting question is highly complex, human customer service can intervene in time to ensure user experience.

[0147] In summary, through the HumanCopilot workbench, it is possible to divide the user's consulting questions into pre-sales, sales, and after-sales, and perform intent recognition. According to the intent recognition results, the user's actual demands can be confirmed. The actual demands can be divided into: knowledge ID, document / FAQ, solution. The embodiment of this application focuses on the entire link, all channels, and all scenarios, and promptly and proactively intervenes in high-value, high-risk consulting services, building a multi-agent interconnection that follows up on consumer issues throughout the process.

[0148] like Figure 8 As shown, it is a schematic diagram of the interface of a human customer service provided by this application. The online customer service interface may include a conversation box and an auxiliary tool box. The conversation box displays the conversation content between the human customer service and the user. The human customer service can enter the conversation content through the input box and click the "Send" button to send it to the user. The auxiliary tool can provide relevant auxiliary information for the human customer service. For example, the AI ​​assistant automatically recommends relevant information to the user. The recommended information may include the key information of the user's consultation, the recommended first question, etc. For example, the key information may include scenarios, user demands, historical calls, merchant communication, etc. Examples are as follows:

[0149] The scenario is "Shipping", which means that the user's current question type is about product shipping, and the request is "request the seller to ship as soon as possible."

[0150] Historical calls: The member said that he had purchased the product 3 days ago, but had not received any response from the seller. He hoped that the product would be shipped as soon as possible.

[0151] Communication with the merchant: The member hopes that the goods will be shipped as soon as possible, but the merchant said that there is currently no stock and asked to wait a few more days.

[0152] Recommended first question: Hello, our waiter sees that you hope to contact the seller as soon as possible to ship the goods. Please wait a moment, we are verifying the relevant information.

[0153] Manual customer service can issue instructions in the auxiliary tool according to their own needs. For example, if the instruction is "reply words", the auxiliary tool can use the AI ​​big model capability to recommend reply words that match the current user's inquiry to the user, such as "Hello, the waiter sees that you hope to contact the seller as soon as possible for shipment. Please wait a moment. We are verifying the relevant information" as auxiliary information. Manual customer service can determine the entry information in the entry box based on the auxiliary information. For example, you can enter "Hello, the waiter sees that you hope to contact the seller as soon as possible for shipment. Please wait a moment. We are verifying the relevant information" in the entry box and click Send to reply to the user's inquiry.

[0154] In one embodiment, auxiliary modules such as customer service assistant, logistics view, search box, service track, etc. can also be configured in the plug-in area of ​​the online chat body of the manual customer service to assist the manual customer service in handling consulting issues.

[0155] In one embodiment, after entering the AI ​​hosting mode service in step 303, it may also include: detecting the current session context of the target consultation request, and using a preset large model to determine whether the current pending problem is solved according to the current session context. If the current pending problem is not solved, the session of the target consultation request is connected to the manual customer service link, and the reply information of the current pending problem is generated according to the input information of the manual customer service.

[0156] In this embodiment, in order to avoid abnormal situations in the AI ​​hosting mode, the current session context of the target consultation request can be detected in the AI ​​hosting mode, and the preset large model determines whether the current pending problem is solved according to the session context. If the AI ​​has replied to the user's current pending problem, but the current pending problem is still not solved, it means that the solution given by the AI ​​in the actual situation fails to meet the user's needs. In order to ensure the user's consulting experience, the session of the target consultation request can be automatically connected to the manual customer service link, and the reply information of the current pending problem is generated according to the input information of the manual customer service. Make full use of the advantages of manual customer service to improve user experience.

[0157] In actual scenarios, in the AI ​​hosting mode, AI can replace human customer service to generate some simple words and solutions and automatically reply without human intervention. Human customer service can still act as an "observer" to supervise the service process. When risks are found to be abnormal, they can intervene in the conversation process in time or directly transfer the conversation to the human customer service link.

[0158] like Fig. 9 As shown, it is a schematic diagram of the interface of an AI workbench provided in an embodiment of the present application. In the AI ​​hosting mode, the AI ​​workbench interface can display the overview of the user consultations currently being processed, for example, there are 7 user consultations in the current service, 2 high-risk user consultations, 4 medium-risk user consultations, and 1 low-risk user consultation. If a user consultation AI fails to solve the user consultation problem in time, it can be automatically transferred to the manual customer service. Assuming that user 3 repeatedly calls in, it means that user 3's consultation problem may not be solved in time, then user 3's consultation session can be transferred to the manual customer service link, and the manual conversation box can be expanded on the AI ​​workbench to display the conversation content between the manual customer service and user 3. The manual customer service can communicate with user 3 directly in the conversation box to facilitate the timely resolution of user 3's consultation problem.

[0159] In one embodiment, after entering the manual customer service mode in step 304, the method further includes: responding to the hosting instruction of the manual customer service, using a preset large model to process the target consultation request.

[0160] In this embodiment, in the manual customer service mode, if the manual customer service determines that the user's consulting problem has been resolved, it can actively trigger the hosting instruction to transfer the standardized closing service to the AI ​​large model for processing, saving labor costs.

[0161] In the embodiment of the present application, the manual customer service mode is used to handle the problems of high-risk and high-value users, and the AI ​​hosting mode is used to handle repetitive and trivial conversational problems during the user consultation process. When the nature of the conversation with the user changes, you can switch between the two modes. For example, in the manual customer service mode, if the user's problem has been solved, it can be hosted in the AI ​​mode for some "cleanup" services. In the AI ​​hosting mode, if some high-risk scenarios are encountered, they can also be automatically transferred to the manual customer service mode for better service. Improve the flexibility and efficiency of consulting request processing.

[0162] In actual scenarios, based on the analysis of a large amount of historical online session data, the session distribution ratio is as follows: Fig.10As shown in the figure, about 33% of the conversations have less than 12 messages. Considering the capabilities of the AI ​​big model, at least these 33% of the conversations can be carried by the AI ​​mode, and the problems of these 33% can be solved based on the capabilities of the AI ​​model. As the capabilities of the AI ​​model become stronger and stronger, AI can handle more and more scenarios and problems, and the system can follow the capabilities of the AI ​​model to make gradual evolution.

[0163] The above-mentioned consultation request processing method is based on the redesigned artificial service product form and service process based on AI, and the complete service capability built based on LLM and LLM Agents (large model agents). From the perspective of the entire service link, a link is designed to enable monitoring of the robot service process, effectively discover risks and intervene manually in time. In the manual customer service mode, the Copilot capability is redesigned based on AI Agent, and Agent can provide more full-process and real-time auxiliary capabilities. The manual workbench is divided into manual customer service mode and AI hosting mode, and tasks can be switched between the two modes according to the specific situation of the service, which can effectively improve the processing capacity of the workbench manual service and reduce service costs. At the same time, it can also more efficiently use precious manual resources to provide higher-quality services to high-value and high-risk users. In addition, the service process can be monitored and evaluated, and risks can be warned in time.

[0164] In actual scenarios, AI can be used to handle at least "33%" of trivial consultations, which can greatly improve service capabilities and efficiency. AI can provide much stronger parallel processing capabilities than humans, surpass the limits of human processing capabilities, and provide 1 to 100 or even higher parallel task processing capabilities. AI can provide standardized service processes to ensure service quality. AI-based services will not produce unnecessary losses, that is, there is no concept of "ineffective service". And as AI processing capabilities increase, more and more scenarios are covered, and the number of manual customer service required by the platform will decrease, especially low-level manual customer service for repetitive work. In addition, the cost of the large model itself will become a marginal cost as the scenarios are covered.

[0165] Please see Fig.11 , which is a method for processing consultation requests for product information in an embodiment of the present application, the method can be Figure 1 The electronic device 1 shown is used to perform and can be applied to Figure 2In the application scenario of consultation request processing shown in , the specific questions of user consultation can be screened and diverted in the manual service stage of the consultation request. For known preset types of consultation, the ability of the AI ​​big model is fully utilized to handle it, reducing the amount of manual customer service involved, improving the efficiency of consultation request processing, and automatically transferring the problem that the big model cannot handle to manual processing, which not only ensures the user's service experience, but also improves resource utilization. This embodiment takes the terminal 220 as an execution end as an example. Compared with the previous embodiment, this embodiment takes the consultation request processing scenario of the user about product information in the e-commerce scenario as an example. The method includes the following steps:

[0166] Step 1101: In response to a manual processing instruction regarding a product information inquiry request, a session context of the product information inquiry request is obtained. The product information here includes but is not limited to product-related knowledge information, product-related order information, product-related service information, and the like.

[0167] Step 1102: According to the session context, identify whether the current question to be processed in the session of the commodity information inquiry request is a preset type of question.

[0168] Step 1103: If the current pending question is a preset type question, use the preset large model to generate reply information corresponding to the current pending question.

[0169] Step 1104: If the current pending issue is not a preset type of issue, the conversation of the product information consultation request is connected to the manual customer service link, and reply information of the current pending issue is generated according to the input information of the manual customer service.

[0170] For details of each step of the above method, please refer to the relevant description of the above embodiment, which will not be repeated here.

[0171] Please see Fig.12 , which is a consultation request processing device 1200 of an embodiment of the present application, which can be applied to Figure 1 The electronic device 1 shown can be applied to Figure 2 In the application scenario of consultation request processing shown in , the specific questions of user consultation can be screened and diverted in the manual service stage of the consultation request. For known preset types of consultation, the ability of AI big model is fully utilized to handle it, reducing the amount of manual customer service involved, improving the efficiency of consultation request processing, and automatically transferring to manual processing for problems that cannot be handled by the big model, which not only ensures the user's service experience, but also improves resource utilization. The device includes: an acquisition module 1201, an identification module 1202, a first processing module 1203 and a second processing module 1204, and the functional principles of each module are as follows:

[0172] The acquisition module 1201 is used to acquire the session context of the target consultation request in response to the manual processing instruction for the target consultation request.

[0173] The identification module 1202 is used to identify whether the current question to be processed in the session of the target consultation request is a preset type of question according to the session context.

[0174] The first processing module 1203 is used to generate reply information corresponding to the current problem to be processed using a preset large model if the current problem to be processed is a preset type of problem.

[0175] The second processing module 1204 is used to connect the conversation of the target consultation request to the manual customer service link if the current problem to be processed is not a preset type of problem.

[0176] In one embodiment, it further includes: a trigger module, which is used to obtain the user characteristics of the user in response to the user's consultation request, determine whether the user is a target user based on the user characteristics, and trigger a manual processing instruction for the consultation request if the user is a target user.

[0177] In one embodiment, it further includes: a trigger module, which is used to respond to the user's consultation request and obtain the session context corresponding to the consultation request. According to the session context, the consultation scenario category to which the consultation request currently belongs is identified. If the consultation scenario category is a target scenario category that requires manual customer service intervention, a manual processing instruction for the consultation request is triggered.

[0178] In one embodiment, the trigger module is further configured to respond to the user's consultation request and connect the consultation request to the robot customer service link to obtain the conversation context between the robot customer service and the user.

[0179] In one embodiment, the identification module is used to input the conversation context into a preset large model, and the preset large model is used to analyze the conversation context and output the consultation scenario category to which the consultation request currently belongs.

[0180] In one embodiment, the second processing module 1204 is used to generate reply information of the current pending problem according to the input information of the manual customer service after connecting the conversation of the target consultation request to the manual customer service link.

[0181] In one embodiment, the second processing module 1204 is used to respond to the auxiliary request of the manual customer service for the current problem to be processed, generate auxiliary information corresponding to the auxiliary request through the preset large model, obtain the input information determined by the manual customer service according to the auxiliary information, and respond to the reply instruction of the manual customer service to the current problem to be processed, and send the input information as the reply information of the current problem to be processed.

[0182] In one embodiment, the first processing module 1203 is used to identify the current intention of the user according to the conversation context, and route the current intention to the corresponding large model proxy module, so that the large model proxy module generates matching reply information according to the current problem to be processed.

[0183] In one embodiment, the first processing module 1203 is used to input the conversation context into a pre-trained intent recognition model, and the intent recognition model is used to analyze the conversation context and identify the user's current intent, wherein the intent recognition model is trained based on a preset conversation sample set, and the corresponding user intent is marked in the preset conversation sample.

[0184] In one embodiment, it also includes: a switching module, which is used to detect the current session context of the target consultation request after using the preset large model to generate the reply information corresponding to the current pending problem, and use the preset large model to determine whether the current pending problem is solved according to the current session context. If the current pending problem is not solved, the session of the target consultation request is connected to the manual customer service link, and the reply information of the current pending problem is generated according to the input information of the manual customer service.

[0185] In one embodiment, the transfer module is also used to connect the conversation of the target consultation request to the manual customer service link, and after generating reply information for the current pending problem based on the input information of the manual customer service, respond to the hosting instructions of the manual customer service and use a preset large model to process the target consultation request.

[0186] For a detailed description of the above-mentioned consultation request processing device 1200, please refer to the description of the relevant method steps in the above-mentioned embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be repeated here.

[0187] Fig.13 The following is a schematic diagram of a cloud device 130 provided in an exemplary embodiment of the present application. The cloud device 130 can be used to run the method provided in any of the above embodiments. Fig.13 As shown, the cloud device 130 may include: a memory 1304 and at least one processor 1305, Fig.13 A processor is taken as an example.

[0188] The memory 1304 is used to store computer programs and can be configured to store various other data to support operations on the cloud device 130. The memory 1304 can be an object storage service (OSS).

[0189] Memory 1304 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0190] The processor 1305 is coupled to the memory 1304 and is used to execute the computer program in the memory 1304 to implement the solution provided by any of the above method embodiments. The specific functions and technical effects that can be achieved are not repeated here.

[0191] Furthermore, if Fig.13 The cloud device also includes: a firewall 1301, a load balancer 1302, a communication component 1306, a power supply component 1303 and other components. Fig.13 Only some components are shown schematically, which does not mean that the cloud device only includes Fig.13 Components shown.

[0192] In one embodiment, the above Fig.13 The communication component 1306 in is configured to facilitate wired or wireless communication between the device where the communication component 1306 is located and other devices. The device where the communication component 1306 is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G, LTE (Long Term Evolution, Long Term Evolution, referred to as LTE), 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component 1306 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 1306 also includes a near field communication (Near Field Communication, referred to as NFC) module to facilitate short-range communication. For example, the NFC module can be based on Radio Frequency Identification (Radio Frequency Identification, referred to as RFID) technology, Infrared Data Association (Infrared Data Association, referred to as IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (bluetooth, referred to as BT) technology and other technologies to achieve.

[0193] In one embodiment, the above Fig.13 The power supply component 1303 provides power to various components of the device where the power supply component 1303 is located. The power supply component 1303 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply component is located.

[0194] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the method of any of the aforementioned embodiments is implemented.

[0195] An embodiment of the present application also provides a computer program product, including a computer program, which implements the method of any of the aforementioned embodiments when executed by a processor.

[0196] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0197] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the methods of various embodiments of the present application.

[0198] It should be understood that the above-mentioned processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the application can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor. The memory may include a high-speed RAM (Random Access Memory) memory, and may also include non-volatile storage NVM (Nonvolatile memory, NVM for short), such as at least one disk storage, and can also be a USB flash drive, a mobile hard disk, a read-only memory, a disk or an optical disk, etc.

[0199] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special computer.

[0200] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic device or a main control device.

[0201] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, clothing or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, clothing or device. In the absence of more restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, clothing or device including the element.

[0202] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0203] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods of each embodiment of the present application.

[0204] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of user data and other information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0205] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for processing a consultation request, characterized in that: include: In response to a manual processing instruction for a target consultation request, obtaining a session context of the target consultation request; According to the conversation context, identifying whether the current question to be processed in the conversation of the target consultation request is a question of a preset type; If the current pending problem is a problem of the preset type, a preset large model is used to generate a reply message corresponding to the current pending problem; If the current problem to be processed is not a problem of the preset type, the session of the target consultation request is connected to the manual customer service link.

2. The method according to claim 1, characterized in that The steps for triggering manual processing instructions for target consultation requests include: In response to a consultation request from a user, obtaining user characteristics of the user; Determining whether the user is a target user according to the user characteristics; If the user is a target user, a manual processing instruction for the consultation request is triggered.

3. The method according to claim 1, characterized in that: The steps for triggering manual processing instructions for target consultation requests include: In response to a consultation request from a user, obtaining a session context corresponding to the consultation request; According to the session context, identifying the consultation scenario category to which the consultation request currently belongs; If the consultation scenario category is a target scenario category that requires manual customer service intervention, a manual processing instruction for the consultation request is triggered.

4. The method according to claim 3, characterized in that The step of obtaining, in response to a consultation request from a user, a session context corresponding to the consultation request includes: In response to the user's consultation request, connecting the consultation request to a robot customer service link; Obtain the conversation context between the robot customer service and the user.

5. The method according to claim 3, characterized in that: The step of identifying the consultation scenario category to which the consultation request currently belongs according to the session context includes: The session context is input into a preset large model, and the preset large model is used to analyze the session context and output the consultation scenario category to which the consultation request currently belongs.

6. The method according to claim 1, characterized in that After the session of the target consultation request is connected to the manual customer service link, the method further includes: generating reply information of the current pending problem according to the input information of the manual customer service; Generating reply information of the current pending issue according to the input information of the manual customer service includes: In response to the auxiliary request of the manual customer service for the current problem to be processed, generating auxiliary information corresponding to the auxiliary request through a preset large model; Obtaining the input information determined by the manual customer service according to the auxiliary information; In response to the manual customer service's reply instruction to the current pending issue, the input information is sent out as reply information of the current pending issue.

7. The method according to claim 1, characterized in that The using of the preset large model to generate the reply information corresponding to the current problem to be processed includes: identifying a user's current intent based on the session context; The current intention is routed to the corresponding large model proxy module so that the large model proxy module generates matching reply information according to the current problem to be processed.

8. The method according to claim 7, characterized in that The identifying the user's current intention according to the session context includes: The session context is input into a pre-trained intent recognition model, and the intent recognition model is used to analyze the session context and identify the user's current intent, wherein the intent recognition model is trained based on a preset session sample set, and the preset session samples are marked with corresponding user intents.

9. The method according to claim 1, characterized in that: After the preset large model is used to generate the reply information corresponding to the current problem to be processed, the method further includes: Detecting the current session context of the target consultation request, and using the preset large model to determine whether the current pending problem is solved according to the current session context; If the current pending issue is not resolved, the session of the target consultation request is connected to a manual customer service link, and reply information of the current pending issue is generated according to the input information of the manual customer service.

10. The method according to claim 1, characterized in that After the session of the target consultation request is connected to the manual customer service link and the reply information of the current pending problem is generated according to the input information of the manual customer service, the method further includes: In response to the hosting instruction of the manual customer service, the target consultation request is processed using the preset large model.

11. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the electronic device to perform the method described in any one of claims 1 to 10.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the processor executes the computer-executable instructions, the method according to any one of claims 1 to 10 is implemented.

13. A computer program product, characterized in that The method comprises a computer program, which implements the method according to any one of claims 1 to 10 when being executed by a processor.

Images