CN119168649A

CN119168649A - Consultation request processing method, device and storage medium

Info

Publication number: CN119168649A
Application number: CN202411017553.8A
Authority: CN
Inventors: 徐祁; 曹海峰; 李历岷; 赵东禹; 谭彦伯
Original assignee: Zhejiang Tmall Technology Co Ltd
Current assignee: Zhejiang Tmall Technology Co Ltd
Priority date: 2024-07-26
Filing date: 2024-07-26
Publication date: 2024-12-20

Abstract

The application provides a consultation request processing method, equipment and a storage medium, wherein the method comprises the steps of responding to a manual processing instruction aiming at a target consultation request, obtaining a session context of the target consultation request, identifying whether a current problem to be processed in the session of the target consultation request is a preset type problem or not according to the session context, generating reply information corresponding to the current problem to be processed by using a preset large model if the current problem to be processed is the preset type problem, and accessing the session of the target consultation request into a manual customer service link if the current problem to be processed is not the preset type problem. According to the application, the manual service stage of the consultation request is realized, the capability of the AI large model is fully utilized for processing aiming at the consultation of the known preset type, and the manual processing is automatically switched to aim at the problem that the large model cannot be processed, so that the service experience of a user is ensured, and the resource utilization rate is improved.

Description

Consultation request processing method, equipment and storage medium

Technical Field

The present application relates to the field of information processing technologies, and in particular, to a method, an apparatus, and a storage medium for processing a consultation request.

Background

With the development of the internet, users increasingly prefer to purchase goods online through an e-commerce platform. Conventional e-commerce platforms typically provide a search function for users to search for merchandise information of interest through the e-commerce platform.

In the e-commerce scene, an important bridge for communication is established between e-commerce customer service and users, and although the solution capability of robot customer service is strong, the service of high-quality and high-value users still cannot be completely replaced by manual customer service, and huge manual customer service still plays a very important role in the task of guaranteeing member experience. Under the traditional manual customer service mode, the customer service assistant online workbench can serve member users consulted by the customer service robot in line. In the service process, because personal experiences, working years and complexity of processing task scenes of two customer service groups are different, generally, the two customer service groups can simultaneously serve consultations of 1-6 member users, and service capacity is limited. Because of the limitation of service capability and upper limit, when the consultation amount is larger, a large amount of customer service is usually required to support, so that not only is the labor cost huge, but also the efficiency is low, and the ever-increasing consultation requirement of users cannot be met.

Disclosure of Invention

The main purpose of the embodiment of the application is to provide a consultation request processing method, equipment and storage medium, which are used for screening and shunting specific problems of user consultation in the manual service stage of the consultation request, processing the specific problems by fully utilizing the capacity of an AI large model aiming at the consultation of a known preset type, reducing the participation of manual customer service, improving the consultation request processing efficiency, automatically switching manual processing aiming at the problem that the large model cannot process, ensuring the service experience of the user and improving the resource utilization rate.

In a first aspect, an embodiment of the present application provides a method for processing a consultation request, including obtaining a session context of a target consultation request in response to a manual processing instruction for the target consultation request, identifying whether a current problem to be processed in a session of the target consultation request is a preset type problem according to the session context, generating reply information corresponding to the current problem to be processed by using a preset large model if the current problem to be processed is the preset type problem, and accessing the session of the target consultation request to a manual customer service link if the current problem to be processed is not the preset type problem.

In one embodiment, the step of triggering the manual processing instruction for the target consultation request comprises the steps of responding to the consultation request of a user, acquiring the user characteristics of the user, judging whether the user is the target user according to the user characteristics, and triggering the manual processing instruction for the consultation request if the user is the target user.

In one embodiment, the step of triggering the manual processing instruction for the target consultation request comprises the steps of responding to the consultation request of a user, acquiring a session context corresponding to the consultation request, identifying the current consultation scene category of the consultation request according to the session context, and triggering the manual processing instruction for the consultation request if the consultation scene category is the target scene category requiring manual customer service intervention.

In an embodiment, the step of responding to the consultation request of the user to obtain the session context corresponding to the consultation request comprises the step of responding to the consultation request of the user, accessing the consultation request into a robot customer service link, and obtaining the session context of the robot customer service and the user.

In an embodiment, the identifying the category of the consultation scene to which the consultation request currently belongs according to the session context includes inputting the session context into a preset large model, wherein the preset large model is used for analyzing the session context and outputting the category of the consultation scene to which the consultation request currently belongs.

In one embodiment, after the session of the target consultation request is accessed to the artificial customer service link, the method further comprises generating reply information of the current to-be-processed problem according to input information of the artificial customer service.

In an embodiment, the generating the reply information of the current problem to be processed according to the input information of the manual service includes generating auxiliary information corresponding to the auxiliary request through a preset large model in response to an auxiliary request of the manual service for the current problem to be processed, acquiring the input information determined by the manual service according to the auxiliary information, and sending the input information as the reply information of the current problem to be processed in response to a reply instruction of the manual service for the current problem to be processed.

In an embodiment, the generating the reply message corresponding to the current problem to be processed by using the preset large model includes identifying a current intention of a user according to the session context, and routing the current intention to a corresponding large model proxy module so that the large model proxy module generates the matched reply message according to the current problem to be processed.

In an embodiment, the method for identifying the current intention of the user according to the session context comprises the steps of inputting the session context into a pre-trained intention identification model, wherein the intention identification model is used for analyzing the session context and identifying the current intention of the user, the intention identification model is obtained through training according to a preset session sample set, and the preset session sample is marked with the corresponding user intention.

In an embodiment, after the reply message corresponding to the current problem to be processed is generated by using a preset large model, the method further comprises detecting a current session context of the target consultation request, judging whether the current problem to be processed is solved or not by using the preset large model according to the current session context, accessing the session of the target consultation request into a manual customer service link if the current problem to be processed is not solved, and generating the reply message of the current problem to be processed according to the input information of the manual customer service.

In an embodiment, after the session of the target consultation request is accessed to a manual service link and the reply information of the current to-be-processed problem is generated according to the input information of the manual service, the method further comprises responding to the hosting instruction of the manual service and processing the target consultation request by using the preset large model.

In a second aspect, an embodiment of the application provides a method for processing a consultation request about commodity information, which comprises the steps of responding to a manual processing instruction about the commodity information consultation request, obtaining a conversation context of the commodity information consultation request, identifying whether a current problem to be processed in the conversation of the commodity information consultation request is a preset type problem according to the conversation context, generating reply information corresponding to the current problem to be processed by using a preset large model if the current problem to be processed is the preset type problem, accessing the conversation of the commodity information consultation request into a manual customer service link if the current problem to be processed is not the preset type problem, and generating the reply information of the current problem to be processed according to input information of the manual customer service.

In a third aspect, an embodiment of the present application provides a consultation request processing apparatus, including:

the acquisition module is used for responding to a manual processing instruction aiming at a target consultation request and acquiring a session context of the target consultation request;

The identification module is used for identifying whether the current problem to be processed in the session of the target consultation request is a preset type problem or not according to the session context;

The first processing module is used for generating reply information corresponding to the current problem to be processed by using a preset large model if the current problem to be processed is the preset type problem;

And the second processing module is used for accessing the session of the target consultation request to the artificial customer service link if the current problem to be processed is not the preset type problem.

In one embodiment, the system further comprises a triggering module, wherein the triggering module is used for responding to the consultation request of the user, acquiring the user characteristics of the user, judging whether the user is a target user according to the user characteristics, and triggering a manual processing instruction aiming at the consultation request if the user is the target user.

In one embodiment, the system further comprises a triggering module, wherein the triggering module is used for responding to the consultation request of the user, acquiring a session context corresponding to the consultation request, identifying the current consultation scene category of the consultation request according to the session context, and triggering a manual processing instruction aiming at the consultation request if the consultation scene category is a target scene category requiring manual customer service intervention.

In an embodiment, the triggering module is further configured to respond to a consultation request of the user, access the consultation request to a robot service link, and obtain the session context of the robot service and the user.

In an embodiment, the identification module is configured to input the session context into a preset large model, where the preset large model is used to analyze the session context, and output a category of the consultation scene to which the consultation request currently belongs.

In an embodiment, the second processing module is configured to generate, after the session for the target consultation request is accessed to the artificial customer service link, reply information of the current problem to be processed according to input information of the artificial customer service.

In an embodiment, the second processing module is configured to respond to an auxiliary request of the human customer service for the current problem to be processed by generating auxiliary information corresponding to the auxiliary request through a preset large model, acquire the input information determined by the human customer service according to the auxiliary information, respond to a reply instruction of the human customer service for the current problem to be processed, and send the input information as reply information of the current problem to be processed.

In an embodiment, the first processing module is configured to identify a current intention of a user according to the session context, and route the current intention to a corresponding large model proxy module, so that the large model proxy module generates matched reply information according to the current problem to be processed.

In an embodiment, the first processing module is configured to input the session context into a pre-trained intent recognition model, where the intent recognition model is configured to analyze the session context and recognize a current intent of the user, where the intent recognition model is obtained by training according to a preset session sample set, and the preset session sample is marked with a corresponding user intent.

In an embodiment, the system further comprises a transfer module, wherein the transfer module is used for detecting the current session context of the target consultation request after the reply information corresponding to the current problem to be processed is generated by using a preset large model, judging whether the current problem to be processed is solved or not according to the current session context by using the preset large model, accessing the session of the target consultation request into a manual customer service link if the current problem to be processed is not solved, and generating the reply information of the current problem to be processed according to the input information of the manual customer service.

In an embodiment, the transfer module is further configured to, after the session for the target consultation request is accessed to a manual service link and the reply information of the current problem to be processed is generated according to input information of the manual service, process the target consultation request by using the preset large model in response to a hosting instruction of the manual service.

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

At least one processor, and

A memory communicatively coupled to the at least one processor;

wherein the memory stores instructions executable by the at least one processor to cause the electronic device to perform the method of any of the above aspects.

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

At least one processor, and

A memory communicatively coupled to the at least one processor;

Wherein the memory stores instructions executable by the at least one processor to cause the cloud device to perform the method of any of the above aspects.

In a sixth aspect, an embodiment of the present application provides a computer readable storage medium, where computer executable instructions are stored, and when executed by a processor, implement the method according to any one of the above aspects.

In a seventh aspect, embodiments of the present application provide a computer program product comprising a computer program which, when executed by a processor, implements the method of any of the above aspects.

According to the consultation request processing method, the equipment and the storage medium, through the consultation request needing manual processing, after the manual service stage is entered, the conversation context of the target consultation request is identified and analyzed, whether the current problem to be processed in the conversation of the target consultation request is a known consultation problem of a preset type or not is identified, and if the current problem to be processed is the consultation problem of the preset type, the problem to be processed can be processed by the preset large model, so that the preset large model generates reply information corresponding to the current problem to be processed. If the problem to be processed is not the preset type consultation problem, the real manual customer service is used for receiving and managing the session, so that the specific problem of the user consultation is screened and split in the manual service stage of the consultation request, the capacity of the AI large model is fully utilized for processing aiming at the known preset type consultation, the participation amount of the manual customer service is reduced, the consultation request processing efficiency is improved, the manual processing is automatically switched aiming at the problem that the large model cannot be processed, the service experience of the user is ensured, and the resource utilization rate is improved.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application. It will be apparent to those of ordinary skill in the art that the drawings in the following description are of some embodiments of the application and that other drawings may be derived from them without inventive faculty.

Fig. 1 is a schematic structural diagram of an electronic device according to an embodiment of the present application;

FIG. 2 is a schematic diagram of an application scenario of a consultation request processing system according to an embodiment of the present application;

FIG. 3 is a flow chart of a method for processing a consultation request according to an embodiment of the present application;

FIG. 4 is a schematic diagram of an exemplary consultation request processing service according to an embodiment of the present application;

FIG. 5A is a schematic diagram of a link diagram of an advisory request service reshaped by AI according to an embodiment of the present application;

FIG. 5B is an interface schematic diagram of an AI workstation provided in an embodiment of the application;

FIG. 6 is a schematic diagram of a scenario link for AI processing consultation request according to an embodiment of the present application;

FIG. 7 is a full-link diagram of a method for processing information of a consultation request according to the present application;

FIG. 8 is a schematic view of an interface for a human customer service provided by the present application;

FIG. 9 is an interface schematic diagram of an AI workstation according to an embodiment of the application;

FIG. 10 is a schematic diagram of a session distribution ratio of a consultation request according to an embodiment of the present application;

FIG. 11 is a flowchart of a method for processing a consultation request regarding merchandise information according to an embodiment of the present application;

FIG. 12 is a schematic diagram of a device for processing a consultation request according to an embodiment of the present application;

fig. 13 is a schematic structural diagram of a cloud device according to an embodiment of the present application.

Specific embodiments of the present application have been shown by way of the above drawings and will be described in more detail below. The drawings and the written description are not intended to limit the scope of the inventive concepts in any way, but rather to illustrate the inventive concepts to those skilled in the art by reference to the specific embodiments.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples do not represent all implementations consistent with the application.

The term "and/or" is used herein to describe an association relationship of an associated object, and specifically indicates that three relationships may exist, for example, a and/or B may indicate that a exists alone, while a and B exist together, and B exists alone.

It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or fully authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region, and provide corresponding operation entries for the user to select authorization or rejection.

In order to clearly describe the technical solution of the embodiments of the present application, firstly, the terms involved in the present application are explained:

AI ARTIFICIAL INTELLIGENCE, artificial intelligence.

LLM Large Language Model, large language model, is a deep learning based natural language processing model that can learn the grammar and semantics of natural language so that human readable text can be generated. The large language model can process various natural language tasks, such as text classification, question-answering, dialogue and the like, and is an important path to artificial intelligence.

API Application Programming Interface, application programming interface.

Prompt, a Prompt word, is a natural language input, similar to a command or instruction, that allows the AI model to know what it needs to do.

IM INSTANT MESSAGING, instant messaging.

The online customer service generally refers to that customer service interacts with members in a text online IM mode, so that the requirements of the members are solved.

Copilot an intelligent auxiliary product in a customer service workbench can provide intelligent conversation, scheme recommendation, intelligent abstract and other auxiliary capabilities for a customer service staff.

Agent, large model Agent, possessing advanced ability of sensing, reasoning, memorizing, etc.

Small model-A small model is a specific model designed and trained by an algorithm to handle a particular task, as opposed to a general large model.

HumanCopilot an intelligent artificial service design with AI as a service subject and human as an assist.

ID Identity document, ID card identification number, account number and unique code.

ChatGLM: chat Global Language Model, chat global model.

SOP Standard Operating Procedure, standard working procedure.

GPT GENERATIVE PRE-Trained Transformer, an Internet-based, data-trained, text-generated deep learning model.

RAG RETRIEVAL-Augmented Generation, search enhancement generation.

ISO: service flow digitization.

BERT Bidirectional Encoder Representations from Transformers, a pre-trained language model based on a multi-layer transducer encoder.

StructBERT A structured BERT model, a pre-trained model for natural language processing, is an improved version based on the BERT model.

FAQ Frequently Asked Questions, common problem solving.

The consultation request processing mode of the embodiment of the application can be applied to any field needing to process information requests.

Taking a customer service of an e-commerce platform as an example to process a consultation request scene of a user, in the e-commerce scene, the e-commerce customer service and the user establish an important bridge for communication, although the solution capability of the AI robot customer service is strong, the manual customer service still cannot be completely replaced for the service of the high-quality high-value user, and the huge manual customer service still plays a very important role in the task of guaranteeing the member experience. Under the traditional manual customer service mode, the customer service assistant online workbench can serve member users consulted by the customer service robot in line. In the service process, because personal experiences, working years and complexity of processing task scenes of two customer service groups are different, generally, the two customer service groups can simultaneously serve consultations of 1-6 member users, and service capacity is limited. Because of the limitation of service capability and upper limit, when the consultation amount is larger, a large amount of customer service is usually required to support, so that not only is the labor cost huge, but also the efficiency is low, and the ever-increasing consultation requirement of users cannot be met.

In alternative embodiments, the customer service system may provide a SOP/ISO-like tool through product capability to assist the customer service party in providing a standardized solution, but the final quality of service may still be affected by factors such as individual differences in customer service party (skills, calendaring, competence, emotion, personality) and product design of the customer service system itself. In a practical scenario, many user incoming lines do not generate any valuable consultation (which can be usually solved through document questions or intelligent speech), but occupy an artificial customer service channel, so that unnecessary loss is caused.

With the development of artificial intelligence technology, the understanding capability, logic reasoning capability and multi-mode capability of a large model are greatly improved, and the AI can accurately understand the consultation problem of the member user in most simple service scenes and can automatically call the system capability (ISO, SOP and action) to solve the consultation problem of the member user.

However, in the intelligent age, the manual customer service still has the following advantages:

1. The manual service still has irreplaceability, as "people" can:

a. and better experience is brought to high-value users.

B. and (5) processing risks.

C. Active service.

D. providing emotional value.

In summary, how to build an extensible artificial service model and product design based on AI technology to bear the capability of a continuously strong large model becomes a research-worthy problem.

In order to solve at least one of the above problems, an embodiment of the present application provides a solution for processing a query request, by identifying and analyzing a session context of a target query request after entering a manual service stage, to identify whether a current problem to be processed in a session of the target query request is a known preset type of query problem, and if the current problem is a preset type of query problem, the current problem to be processed may be submitted to a preset large model to be processed, so that the preset large model generates reply information corresponding to the current problem to be processed. If the problem to be processed is not the preset type consultation problem, the real manual service is used for receiving and managing the session, and the reply information of the current problem to be processed is generated according to the input information of the manual service, so that the specific problem consulted by the user is screened and shunted in the manual service stage of the consultation request, the capability of the AI large model is fully utilized for processing the known preset type consultation, the participation amount of the manual service is reduced, the processing efficiency of the consultation request is improved, and the manual processing is automatically switched according to the problem that the large model cannot process, so that the service experience of the user is ensured, and the resource utilization rate is improved.

Specifically, through reconstructing a service flow and a manual service workbench based on an AI, the problem of simple and repeated consultation is solved completely by the AI, and manual customer service is focused on providing higher-quality active service for high-value users and high-risk scenes.

Some embodiments of the present application are described in detail below with reference to the accompanying drawings. In the case where there is no conflict between the embodiments, the following embodiments and features in the embodiments may be combined with each other. In addition, the sequence of steps in the method embodiments described below is only an example and is not strictly limited.

As shown in fig. 1, the present embodiment provides an electronic device 1, which includes at least one processor 11 and a memory 12, and one processor is exemplified in fig. 1. The processor 11 and the memory 12 are connected by a bus 10. The memory 12 stores instructions executable 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 methods in the following embodiments, so as to implement screening and splitting of specific problems of user consultation in a manual service stage of a consultation request, process the specific problems by fully utilizing the capability of an AI large model for the known preset type consultation, reduce the participation of manual service, improve the processing efficiency of the consultation request, and automatically switch manual processing for the problem that the large model cannot process, thereby not only ensuring the service experience of the user, but also improving the resource utilization rate.

In an embodiment, the electronic device 1 may be a mobile phone, a tablet computer, a notebook computer, a desktop computer, or a large computing system composed of a plurality of computers.

Fig. 2 is a schematic diagram of an application scenario 200 of a consultation request processing system according to an embodiment of the present application. As shown in fig. 2, the system includes a server 210 and a terminal 220, wherein:

The server 210 may be a data platform providing a consultation request processing service, for example, an e-commerce customer service platform. In a practical scenario, an e-commerce customer service platform may have multiple servers 210, for example, 1 server 210 in fig. 2.

The terminal 220 may be a computer, a mobile phone, a tablet, or other devices used when the user logs in to the e-commerce customer service platform, or a plurality of terminals 220 may be provided, and 2 terminals 220 are illustrated in fig. 2 as an example.

Information transmission between the terminal 220 and the server 210 may be performed through the internet, so that the terminal 220 may access data on the server 210. The terminal 220 and/or the server 210 may be implemented by the electronic device 1.

The consultation request processing scheme of the embodiment of the application can be deployed on the server 210, the terminal 220 or the server 210 and the terminal 220. The actual scene may be selected based on actual requirements, which is not limited in this embodiment.

When the consultation request processing scheme is fully or partially deployed on the server 210, a call interface may be opened to the terminal 220 to provide algorithm support to the terminal 220.

The method provided by the embodiment of the application can be realized by the electronic equipment 1 executing corresponding software codes and by carrying out data interaction with a server. The electronic device 1 may be a local terminal device. When the method is run on a server, the method can be implemented and executed based on a cloud interaction system, wherein the cloud interaction system comprises the server and the client device.

In a possible implementation manner, the method provided by the embodiment of the present application provides a graphical user interface through a terminal device, where the terminal device may be the aforementioned local terminal device or the aforementioned client device in the cloud interaction system.

Please refer to fig. 3, which is an embodiment of the present application, which is a consultation request processing method, which can be executed by the electronic device 1 shown in fig. 1 and can be applied to the consultation request processing application scenario shown in fig. 2, so as to implement screening and splitting of specific problems of user consultation in a manual service stage of the consultation request, fully utilize the capability of the AI large model for processing for the known preset type consultation, reduce the participation of manual customer service, improve the processing efficiency of the consultation request, and automatically switch manual processing for the problem that the large model cannot process, thereby not only ensuring the service experience of the user, but also improving the resource utilization rate. In this embodiment, taking the terminal 220 as an executing terminal as an example, the method includes the following steps:

Step 301, responding to a manual processing instruction aiming at a target consultation request, and acquiring a session context of the target consultation request.

In this step, the target consultation request may be a consultation request initiated by the user for a specific query, for example, a consultation request initiated by a consumer user or a merchant user for a specific commodity or specific order information in an e-commerce scenario. The consultation request initiated by the user in the actual scene can be processed by the robot, or can be processed manually, and the manual processing instruction can be actively triggered by the user initiating the consultation request or automatically triggered by the system. When a manual processing instruction for the target consultation request is detected, a session context of the target consultation request is firstly acquired in response to the instruction, wherein the session context can contain all session contents from the initiation to the current moment of the target consultation request so as to analyze the actual requirements of the user according to the session context.

In one embodiment, the step of triggering the manual processing instruction for the target consultation request includes acquiring user characteristics of the user in response to the consultation request of the user. And judging whether the user is a target user according to the user characteristics. And if the user is a target user, triggering a manual processing instruction aiming at the consultation request.

In this embodiment, taking a user consultation scenario of an e-commerce platform as an example, when a user has a question about certain commodity information or order information, a customer service consultation request is triggered, in response to the consultation request, user characteristics of the user are firstly obtained, where the user characteristics include, but are not limited to, identity characteristics, character characteristics, member characteristics and other information of the user, then whether the user is a specific target user is judged according to the user characteristics, and specific characteristics of the target user can be set according to actual requirements, for example, the target user can be a user with a member registered in the platform, or can be a user with a specific character characteristic (such as splenic dysphoria) or can be a marked user in a history consultation record, a user with consumption reaching a certain limit, and the like. In an actual scene, the consultation request of the target user has great influence on the e-commerce platform, belongs to a key service object in the consultation service, and can automatically trigger a manual processing instruction for the consultation request of the user if the current consultation user is judged to be the target user in order to ensure the consultation experience of the target user.

In one embodiment, the step of triggering the manual processing instruction for the target consultation request includes responding to the consultation request of the user and acquiring the session context corresponding to the consultation request. And identifying the category of the consultation scene to which the consultation request currently belongs according to the session context. And if the consultation scene category is a target scene category requiring manual customer service intervention, triggering a manual processing instruction aiming at the consultation request.

In this embodiment, taking a user consultation scenario of an e-commerce platform as an example, when a user triggers a customer service consultation request on certain commodity information or order information, a session context of the consultation request can be detected in real time, and a consultation scenario category to which the consultation request currently belongs is identified. Judging whether the current consultation scene category of the consultation request is a target scene category or not, wherein the target scene category refers to a high-risk scene category requiring manual customer service intervention, such as a scene with unconditional vocabulary such as abuse, threat and the like caused by a language sent by a user in a conversation, or a scene with the user desiring to injure the user or other people. The method can preset which scene categories need manual customer service intervention, judge whether the scene categories are target scene categories by comparing the consultation scene categories to which the current consultation scene categories belong with preset scene categories, and also can directly identify whether the consultation scene categories are target scene categories according to the conversation context by a preset large model. If the consultation scene category is a target scene category requiring manual customer service intervention, the system can automatically trigger a manual processing instruction aiming at the consultation request so as to facilitate the manual customer service intervention processing in time, thereby improving the consultation processing efficiency and the consultation experience of the user.

In one embodiment, responding to the consultation request of the user, acquiring the session context corresponding to the consultation request specifically comprises responding to the consultation request of the user, and accessing the consultation request into the robot customer service link. And acquiring the conversation context of the robot customer service and the user.

In this embodiment, taking a user consultation scenario of an e-commerce platform as an example, when a user triggers a customer service consultation request for certain commodity information or order information, a preset robot customer service can process the consultation request of the user, and the robot customer service can provide conversational intelligent service for the user, so that labor cost is saved, and efficiency is improved. In the process of processing the user consultation by the robot customer service, the conversation content between the robot customer service and the user is detected in real time to obtain the conversation context, wherein the conversation context can refer to chat corpus of the robot customer service and the user. By setting up session content detection in the robot service stage, whether manual customer service intervention is needed or not can be screened in real time, and the capability of active service of the system is improved.

In one embodiment, identifying the category of the counseling scene to which the counseling request currently belongs according to the session context comprises inputting the session context into a preset large model for analyzing the session context and outputting the category of the counseling scene to which the counseling request currently belongs.

In this embodiment, the process of handling the user consultation by the robot customer service may be monitored by using the AI large model, and the session context between the robot customer service and the user is input into the preset large model, so that the preset large model analyzes the session context, identifies the type of the consultation scene to which the consultation request currently belongs, fully utilizes the capability of the AI large model, and improves the processing efficiency.

As shown in fig. 4, a typical service process of processing a consultation request is provided in an embodiment of the present application, and in a typical service process, a user who initiates a consultation request first enters a service with a robot, and the service with the robot provides conversational intelligent service for the user. When the robot cannot solve the problem or the user has strong appeal to the manual service, the user can apply for the manual service from the robot customer service, then manually overcome the connection between the secondary customer service workbench and the user, and can provide more accurate service and solving capability for the member user by means of tools such as a customer service assistant, copilot (an AI assistant) and the like. If the problems of some scenes still cannot be solved by online chat, the work order can be recorded and then be subsequently solved by offline service.

However, in the robot service stage, although the intelligent service capability provided by the ai+ rule is provided, the damage of the user experience of the member in the lover self-service stage is still avoided due to the limitation of the AI and the uncertainty of the robot service, and particularly for high-value and high-risk users, the system still needs to have the capability of actively carrying out the process, and cannot wait for the layer-by-layer funnels to be finally distributed.

As shown in fig. 5A, in order to solve the problem of the embodiment in fig. 4, an observation tray may be provided in the robot service process, and user features and/or session context for detecting incoming line consultation may be implemented, so as to screen out target users with high value, or identify target scene categories requiring human customer service intervention at high risk in the consultation session process, so that the consultation problem of the users can be actively accepted by the system, and be timely handled by human customer service intervention, thereby reducing loss of service experience. At this stage, real-time chat analysis can be effectively performed on the session context by utilizing the characteristics of the AI large model, and scenes such as service experience, risk, value, emotion and the like are identified.

As shown in fig. 5B, an interface schematic diagram of an AI workbench provided in an application embodiment may display, in real time, scene category recognition results of multiple consultation users in an interface of the AI workbench in a robot service process, so as to facilitate real-time observation of actual situations of different users. Take the example of a 4 user observation tray in fig. 5B, wherein:

The consultation type of the user 1 is ' goods replacement/return ', the current progress is that the quality problem of goods is determined, the conversation content sent by the user at the time of 00:50 is ' hello ', the user first prompts the logistics and dispatch ', the category of the consultation scene to which the consultation request of the user 1 currently belongs is identified as ' high-risk user ', the category of the target scene belongs to, and the user is characterized as ' anger '. The system can automatically access the manual customer service channel for processing, and reduces the experience loss of users.

The consultation type of the user 2 is commodity quality problem, the current processing progress is whether a consumer wants to return goods and refund because of commodity quality problem, customer service can assist the consumer to apply for after-sale and explain the maintenance step, the user characteristic of the user 2 is identified as a member user, and the member user is a high-value user for an e-commerce platform, so the user belongs to a target user, and can automatically access a manual customer service channel for processing, and user experience is improved.

The type of consultation by user 3 is "repeat incoming line, upgrade process", and the current demand by the user is "consumer wants to know after-sales process flow". The user 3 does not belong to a target user, the current consultation scene category is not the target scene category, and the current consultation scene category belongs to the consultation scene which can be completed by the AI, so that the consultation content of the processor can be continued by the robot customer service or processed by the AI large model, and the participation of the manual customer service is saved.

The consultation type of the user 4 is 'goods replacement/return', the current progress is the commodity quality problem, the current consultation scene category of the consultation request of the user 4 is identified as 'high-risk user', the consultation scene category belongs to the target scene category, and the consultation scene category can be automatically accessed into a manual customer service channel for processing. And the experience loss of the user is reduced.

Step 302, according to the session context, identifying whether the current problem to be processed in the session of the target consultation request is a preset type problem. If yes, go to step 303, otherwise go to step 304.

In this step, after entering the manual service stage, the session context of the target consultation request is identified and analyzed, and whether the currently pending problem in the session of the target consultation request is a known consultation problem of a preset type is identified, where the preset type problem may include some problems that already have a fixed answer or a standard answer flow, such as a knowledgeable problem, a simple repeated problem, and the like. These types of questions may directly invoke the relevant knowledge base or preset question-answer databases to query the corresponding answers, thus eliminating the need for human customer service to handle them. If the current problem to be processed is a preset type problem, the step 303 is performed directly by the preset large model, otherwise the step 304 is performed by the manual service.

Step 303, if the current problem to be processed is a preset type problem, generating reply information corresponding to the current problem to be processed by using a preset large model.

In this step, the preset type of questions generally belong to repeated, trivial and worthless dialogues and questions, an AI hosting mode may be adopted, a preset large model is combined with a related knowledge base or a preset question-answer database to query corresponding answers, and answer information corresponding to the current questions to be processed is generated according to the query result, so that participation of manual customer service is reduced, and processing efficiency is improved. The reply information here includes, but is not limited to, text information, image information, audio-video information, operation processing, and the like.

In one embodiment, step 303 may include specifically identifying a current intent of the user based on the session context. And routing the current intention to a corresponding large model proxy module so that the large model proxy module generates matched reply information according to the current problem to be processed.

In this embodiment, in the AI hosting mode, the AI may replace a manual customer service to perform some simple speech operation and scheme generation, and automatically reply, without manual participation, may identify the current intention of the user according to the conversation context corpus, and then route to each corresponding large model Agent module Agent according to the current intention of the user, where the large model Agent modules are used to implement specific information query functions, such as a credential identification module, a speech operation generation module, a knowledge base RAG, etc., and these large model Agent modules may match corresponding query results according to the current intention of the user, and may generate, via a preset large model, response information of the current problem to be processed according to the query results to generate a match. The accuracy and the effectiveness of the reply information are improved.

The great advantage of the AI hosting mode is that the cost reduction and efficiency enhancement can be achieved by utilizing the strong processing capacity of the AI to improve the manual parallel service capacity, and meanwhile, the customer service capability is liberated from the problem that the AI can process. AI mode, on the other hand, is evolutionarily capable of progressively enhancing as models and intelligence capabilities evolve, assuming more and more of the original service processes that require human customer service participation.

In one embodiment, the method for identifying the current intention of the user according to the conversation context comprises the steps of inputting the conversation context into a pre-trained intention identification model, wherein the intention identification model is used for analyzing the conversation context and identifying the current intention of the user, the intention identification model is obtained through training according to a preset conversation sample set, and the preset conversation sample is marked with the corresponding user intention.

In this embodiment, the intent recognition model may be a small model specifically trained for the advisory scenario, such as training StructBERT small models for intent recognition. Taking an e-commerce scene as an example, during model training, consultation session samples of different users in the e-commerce scene can be collected, corresponding user intentions are marked in each session sample to form a preset session sample set, then a StructBERT small model is trained by adopting the preset session sample set to obtain an intention recognition model suitable for the e-commerce scene, in actual use, the session context of a target consultation request is input into the intention recognition model, the intention recognition model analyzes the session context, and the current intention of the user is recognized. The intention recognition model can be deployed in the proxy module, and when the intention recognition model needs to be used, the intention recognition model is directly called, so that the calculation amount of the end side is reduced.

And 304, if the current problem to be processed is not a preset type of problem, accessing the session of the target consultation request into the manual customer service link.

In this step, if the current problem to be processed of the consulting user is not a preset type problem, it is indicated that the current problem to be processed may not be processed depending on a preset large model, in order to ensure the consulting experience of the user, a manual customer service needs to be accessed for processing, a session of the target consulting request is automatically accessed to a manual customer service link,

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

As shown in fig. 6, in a schematic view of a scenario link of an AI processing consultation request provided by the embodiment of the present application, in a manual service stage, specific problems of user consultation are screened and shunted by detecting and identifying a session context, and for known preset types of problems, such as repeated, trivial, and worthless conversations and problems, AI processing is performed, capabilities of an AI large model are fully utilized, participation of manual service is reduced, processing efficiency of the consultation request is improved, and for the problem that the large model cannot be processed, the system actively switches manual processing, and the manual service is focused on a service process and active service of a high-value user and a high-risk scenario as a person. The service experience of the user is guaranteed, and the resource utilization rate is improved.

In an embodiment, generating the reply information of the current problem to be processed according to the input information of the manual service may specifically include generating auxiliary information corresponding to the auxiliary request through a preset large model in response to the auxiliary request of the manual service for the current problem to be processed. And acquiring input information determined by the manual customer service according to the auxiliary information. Responding to a reply instruction of the manual customer service to the current problem to be processed, and sending the input information as reply information of the current problem to be processed.

In this embodiment, the manual service may process the consultation problem of the user in an online session manner, and in the manual service module, an auxiliary module may be configured to assist the manual service in answering the consultation problem of the user, so as to improve the processing efficiency of the consultation request. For example, an agent application Copilot with the common capabilities of scene recognition, first question recommendation, speaking reply, scheme generation and the like can be placed in the plug-in area of the manual customer service module. The manual customer service can send an auxiliary request aiming at a specific consultation problem at the auxiliary module or open an automatic auxiliary function, and the auxiliary request is automatically triggered by the system. The auxiliary request can contain contents such as conversation corpus and instructions of the user, copilot can identify the intention of the user according to the instructions of the conversation corpus or the customer service, then the intention is routed to each large model Agent module Agent (such as certificate identification, voice operation generation, knowledge base RAG and the like) for processing according to the intention, a corresponding query result is returned as auxiliary information, such as recommended reply information, then the manual customer service can determine information required to be recorded according to the auxiliary information returned by the large model, or directly takes the auxiliary information as recording information, then the manual customer service triggers the reply instruction, and the system sends the recording information as reply information of the current problem to be processed to complete the conversation with the user.

As shown in fig. 7, for the full-link schematic diagram of the method for processing the consultation request information provided by the present application, taking the consultation processing scenario of the e-commerce user as an example, in order to implement the above layering, humanCopilot designs two modes for online service, namely a manual mode and an AI hosting mode, the whole system may include a robot Agent module (Agent), a HumanCopilot workbench, a HumanCopilot Agent module, and an ISO solution center, wherein:

The robot Agent is configured with a robot customer service, can realize the functions of starting a session with a consultation user, issuing a first question, carrying out multiple rounds of conversations with the consultation user, calling a question-answer knowledge base, giving a solution (such as refund only or complaint sellers) and the like. In the conversation process of the robot Agent and the consultation user, the user can trigger manual processing instructions through the robot customer service, or the system detects that the target user or the current consultation scene is the target scene type through detection and identification of the conversation, automatically triggers manual processing instructions, and the robot Agent requests manual customer service processing to a HumanCopilot workbench.

And HumanCopilot, when receiving a manual processing request of the robot Agent, acquiring a session context and/or other information between the robot customer service and the user, identifying whether the current problem to be processed in the session is of a problem type which can be processed by an AI according to the session context and/or other information, if not, switching to a manual mode, accessing the current session of the user into a manual customer service link, and generating reply information of the current problem to be processed according to the input information of the manual customer service. If the current problem to be processed is of a problem type which can be processed by the AI, the AI full-support mode can be entered, in the AI full-support mode, the AI large model can process a first question/dialogue with a user, and the intention recognition is carried out on the user according to a conversation corpus, and the intention recognition can directly call a pre-trained StructBERT small model to carry out the intention recognition. And then classifying and routing to different large model proxy modules according to the intention recognition result, for example, classifying the consultation problem of the user into types such as pre-sale consultation, after-sale appeal, certificate recognition Agent and the like through the intention routing.

If the current information inquiry questions of the user are pre-sale consultation, for example, the related contents of some knowledge about the commodity can be consulted, the pre-sale consultation-RAG mode of the HumanCopilot agency module can be called through the RAG, in the mode, query word rewriting, knowledge retrieval engine and retrieval result giving can be realized according to the requirements, the query word rewriting and the retrieval result are combined at the Prompt, and the new Prompt is input into the LLM model, so that the LLM model generates an answer corresponding to the specific problem.

If the problem currently consulted by the user belongs to an after-sales appeal, an after-sales service-Agent mode of the HumanCopilot Agent module can be called by the solution Agent, and in the mode, a candidate set matched with the appeal can be recalled by the ISO solution center according to multiple appeal. Specifically, the multiple requirements can be matched in an ISO solution center, and the ISO solution center screens out one or more matched schemes, such as scheme 1 and scheme 2, through service policy group processing and security policy group processing, wherein each scheme comprises the stages of admission, flow, execution and the like. After the scheme candidate set of resort matching is obtained, the after-sales service-agent mode is used for reasoning and planning the scheme candidate set, extracting influence factors, and finally, carrying out scheme sorting on the scheme candidate set according to the factor extraction result, so that an AI model can select a more suitable solution according to the scheme sorting to reply to the consultation problem of the user.

If the current consultation problem of the user belongs to the certificate identification request, a preset qwen-VL-Agent multi-mode can be directly called by the certificate identification Agent, so that the certificate picture uploaded by the user can be identified, and an analysis result is given.

On the other hand, humanCopilot the workstation can also carry out emotion recognition, consult problem complexity discernment to the user to when discernment user's emotion is bad or consult problem complexity is higher, in time intervene artifical customer service, guarantee user experience.

In sum, through HumanCopilot's workstation, can realize dividing user's consultation problem into before, in, after selling, carry out intention discernment, confirm user's actual appeal according to intention discernment result, actual appeal can be divided into knowledge ID, document/FAQ, solution. The embodiment of the application constructs multi-Agent interconnection which completely follows the problems of consumers by focusing all links, all channels and all scenes and actively intervening in high-value and high-risk consultation services in time.

As shown in fig. 8, an interface schematic diagram of a manual customer service provided by the application may include a session frame and an auxiliary tool frame in an online customer service interface, wherein the session frame displays session contents of the manual customer service and a user, and the manual customer service can input the session contents through an input frame and click a "send" button to send to the user. The auxiliary tool can provide first-aid auxiliary information for the manual customer service, for example, an AI assistant automatically recommends relevant information for the user, the recommended information can comprise key information of the user consultation, recommended first questions and the like, and the key information can comprise contents such as scenes, user appeal, historical incoming lines, merchant communication and the like, for example, the following contents are included:

the scenario is "shipping", which means that the type of problem currently consulted by the user is related to commodity shipping, and the complaint is "ask the seller to ship as soon as possible.

Historical incoming line, that is, the member expresses that the commodity has been taken for 3 days, and the contact seller does not answer, so that the commodity is hoped to be shipped as soon as possible.

And the merchant communicates that the member hopes to deliver goods as soon as possible, and the merchant expresses that no goods exist at present for a few days.

The recommendation asks you good, you see you like you to get in contact with the seller as soon as possible, please get your mind, etc., and verify the relevant information.

The manual customer service can send instructions in an auxiliary tool according to own requirements, for example, the instructions are "answer call operation", the auxiliary tool can recommend answer call operation matched with the current user consultation for the user by means of AI large model capacity, "you good, you can see that you can get the relevant information to be connected with the seller as soon as possible, you wait, and verify relevant information" as auxiliary information, and the manual customer service can determine the input information in the input frame according to the auxiliary information, for example, you good, you can see that you can get the seller to be connected with the seller as soon as possible, you wait, verify relevant information "is input in the input frame, and click and send to answer the consultation of the user.

In an embodiment, auxiliary modules such as a customer service assistant, a logistics view, a search frame, a service track and the like can be configured in a plug-in area of an online chat subject of the manual customer service to assist the manual customer service in handling the consultation problem.

In one embodiment, after the AI-hosting mode service is entered in step 303, the method may further include detecting a current session context of the target consultation request, and determining whether the current pending problem is solved according to the current session context using a preset large model. If the current problem to be processed is not solved, accessing the session of the target consultation request into the artificial customer service link, and generating reply information of the current problem to be processed according to the input information of the artificial customer service.

In this embodiment, in order to avoid an abnormal situation in the AI hosting mode, the current session context of the target consultation request may be detected in the AI hosting mode, and a preset large model is used to determine whether the current problem to be processed is solved according to the session context, if the current problem to be processed of the user has been replied to by the AI, but the current problem to be processed is still not solved, which indicates that the solution given by the AI in the actual situation fails to satisfy the user requirement, in order to ensure the consultation experience of the user, the session of the target consultation request may be automatically accessed to the artificial customer service link, and reply information of the current problem to be processed may be generated according to the input information of the artificial customer service. The advantage of manual customer service is fully utilized, and the user experience is improved.

In an actual scene, in an AI hosting mode, AI can replace manual customer service to perform some simple speaking operations and scheme generation and automatically reply without manual participation. The manual customer service can still be used as an observer to supervise the service process, and when the risk abnormality is found, the conversation process is timely intervened or the conversation is directly transferred to the manual customer service link.

As shown in fig. 9, an interface schematic diagram of an AI workbench according to an embodiment of the present application may show a user consultation profile currently being processed in an AI hosting mode, for example, 7 users consulting in the current service, 2 users consulting with high risk, 4 users consulting with medium risk, and 1 user consulting with low risk. If a user consultation AI fails to solve the problem of user consultation in time, the user consultation AI can be automatically transferred to a manual customer service, and if the user 3 repeatedly enters the line, the consultation problem of the user 3 is indicated that the consultation problem of the user 3 may not be solved in time, the consultation session of the user 3 can be transferred to a manual customer service link, a manual session frame can be unfolded on an AI workbench, the session content of the manual customer service and the user 3 is displayed, and the manual customer service can be directly communicated with the user 3 in the session frame, so that the consultation problem of the user 3 is conveniently solved in time.

In one embodiment, after entering the manual service mode in step 304, the method further includes processing the target consultation request using a pre-defined large model in response to the instructions for hosting the manual service.

In the embodiment, in the manual customer service mode, if the manual customer service determines that the consultation problem of the user is solved, the hosting instruction can be actively triggered, and the standardized ending service is transferred to the AI large model for processing, so that the labor cost is saved.

In the embodiment of the application, the manual customer service mode is used for processing the problem of high-risk high-value users, the AI hosting mode is used for processing the repeated and trivial dialogue type problem in the consultation process of the users, and when the property of the dialogue with the users is changed, the two modes can be switched. For example, in a manual customer service mode, if the user's problem has been solved, the AI mode can be hosted for some "tail-sweeping" services. In the AI hosting mode, if some high-risk scenes are encountered, the AI hosting mode can be automatically forwarded to the manual customer service mode for better service. The flexibility and the efficiency of the consultation request processing are improved.

In a practical scenario, where analysis is based on a large amount of historical online session data, the session distribution is less than 12 messages for about 33% of sessions than for example shown in fig. 10, at least 33% of sessions can be carried in AI mode considering the AI large model capability, and the 33% of problems can be solved based on AI model capability. As AI models become more capable, AI-treatable scenarios and problems become more and more, the system may follow the capabilities of AI models to evolve progressively.

According to the consultation request processing method, based on the manual service product form and service flow redesigned by the AI, the link for starting and monitoring the robot service process is designed on the basis of the complete service capability constructed by LLM and LLM AGENTS (large model agency) on the basis of the service full link, so that risks are effectively found and manual intervention is timely performed. Under the manual customer service mode, copilot capabilities are redesigned based on AI agents, and the agents can provide more complete-flow and real-time auxiliary capabilities. The manual workbench is split into the manual customer service mode and the AI hosting mode, and tasks can be switched between the two modes according to specific service conditions, so that the processing capacity of manual service of the workbench can be effectively improved, and the service cost is reduced. Meanwhile, precious manual resources can be utilized more efficiently, and better quality service can be provided for high-value high-risk users. And the service process can be monitored and evaluated, and the risk can be early warned in time.

In a practical scene, at least '33%' trivial consultation of AI processing can be used, so that service capability and efficiency can be greatly improved. AI can provide much more parallel processing capability than human, can surpass the limit of human handling capability, provide 1 to 100 or even higher parallel task processing capability, AI can provide standardized service process, guarantee service quality, AI-based service will not produce unnecessary loss, i.e. there is no concept of "invalid service". And with the enhancement of AI processing capability, the covered scene is more and more, the number of manual customer service required by the platform is less and less, and particularly, the manual customer service of low level for repeated labor is performed. Furthermore, the cost of the large model itself can become marginal as the scene is overlaid.

Please refer to fig. 11, which is an embodiment of a method for processing a consultation request about merchandise information, the method can be executed by the electronic device 1 shown in fig. 1, and can be applied to the application scenario of the consultation request processing shown in fig. 2, so as to implement screening and splitting of specific problems of user consultation in the manual service stage of the consultation request, process the specific problems by fully utilizing the capability of an AI large model for the known preset type consultation, reduce the participation of manual customer service, improve the processing efficiency of the consultation request, and automatically switch manual processing for the problem that the large model cannot process, thereby not only ensuring the service experience of the user, but also improving the resource utilization rate. In this embodiment, the terminal 220 is taken as an execution end, and compared with the foregoing embodiment, in this embodiment, a consultation request processing scenario for a user about commodity information in an e-commerce scenario is taken as an example, the method includes the following steps:

Step 1101, responding to the manual processing instruction about the commodity information consultation request, and acquiring the session context of the commodity information consultation request. The commodity information includes, but is not limited to, knowledge information related to the commodity, order information related to the commodity, service information related to the commodity, and the like.

Step 1102, identifying whether the current problem to be processed in the session of the commodity information consultation request is a preset type problem according to the session context.

And 1103, if the current problem to be processed is a preset type problem, generating reply information corresponding to the current problem to be processed by using a preset large model.

And 1104, if the current problem to be processed is not a preset type of problem, accessing the conversation of the commodity information consultation request into the manual customer service link, and generating reply information of the current problem to be processed according to the input information of the manual customer service.

The details of each step of the above method may be referred to the related descriptions of the above embodiments, which are not repeated herein.

Please refer to fig. 12, which is an apparatus 1200 for processing a consultation request according to an embodiment of the present application, which is applicable to the electronic device 1 shown in fig. 1 and is applicable to the application scenario of the consultation request processing shown in fig. 2, so as to implement screening and splitting of specific problems of user consultation in a manual service stage of the consultation request, and fully utilize the capability of an AI large model for processing for the consultation of a known preset type, thereby reducing the participation of manual service, improving the processing efficiency of the consultation request, and automatically switching manual processing for the problem that the large model cannot process, so as to not only ensure the service experience of the user, but also improve the resource utilization rate. The device comprises an acquisition module 1201, an identification module 1202, a first processing module 1203 and a second processing module 1204, wherein the functional principle of each module is as follows:

an obtaining module 1201 is configured to obtain a session context of the target consultation request in response to a manual processing instruction for the target consultation request.

The identifying module 1202 is configured to identify, according to the session context, whether a current problem to be processed in the session of the target consultation request is a preset type of problem.

The first processing module 1203 is configured to generate reply information corresponding to the current problem to be processed by using a preset large model if the current problem to be processed is a preset type problem.

And the second processing module 1204 is configured to access the session of the target consultation request to the artificial customer service link if the current problem to be processed is not a preset type of problem.

In one embodiment, the system further comprises a triggering module, which is used for responding to the consultation request of the user and acquiring the user characteristics of the user. And judging whether the user is a target user according to the user characteristics. And if the user is a target user, triggering a manual processing instruction aiming at the consultation request.

In one embodiment, the system further comprises a triggering module, which is used for responding to the consultation request of the user and acquiring the session context corresponding to the consultation request. And identifying the category of the consultation scene to which the consultation request currently belongs according to the session context. And if the consultation scene category is a target scene category requiring manual customer service intervention, triggering a manual processing instruction aiming at the consultation request.

In an embodiment, the triggering module is further configured to access the consultation request to the robot customer service link in response to the consultation request of the user. And acquiring the conversation context of the robot customer service and the user.

In an embodiment, the identification module is configured to input the session context into a preset large model, where the preset large model is configured to analyze the session context, and output a category of the consultation scene to which the consultation request currently belongs.

In an embodiment, the second processing module 1204 is configured to generate, after the session of the target consultation request is accessed to the artificial customer service link, reply information of the current to-be-processed problem according to the input information of the artificial customer service.

In an embodiment, the second processing module 1204 is configured to generate, in response to an auxiliary request for the current problem to be processed by the human customer service, auxiliary information corresponding to the auxiliary request through a preset large model. And acquiring input information determined by the manual customer service according to the auxiliary information. Responding to a reply instruction of the manual customer service to the current problem to be processed, and sending the input information as reply information of the current problem to be processed.

In an embodiment, the first processing module 1203 is configured to identify a current intention of the user according to the session context. And routing the current intention to a corresponding large model proxy module so that the large model proxy module generates matched reply information according to the current problem to be processed.

In an embodiment, the first processing module 1203 is configured to input the session context into a pre-trained intent recognition model, where the intent recognition model is configured to analyze the session context and recognize a current intent of the user, and the intent recognition model is obtained by training according to a preset session sample set, and the preset session sample is marked with a corresponding user intent.

In an embodiment, the system further comprises a transfer module, wherein the transfer module is used for detecting the current session context of the target consultation request after the reply information corresponding to the current problem to be processed is generated by using the preset large model, and judging whether the current problem to be processed is solved or not according to the current session context by using the preset large model. If the current problem to be processed is not solved, accessing the session of the target consultation request into the artificial customer service link, and generating reply information of the current problem to be processed according to the input information of the artificial customer service.

In an embodiment, the transfer module is further configured to process the target consultation request by using a preset large model in response to a hosting instruction of the human customer service after accessing the session of the target consultation request to the human customer service link and generating the reply information of the current to-be-processed problem according to the input information of the human customer service.

For a detailed description of the above consultation request processing apparatus 1200, please refer to the description of the related method steps in the above embodiment, the implementation principle and technical effects are similar, and the detailed description of this embodiment is omitted herein.

Fig. 13 is a schematic structural diagram of a cloud device 130 according to an exemplary embodiment of the present application. The cloud device 130 may be used to run the methods provided in any of the embodiments described above. As shown in fig. 13, the cloud device 130 may include a memory 1304 and at least one processor 1305, one processor being exemplified in fig. 13.

Memory 1304, for storing computer programs, may be configured to store various other data to support operations on cloud device 130. The memory 1304 may be an object store (Object Storage Service, OSS).

The memory 1304 may be implemented by any type or combination of volatile or nonvolatile memory devices 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 or optical disk.

The processor 1305 is coupled to the memory 1304, and is configured to execute the computer program in the memory 1304, so as to implement the solutions provided by any of the method embodiments, and specific functions and technical effects that can be implemented are not described herein.

Further, as shown in fig. 13, the cloud device further includes a firewall 1301, a load balancer 1302, a communication component 1306, a power component 1303, and other components. Only some components are schematically shown in fig. 13, which does not mean that the cloud device only includes the components shown in fig. 13.

In one embodiment, the communication component 1306 in fig. 13 is configured to facilitate wired or wireless communication between the device in which the communication component 1306 is located and other devices. The device in which the communication component 1306 is located may access a wireless network based on a communication standard, such as a WiFi,2G, 3G, 4G, LTE (Long Term Evolution, long term evolution, LTE for short), 5G, or a combination thereof. In one exemplary embodiment, the communication component 1306 receives broadcast signals or broadcast related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the Communication component 1306 further includes a Near Field Communication (NFC) module to facilitate short range communications. For example, the NFC module may be implemented based on radio frequency identification (Radio Frequency Identification, RFID) technology, infrared data Association (IrDA) technology, ultra Wide Band (UWB) technology, bluetooth (BT) technology, and other technologies.

In one embodiment, the power supply unit 1303 shown in fig. 13 provides power to various components of the device in which the power supply unit 1303 is located. Power component 1303 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device in which the power component resides.

The embodiment of the application also provides a computer readable storage medium, wherein computer executable instructions are stored in the computer readable storage medium, and when the processor executes the computer executable instructions, the method of any of the previous embodiments is realized.

Embodiments of the present application also provide a computer program product comprising a computer program which, when executed by a processor, implements the method of any of the preceding embodiments.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described device embodiments are merely illustrative, e.g., the division of modules is merely a logical function division, and there may be additional divisions of actual implementation, e.g., multiple modules may be combined or integrated into another system, or some features may be omitted or not performed.

The integrated modules, which are implemented in the form of software functional modules, may be stored in a computer readable storage medium. The software functional modules described above are stored in a storage medium and include instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or processor to perform some of the steps of the methods of the various embodiments of the application.

It should be appreciated that the Processor may be a central processing unit (Central Processing Unit, abbreviated as CPU), or may be other general purpose Processor, digital signal Processor (DIGITAL SIGNAL Processor, abbreviated as DSP), application SPECIFIC INTEGRATED Circuit (ASIC), or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present application may be embodied directly in a hardware processor for execution, or in a combination of hardware and software modules in a processor for execution. The memory may include a high-speed RAM (Random Access Memory ) memory, and may further include a nonvolatile memory NVM (Nonvolatile memory, abbreviated as NVM), such as at least one magnetic disk memory, and may further be a U-disk, a removable hard disk, a read-only memory, a magnetic disk, or an optical disk.

The storage medium may be implemented by any type of volatile or non-volatile Memory device or combination thereof, such as Static Random-Access Memory (SRAM), electrically erasable programmable Read-Only Memory (ELECTRICALLY ERASABLE PROGRAMMABLE READ ONLY MEMORY EEPROM), erasable programmable Read-Only Memory (Erasable Programmable Read-Only Memory (EPROM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), read-Only Memory (ROM), magnetic Memory, flash Memory, magnetic disk, or optical disk. A storage media may be any available media that can be accessed by a general purpose or special purpose computer.

An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an Application SPECIFIC INTEGRATED Circuits (ASIC). It is also possible that the processor and the storage medium reside as discrete components in an electronic device or a master device.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article of apparel, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article of apparel, or apparatus. Without further limitation, an element defined by the phrase "comprising one does not exclude the presence of other like elements in a process, method, article of apparel, or apparatus that comprises the element.

The foregoing embodiment numbers of the present application are merely for the purpose of description, and do not represent the advantages or disadvantages of the embodiments.

From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g. ROM/RAM, magnetic disk, optical disk) comprising several instructions for causing a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method of the embodiments of the present application.

In the technical scheme of the application, the related information such as user data and the like is collected, stored, used, processed, transmitted, provided, disclosed and the like, which are all in accordance with the regulations of related laws and regulations and do not violate the popular public order.

The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the application, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.

Claims

1. A consultation request processing method, characterized by comprising:

responding to a manual processing instruction aiming at a target consultation request, and acquiring a session context of the target consultation request;

according to the session context, identifying whether the current problem to be processed in the session of the target consultation request is a preset type problem or not;

if the current problem to be processed is the preset type problem, generating reply information corresponding to the current problem to be processed by using a preset large model;

And if the current problem to be processed is not the problem of the preset type, accessing the session of the target consultation request into a manual customer service link.

2. The method of claim 1, wherein triggering the manual processing instruction for the target advisory request comprises:

Responding to a consultation request of a user, and acquiring user characteristics of the user;

judging whether the user is a target user or not according to the user characteristics;

and if the user is a target user, triggering a manual processing instruction aiming at the consultation request.

3. The method of claim 1, wherein triggering the manual processing instruction for the target advisory request comprises:

Responding to a consultation request of a user, and acquiring a session context corresponding to the consultation request;

identifying the category of the consultation scene to which the consultation request currently belongs according to the session context;

and if the consultation scene category is a target scene category requiring manual customer service intervention, triggering a manual processing instruction aiming at the consultation request.

4. The method of claim 3, wherein the obtaining, in response to the consultation request of the user, the session context corresponding to the consultation request includes:

responding to the consultation request of the user, and accessing the consultation request into a robot customer service link;

And acquiring the session context of the robot customer service and the user.

5. The method of claim 3, wherein the identifying the category of the counseling scenario to which the counseling request currently belongs according to the session context comprises:

and inputting the session context into a preset large model, wherein the preset large model is used for analyzing the session context and outputting the category of the consultation scene to which the consultation request currently belongs.

6. The method of claim 1, further comprising generating reply information of the current pending problem based on entered information of a human customer service after the session of the target consultation request is accessed to a human customer service link;

the generating the reply information of the current to-be-processed problem according to the input information of the manual customer service comprises the following steps:

Responding to an auxiliary request of the manual customer service for the current problem to be processed, and generating auxiliary information corresponding to the auxiliary request through a preset large model;

acquiring the input information determined by the manual customer service according to the auxiliary information;

And responding to a reply instruction of the manual customer service to the current problem to be processed, and sending the input information as reply information of the current problem to be processed.

7. The method of claim 1, wherein the generating reply information corresponding to the current problem to be processed using a preset large model includes:

identifying the current intention of the user according to the session context;

and routing the current intention to a corresponding large model proxy module so that the large model proxy module generates matched reply information according to the current to-be-processed problem.

8. The method of claim 7, wherein the identifying the current intent of the user based on the session context comprises:

Inputting the conversation context into a pre-trained intention recognition model, wherein the intention recognition model is used for analyzing the conversation context and recognizing the current intention of the user, the intention recognition model is obtained by training according to a preset conversation sample set, and the preset conversation sample is marked with the corresponding user intention.

9. The method of claim 1, further comprising, after the generating reply information corresponding to the current problem to be processed using a preset large model:

Detecting the current session context of the target consultation request, and judging whether the current problem to be processed is solved or not according to the current session context by using the preset large model;

if the current problem to be processed is not solved, accessing the session of the target consultation request into a manual customer service link, and generating reply information of the current problem to be processed according to the input information of the manual customer service.

10. The method of claim 1, further comprising, after the session of the target consultation request is accessed to a manual service link and the reply message of the current pending problem is generated according to input information of the manual service:

And responding to the hosting instruction of the manual customer service, and processing the target consultation request by using the preset large model.

11. An electronic device, comprising:

At least one processor, and

A memory communicatively coupled to the at least one processor;

Wherein the memory stores instructions executable by the at least one processor to cause the electronic device to perform the method of any one of claims 1-10.

12. A computer readable storage medium having stored therein computer executable instructions which, when executed by a processor, implement the method of any of claims 1-10.

13. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any of claims 1-10.