Request processing method and device, storage medium and program product

Abstract

The embodiment of the invention provides a request processing method and device, a storage medium and a program product. In the request processing method, when a user has a service request, a main intelligent processing module can obtain request context information, and can automatically plan a target processing scheme based on the request context information. The main intelligent processing module can schedule the sub-intelligent processing modules corresponding to the processing links in the target processing scheme to process the service request, so as to dynamically distribute the processing task of the service request to the appropriate sub-intelligent processing modules, and reduce the dependence on manual customer service. A part of the sub-intelligent processing modules can process uncertain and fuzzy information by using the target language model, and can make a reasonable decision according to incomplete or uncertain data. And the main intelligent processing module and the sub intelligent processing module cooperate with each other, so that service requests of users in different scenes can be efficiently coped with. And furthermore, the automatic processing proportion of user complaints can be improved, and the labor cost brought by manual customer service is reduced.

Description

Technical Field

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

[0002] With the rapid development of digital commerce, e-commerce platforms have become an integral part of people's daily lives. Consumers can easily purchase a wide range of goods and services through e-commerce platforms, enjoying a convenient shopping experience. However, as transaction volume continues to increase, first-time users may encounter various problems during the shopping process, such as product quality issues, logistics delays, and inadequate after-sales service. These issues can lead to first-time user dissatisfaction and complaints. E-commerce platforms need to establish efficient and professional e-commerce customer service systems to handle first-time user complaints.

[0003] Traditional e-commerce customer service systems primarily rely on process and rule engines to handle simple user complaints, while manual customer service representatives handle more complex and varied complaints. This traditional approach lacks automated processing capabilities, resulting in significant manual processing pressure and significant labor costs. Therefore, a new solution is needed. Summary of the Invention

[0004] Embodiments of the present application provide a request processing method, device, storage medium, and program product to reduce the labor cost of request processing.

[0005] An embodiment of the present application provides a request processing method, which is applied to a main intelligent processing module in a request processing system, wherein the request processing system also includes multiple sub-intelligent processing modules that cooperate with the main intelligent processing module, including: responding to a service request of a first user, and obtaining the request context information of the first user; planning a processing plan for the service request based on the request context information to obtain a target processing plan, wherein the target processing plan includes at least one processing link; scheduling the sub-intelligent processing modules corresponding to each of the at least one processing link to process the service request based on the request context information; among the at least one processing link, the sub-intelligent processing modules corresponding to at least some of the processing links are used to call the target model to process the service request based on the request context information.

[0006] An embodiment of the present application also provides an electronic device, comprising: a memory and a processor; the memory is used to store one or more computer instructions; the processor is used to execute the one or more computer instructions to: execute the steps in the method provided in the embodiment of the present application.

[0007] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the steps of the method provided in the embodiment of the present application.

[0008] An embodiment of the present application further provides a computer program product, comprising: a computer program / instructions, which, when executed by a processor, can implement the steps of the method provided in the embodiment of the present application.

[0009] In an embodiment of the present application, when a user makes a service request, the main intelligent processing module can obtain the user's request context information and automatically plan a target processing solution for processing the service request based on the request context information. The main intelligent processing module can schedule the sub-intelligent processing modules corresponding to at least one processing link in the target processing solution to process the service request according to the request context information. Based on this embodiment, the main intelligent processing module can flexibly schedule the sub-intelligent processing modules according to the request context information, so as to dynamically assign the processing tasks of the service request to the appropriate sub-intelligent processing modules, thereby reducing dependence on manual customer service. Some sub-intelligent processing modules can use the target language model to process uncertain and fuzzy information, and can make reasonable decisions based on incomplete or uncertain data. The main intelligent processing module and the sub-intelligent processing modules cooperate with each other to efficiently respond to user service requests in different scenarios. Furthermore, it is conducive to improving the proportion of automated processing of user complaints in the service platform and significantly reducing the labor costs brought by manual customer service. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0011] Figure 1 A flowchart of a request processing method provided by an exemplary embodiment of the present application;

[0012] Figure 2 A schematic diagram of the structure of a demand processing system with multiple intelligent processing modules provided by an exemplary embodiment of the present application;

[0013] Figure 3 A schematic diagram of the working link of the main intelligent processing module provided by an exemplary embodiment of the present application;

[0014] Figure 4 A schematic diagram of a working link of an initiator module provided in an exemplary embodiment of the present application;

[0015] Figure 5 A schematic diagram of a working link of a negotiation submodule provided in an exemplary embodiment of the present application;

[0016] Figure 6 A schematic diagram of a working link of a first judgment submodule provided in an exemplary embodiment of the present application;

[0017] Figure 7 A schematic diagram of a working link of a second judgment submodule provided in an exemplary embodiment of the present application;

[0018] Figure 8 A schematic structural diagram of an electronic device provided as an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0019] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0020] The terms used in the examples of this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms "a," "the," and "the" used in the examples of this application and the appended claims are also intended to include plural forms, and unless the context clearly indicates otherwise, "a plurality" generally includes at least two, but does not exclude the inclusion of at least one.

[0021] It should be understood that the term "and / or" as used herein is merely a description of the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0022] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a product or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such a product or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the product or system comprising the element.

[0023] The customer service complaint handling system of an e-commerce platform is a complex system, involving multiple interactions (member, merchant, and customer service) between multiple actors (complainants, respondents, and platform customer service). The complaint handling process is completed in multiple stages (initiation, negotiation, adjudication, fulfillment, appeal, and rights protection and mediation). Some e-commerce customer service systems primarily rely on process engines and rule engines to handle simple user complaints, while manual customer service is used to handle more complex and varied complaints. The process engine primarily defines the workflow for handling complaint events in the customer service system. By breaking down complex customer service processes into a series of orderly steps, it ensures that each complaint is properly handled according to the established process. The rule engine primarily defines a series of conditions and actions to enable rapid classification, judgment, and response to complaints. However, process engines and rule engines struggle to effectively cover the complex and varied complaint scenarios, resulting in many complaints being manually handled, which consumes significant labor costs. Due to insufficient human resources, customer service systems face significant efficiency bottlenecks.

[0024] In response to the above technical problems, a solution is provided in some embodiments of the present application. The technical solutions provided in each embodiment of the present application are described in detail below with reference to the accompanying drawings.

[0025] Figure 1 is a flowchart of a request processing method provided by an exemplary embodiment of the present application, which may include: Figure 1 Steps shown:

[0026] Step 101: Respond to a service request from a first user and obtain request context information of the first user.

[0027] Step 102: Plan a processing solution for the service request based on the request context information to obtain a target processing solution, where the target processing solution includes at least one processing step.

[0028] Step 103: Schedule the sub-intelligent processing modules corresponding to each of the at least one processing link to process the service request according to the request context information; in the at least one processing link, the sub-intelligent processing modules corresponding to at least some of the processing links are used to call the target model to process the service request according to the request context information.

[0029] Embodiments of the present application can be applied to a request processing system comprising multiple intelligent processing modules. The request processing system can provide request processing services to multiple complaint channels of a service platform, which may include but is not limited to e-commerce platforms, financial service platforms, online education platforms, etc. The intelligent processing module in the request processing system can be an agent. An agent is an autonomous computer system that can perceive external input information, process the perceived information, and perform goal-oriented reasoning and decision-making actions. In some embodiments, the agent may include: a perception module, a decision / reasoning engine, an action module, and a memory system. The perception module is used to receive external input (such as text, images, voice, etc.) as a basis for decision-making. The decision / reasoning engine is used to plan tasks and perform complex reasoning. In some embodiments, the agent's decision / reasoning engine can be driven by a machine learning model (such as a language model) to implement complex tasks such as context understanding, logical reasoning, task decomposition, and strategy formulation based on the capabilities of the machine learning model. The action module is used to perform operations such as outputting natural language, calling interfaces, operating tools, and generating code. The memory system is used to rely on external storage (such as a vector database or state manager) to store short-term or long-term memory and maintain task continuity. The agent also has the ability to learn and can continuously optimize its performance through prompt engineering, in-context learning (ICL), fine-tuning or reinforcement learning.

[0030] In this embodiment, when applied to request processing scenarios, the agent primarily infers and plans a response plan based on the knowledge base, service specifications, and request-related features. It then uses tools provided by the service platform to execute the planned response plan, thereby intelligently resolving user service requests. In a system where multiple agents collaborate, they can solve complex problems through mutual interaction, collaboration, and negotiation. Each agent possesses a degree of autonomy, intelligence, and adaptability, capable of making decisions and taking actions based on its own perceptions and goals.

[0031] The request processing system provided in the embodiment of the present application may include at least a main intelligent processing module and multiple sub-intelligent processing modules that cooperate with the main intelligent processing module. In this embodiment, the main intelligent processing module can schedule the execution of other sub-intelligent processing modules. The following will be illustrated with reference to the steps.

[0032] In step 101, the main intelligent processing module may respond to a service request from a first user and obtain the request context information of the first user. The first user refers to the user initiating the service request, for example, a buyer in an e-commerce scenario. The first user's service request may be used to query order status, initiate purchase inquiries, initiate returns and exchanges, or file a complaint, although this embodiment does not limit this. For example, if the service request is used to file a complaint, the first user may submit the complaint through a complaint channel provided by the service platform. This complaint channel may include an online customer service channel, a robot customer service channel, a consultation channel, an offline telephone channel, and so on. In some embodiments, the client provides a customer service dialogue portal. The first user may access the customer service dialogue interface through the customer service dialogue portal and submit a service request through one or more rounds of conversation within the customer service dialogue interface. The main intelligent processing module may respond to the first user's service request and engage in at least one round of conversation with the first user to obtain the request context information based on the at least one round of conversation. Optionally, in this embodiment, multiple complaint channels of the service platform may share data. In other words, the context information of complaints initiated by the same user across different channels may be shared with other complaint channels of the service platform. In this embodiment, the main intelligent processing module may obtain the request context information of the first user from at least one complaint channel according to the identification information of the first user.

[0033] Request context information refers to background information related to the first user's service request, used to describe the details of the service request to facilitate accurate understanding of the service request. Request context information may include information provided by the first user during the conversation, or information retrieved from a relevant database on the service platform based on the information provided by the first user.

[0034] For example, the request context information may include at least one of the following: the first user's basic information, order details, transaction history, problem description, timeline, customer service records, and logistics information. The problem description may be provided by the first user. For example, if the service request is used to initiate a complaint, the problem description may include: the complaint type (such as product quality issues, logistics delays, after-sales service issues, etc.), a detailed description of the complaint (problems encountered, desired solutions, etc.), and supporting evidence (photos, videos, chat logs, email screenshots, etc.).

[0035] In step 102, the main intelligent processing module may plan the processing solution for the service request based on the request context information to obtain a target processing solution. The processing solution for the service request is planned to formulate steps and action paths for resolving the service request. Optionally, the main intelligent processing module may use a predefined workflow to plan the processing solution for the service request. Optionally, the main intelligent processing module may use an artificial intelligence (AI) model to plan the processing solution for the service request. Alternatively, the main intelligent processing module may first use a predefined workflow to plan the processing solution for the service request, and then use the AI ​​model to improve the planned processing solution, which is not limited in this embodiment.

[0036] The target processing solution may include at least one processing step. In this embodiment, a single processing step is used to perform a relatively independent task, focusing on completing a specific part of the request processing process. If the target processing solution includes multiple processing steps, the multiple processing steps are generally executed in a specific order. Taking the service request for initiating a complaint in an e-commerce scenario as an example, the processing steps for this type of service request may include: the step of initiating the complaint, the step of negotiating the complaint, the step of intelligent judgment, the step of manual assisted judgment, etc. The target processing solution may include any one of the above processing steps, or it may be a combination of any number of the above processing steps. When multiple processing steps are combined, the multiple processing steps have a certain order of execution. For example, the target processing solution may include: executing the step of initiating the complaint first, then executing the step of negotiating the complaint. Alternatively, the target processing solution may include: executing the step of initiating the complaint first, then executing the step of intelligent judgment, and then executing the step of manual assisted judgment, etc., which are not listed one by one.

[0037] In step 103, the main intelligent processing module can schedule the sub-intelligent processing modules corresponding to each of the at least one processing links to process the service request according to the request context information. In this embodiment, each processing link can correspond to a sub-intelligent processing module, and the sub-intelligent processing module can use the request context information as input and use a specific workflow to perform specific processing operations on the service request. The sub-intelligent processing modules corresponding to different processing links use different workflows to process the service request. The sub-intelligent processing modules corresponding to at least some processing links can be used to call the target model to process the service request according to the request context information, which will be specifically introduced in subsequent embodiments.

[0038] In this embodiment, when a user makes a service request, the main intelligent processing module can obtain the user's request context information and automatically plan a target processing solution for processing the service request based on the request context information. The main intelligent processing module can schedule the sub-intelligent processing modules corresponding to at least one processing link in the target processing solution to process the service request based on the request context information. Based on this embodiment, the main intelligent processing module can flexibly schedule the sub-intelligent processing modules based on the request context information, thereby dynamically assigning the processing task of the service request to the appropriate sub-intelligent processing module, reducing dependence on manual customer service. Some sub-intelligent processing modules can use the target language model to process uncertain and ambiguous information, and can make reasonable decisions based on incomplete or uncertain data. The main intelligent processing module and the sub-intelligent processing modules work together to efficiently respond to user service requests in different scenarios. Furthermore, it is conducive to increasing the proportion of automated processing of user complaints in the service platform and significantly reducing the labor costs brought by manual customer service.

[0039] Furthermore, the platform's multiple service channels can share data, allowing the contextual information of service requests initiated by the same user across different channels to be shared with other channels. This contextual information can be used to develop consistent solutions, ensuring consistent problem resolution across the platform's multiple channels. This eliminates the need for users to repeatedly consult between channels, improving the user experience.

[0040] In some optional embodiments, the request processing system may be a three-layer structure, which may include, in addition to the main intelligent processing module and multiple sub-intelligent processing modules, multiple general capability modules, which may be general capability agents. Figure 2 As shown, in an optional embodiment, the multiple sub-intelligent processing modules of the request processing system may include: an initiating agent (i.e., an initiating sub-module), a negotiating agent (i.e., a negotiating sub-module), a first decision agent (i.e., a first decision sub-module), and a second decision agent (i.e., a second decision sub-module). Figure 2 As shown, the multiple general capability modules may include, but are not limited to, at least one of: a demand identification agent (i.e., a demand identification module), a notification agent (i.e., a notification module), an outbound call agent (i.e., an outbound call module), a credential agent (i.e., a credential intelligent processing module), and a risk control agent (i.e., a risk control module). Any general capability module can be called by any sub-intelligent processing module, and any sub-intelligent processing module can call one or more general capability modules.

[0041] In different scenarios, service requests initiated by users for different reasons are handled in different ways. Each module in the request processing system can rely on the knowledge base to obtain the knowledge required to process service requests in different scenarios, such as Figure 2 The following are the shipping scenario knowledge base, attitude scenario knowledge base, and commitment scenario knowledge base. Taking the complaint scenario as an example, the shipping scenario knowledge base stores relevant knowledge for handling complaints in shipping scenarios, the attitude scenario knowledge base stores relevant knowledge for handling complaints in attitude scenarios, and the commitment scenario knowledge base stores relevant knowledge for handling complaints in commitment scenarios. For example, the knowledge for handling service requests in the shipping scenario may include: If the complaint is due to delayed shipment, the response method is to provide the customer with the latest logistics information and explain the delay reason (such as during peak holiday periods). If the delay is long, consider offering a small amount of compensation or a coupon. If the complaint is due to package loss, the response method is to immediately contact the logistics company to confirm the package status. If it is indeed lost, quickly arrange for reshipment or a full refund, and provide the customer with a detailed follow-up plan. If the complaint is due to damaged goods, the response method is to instruct the customer to take photos of the damage as evidence. A replacement or refund is determined based on the severity of the damage, and a quick response is ensured to minimize customer inconvenience.

[0042] Optionally, the main intelligent processing module may search a target knowledge base based on the request context information to obtain the target knowledge information required to process the service request. The target knowledge base may be at least one of a shipping scenario knowledge base, an attitude scenario knowledge base, and a commitment scenario knowledge base. Before searching the target knowledge base based on the request context information, the main intelligent processing module may further identify the complaint scenario corresponding to the service request based on the request context information and determine the target knowledge base based on the complaint scenario. For example, if the complaint scenario is a shipping scenario, the target knowledge base may be determined to be a shipping scenario knowledge base.

[0043] Optionally, the main intelligent processing module can extract features from the request context information to obtain target request features corresponding to the request context information; optionally, target request features refer to quantifiable or categorizable data indicators extracted from the first user's request context information that can reflect key dimensions such as the content of the service request, user behavior, order status, and service process, and are used to support classification and identification of service requests, formulation of processing strategies, priority judgment, risk assessment, and automated decision-making. The main intelligent processing module can utilize Figure 2 The feature center’s ability to extract target request features from request context information and store the extracted target request features in Figure 2 The feature centers shown are for subsequent use and can be reused among multiple agents.

[0044] Optionally, the main intelligent processing module can extract features from the request context information from at least one dimension to obtain target request features. The at least one dimension may include at least one of the order dimension, the membership dimension, and the logistics dimension. Taking the service request for initiating a complaint as an example, the complaint feature of the order dimension (i.e. Figure 2 The order characteristics (illustrative order characteristics) may include: the first-level complaint reason ID (such as "product quality problem", "delayed delivery", etc.), whether the delivery is urged, whether the merchant fulfills the contract, the total order amount, order status, the time interval between the order time and the complaint, the number of after-sales records, the return type, and at least one of the promotion participation. Member-level complaint characteristics (i.e. Figure 2 The member characteristics shown in the figure may include: member level, registration time, number of historical complaints, repurchase rate, customer satisfaction score, blacklist identification and channel preference. Among them, the complaint characteristics of the logistics dimension (i.e. Figure 2 Illustrated logistics characteristics may include at least one of the following: estimated delivery time, actual delivery time, logistics status, logistics company, delivery method, logistics track update frequency, receipt status, abnormal logistics flags, and freight insurance usage. In addition to the aforementioned dimensions, customer service interaction characteristics may also be captured, such as initial response time and number of manual interventions.

[0045] Optionally, the main intelligent processing module may use a natural language processing (NLP)-based processing method or a rule engine-based method when extracting features from the request context information. The following uses a service request for initiating a complaint as an example to provide exemplary explanations.

[0046] Among them, the natural language-based processing method is used to process the text content of the request context information and extract information at the semantic level. Optionally, one or more tools can be used to perform keyword recognition on the request context information, such as identifying keywords such as "slow delivery", "quality issues", and "poor customer service attitude" through a dictionary or model. Optionally, one or more tools can be used to perform intent recognition on the request context information, such as determining the core demands of the complaint, such as "I want a refund" and "I want an exchange". Optionally, one or more tools can be used to perform sentiment analysis on the request context information, such as identifying whether the user's emotions are positive, neutral, or negative. Optionally, one or more tools can be used to perform entity recognition on the request context information, such as identifying entity information such as order number, product name, logistics company, etc. Among them, the one or more tools can be at least one of: regular expression, named entity recognition, and intent classification model, which is not limited in this embodiment.

[0047] The rule engine-based approach extracts features based on pre-defined logical rules. For example, if the request context contains "not received yet" and "urgent need," the complaint feature is marked as "expedited delivery." If the request context contains "order amount > 1,000 yuan" and "complaint frequency ≥ 2 times / month," the complaint feature is marked as "key customer complaint."

[0048] Based on the above implementation, identifying target request features in request context information from at least one dimension can assist in quickly identifying complaint-related information, thereby assisting the target model in customizing personalized solutions.

[0049] The main intelligent processing module can plan the processing plan for the service request based on the acquired target knowledge information and target request characteristics to obtain the target processing plan.

[0050] In some optional embodiments, the main intelligent processing module may utilize the target language model to plan a processing solution for the service request. Alternatively, the main intelligent processing module may construct a target prompt word based on the target knowledge information and the target request characteristics, and based on the target prompt word, invoke the target language model to plan a processing solution for the service request, thereby obtaining the target processing solution.

[0051] In the above and following embodiments of the present application, the target language model can be a large language model, which is a natural language processing (NLP) model that has been trained on a large scale. Large language models are usually built based on deep learning technology and are trained on large-scale training data sets, so that they can show powerful performance when processing natural language tasks. The embodiments of the present application do not limit the number of model parameters supported by the model, with the goal of meeting actual needs. During the training process, these parameters are continuously adjusted and optimized based on the difference between the predicted results and the actual results of the large language model to improve the prediction accuracy of the large language model. In some embodiments, the large language model usually adopts an advanced neural network architecture, such as a transformer (Transformer) architecture to build a model structure, which enables the large language model to capture complex patterns in the text and process long-distance dependencies. After sufficient training, the large language model has a powerful generation capability and can produce coherent and contextual text content based on given prompts.

[0052] In this embodiment, a large language model pre-trained on a large number of general data sets can be used as the base model of the target language model. In order to adapt to the specific application scenario of this application, that is, for processing user requests in e-commerce scenarios, the pre-trained large language model can be fine-tuned on a data set formed by a large number of user request cases (such as user complaint cases) to make the fine-tuned large language model more suitable for executing request processing tasks.

[0053] In other optional embodiments, the main intelligent processing module can use the request processing workflow to plan the processing scheme for the service request. Optionally, the main intelligent processing module determines a request processing workflow containing multiple process nodes, and the jump relationship between some process nodes in the request processing workflow is triggered by a specific jump condition. The main intelligent processing module searches the target knowledge base based on the request context information to obtain the target knowledge information, and extracts the features of the request context information. After obtaining the target request features, the main intelligent processing module can match the process nodes and jump conditions in the request processing workflow based on the target knowledge information and the target request features to obtain the target working link adapted to the service request, and determine the target processing scheme based on the target working link. Specifically, it can start from the starting node and proceed in the direction of the connection between the nodes. Every time a branch point with a condition is reached, it is decided which path to choose to continue moving forward based on the target knowledge information and the target request features until the end node is reached. The link formed between the starting node and the end node serves as the target working link adapted to the service request.

[0054] Based on the above implementation, the main intelligent processing module implements intelligent planning of service request processing solutions. In some embodiments, for simple service requests, the main intelligent processing module can plan request processing solutions through the request processing workflow. For complex service requests, the main intelligent processing module can leverage the powerful task processing capabilities of the target language model to plan solutions, achieving adaptation to different complaint scenarios, which helps reduce manual processing and lower labor costs.

[0055] In some optional embodiments, the service request of the first user is a complaint type request, and the corresponding request context information includes the complaint information of the first user against the second user. Figure 3 As shown, taking the complaint handling system implemented as a multi-agent system as an example, the processing scheme for the service request of the complaint type planned by the main process agent (i.e., the main intelligent processing module) may include: complaint progress check, complaint initiation, negotiation, intelligent judgment, manual judgment or complaint evaluation, etc., or it may include a scheme obtained by combining multiple schemes among the above schemes. Optionally, in the case that a complaint work order already exists for the service request, the main process agent may take over the complaint work order and track the complaint progress of the complaint work order. The complaint work order is an electronic document used to record and track the entire process of the service request. As Figure 3 As shown in the figure, AI models can be used to execute corresponding processing procedures in the steps of complaint initiation, negotiation, intelligent judgment, and manual judgment in the processing plan.

[0056] In some optional embodiments, when there is no complaint ticket for a service request, the main process agent can schedule the initiation submodule corresponding to the complaint initiation link, determine whether the service request meets the complaint access requirements based on the request context information, and create a complaint ticket corresponding to the service request if it meets the complaint access requirements. Figure 4 As shown, the initiating agent (i.e., the initiating submodule) can obtain the complaint form features and complaint access features corresponding to the service request based on the request context information. Among them, the complaint form is used to record the user's complaint information in a structured manner. Usually, the complaint form includes: basic customer information, complaint details, time and place of problem occurrence, expected solution, processing status tracking and other parts. The initiating agent can search the target knowledge base for the complaint form collection rules and complaint access rules corresponding to the service request based on the request context information; if the complaint form features meet the requirements of the form collection rules, and the complaint access features meet the requirements of the complaint access rules, the initiating agent can create the complaint work order. In some optional embodiments, such as Figure 4 As shown, the initiating agent can call the target language model to make form information decisions and initiate access decisions. When making form information decisions, the initiating agent can call the target language model based on the complaint form features and the form collection rules to determine whether the complaint form features meet the requirements of the form collection rules. When making access decisions, the initiating agent can call the target language model based on the complaint access features and the complaint access rules to determine whether the complaint access features meet the requirements of the complaint access rules.

[0057] Optionally, if the complaint form features do not meet the requirements of the form collection rules, the initiating agent may determine target form information to be collected and may engage in at least one round of dialogue with the first user to collect the target form information. After collecting the target form information, the initiating agent may re-execute the steps for obtaining the complaint form features, which will not be further described.

[0058] In some embodiments, as Figure 4 As shown, the initiating agent may first perform the form information decision step and, when the form information is complete, then perform the initiating admission decision. Optionally, if the complaint admission characteristics do not meet the requirements of the complaint admission rules, the initiating agent may determine the reason for the denial and inform the first user of the reason through a dialogue with the first user.

[0059] In some optional embodiments, the main intelligent processing module may schedule the negotiation submodule corresponding to the negotiation phase to execute a negotiation admission process and a solution recommendation process. During the negotiation admission process, the negotiation submodule may determine whether the service request meets the negotiation admission requirements based on the request context information, and, if the service request meets the negotiation admission requirements, determine a target negotiation solution by conducting at least one round of dialogue with the first user and / or the second user.

[0060] Optionally, in the negotiation access process, the negotiation submodule may obtain the real-time conversation information and / or historical negotiation information between the first user and the second user based on the request context information. For example, in an e-commerce scenario, the first user is usually a buyer user, and the second user is usually a seller user. The negotiation submodule may call the target language model based on the real-time conversation information and / or historical negotiation information, so that the target language model determines whether the service request meets the negotiation access requirements. Figure 5 As shown, if the service request does not meet the negotiation access requirements, the negotiation agent (ie, the negotiation submodule) may apply for platform intervention.

[0061] Optionally, in the solution recommendation process, the negotiation submodule can identify the negotiation intention of the second user through at least one round of dialogue with the second user. If the second user does not agree to negotiate, the negotiation submodule can request the platform to intervene. If the second user agrees to negotiate, the negotiation submodule can identify the target appeal information of the first user through at least one round of dialogue with the first user, and determine at least one recommended negotiation solution based on the target appeal information of the first user. The negotiation submodule can directly call the target language model to identify the target appeal information of the first user, or can call Figure 2 The illustrated demand recognition agent (ie, demand recognition module) is used to recognize the target demand information of the first user, which is not limited in this embodiment.

[0062] Alternatively, as Figure 5 As shown, the negotiation agent can call the target language model according to the target demand information of the first user, so that the target language model outputs at least one recommended negotiation solution. Figure 5 As shown, the negotiation agent can provide the at least one recommended negotiation solution to the second user (i.e., the seller) for reference, so that the second user can select a target negotiation solution. The negotiation agent can conduct at least one round of dialogue with the first user based on the target negotiation solution to identify the negotiation intention of the first user; if the first user agrees to negotiate, the target solution is determined to be a consensus solution. If the first user disagrees with the negotiation, the negotiation agent can recommend a new negotiation solution. Figure 5 As shown, if the buyer does not agree to the negotiation, the negotiation agent can re-call the target language model to generate a recommendation solution, which will not be repeated here.

[0063] In some optional embodiments, the main intelligent processing module can schedule the first decision submodule corresponding to the intelligent decision link, obtain the information required for the decision according to the request context information, and determine the target decision solution based on the information required for the decision. Figure 6 As shown, the first decision agent (ie, the first decision submodule) can execute the claim confirmation process, the decision information collection process, the solution decision process, and the solution submission process.

[0064] Optionally, in the request confirmation process, the first judgment submodule obtains the target request information of the first user based on the real-time conversation information between the first user and the second user and the request context information. Optionally, the first judgment submodule can call the target language model to determine whether the target request information of the first user can be recognized based on the real-time conversation information and the request context information. Alternatively, the first judgment submodule can call Figure 2 The illustrated demand recognition agent (ie, demand recognition module) is used to recognize the target demand information of the first user, which is not limited in this embodiment.

[0065] If not, then obtain the first user's supplementary demand information by calling or leaving a message to the first user; if so, determine the first user's target demand information based on the real-time conversation information, the request context information and the supplementary demand information. Figure 6 As shown, in the e-commerce field, for example, the first decision agent can obtain real-time conversation information and complaint information between buyers and sellers and use the target language model to confirm the request. If the information is insufficient, the first decision submodule can call the outbound call agent (i.e., the outbound call module) to obtain the buyer or seller's additional information for claim confirmation through outbound calls or messages. If the information is complete, the first decision agent can calibrate the first user's request information.

[0066] In the judgment information collection process, the first judgment submodule can collect the information required for judgment based on the demand information and target judgment knowledge. Optionally, the first judgment submodule can call the target language model to determine whether it has complete information required for judgment based on the target demand information and target judgment knowledge; if not, the information required for judgment is supplemented through at least one round of dialogue with the first user. Figure 6As shown, taking the e-commerce field as an example, the first judgment agent can divert the demand scenarios according to the buyer's demand information to determine the buyer's target demand scenario. According to the target demand scenario, relevant judgment handling knowledge can be recalled from the target knowledge base. The first judgment agent can call the target language model to perform integrity verification on the information required for judgment based on the target judgment knowledge. If there is missing information, the first judgment agent can call the credential agent (i.e., the credential intelligent processing module) to review the credential information, and the credential agent will resubmit the credential review information to the first judgment agent for information integrity verification. If the information required for the judgment is complete, the solution decision process can be entered.

[0067] In the solution decision process, the first judgment submodule can call the target language model to determine the judgment solution and the confidence level of the judgment solution based on the information required for the judgment. If the confidence level of the judgment solution is lower than the set threshold, the complaint ticket will be submitted to the manual customer service for processing. In the solution submission process, if the confidence level of the judgment solution is higher than or equal to the set threshold, the first judgment submodule can use the judgment solution as the target judgment solution. Figure 6 As shown, the first judgment agent (i.e., the first judgment submodule) can call the target language model to decide the judgment plan based on the judgment information. If the confidence of the judgment plan is higher than or equal to the set threshold, the first judgment agent can call the risk control agent (risk control module) to perform risk control verification on the judgment plan. If the judgment plan passes the risk control verification, the judgment plan is submitted and executed. If the confidence of the judgment plan is lower than the set threshold, the judgment process of the complaint work order can be diverted to the manual customer service, and the manually submitted plan can be obtained. If the confidence of the manually submitted plan is higher than or equal to the set threshold, the judgment plan can be subjected to risk control verification. As Figure 6 As shown, if the solution fails to pass the risk control verification, the first decision agent can obtain the risk control interception information and re-call the target language model based on the risk control interception information to decide on the decision solution.

[0068] In some optional embodiments, the main intelligent processing module can schedule the second decision submodule corresponding to the manual auxiliary decision link to identify the exception type according to the request context information and submit the exception information corresponding to the exception type to the manual customer service to assist the manual customer service in processing the complaint ticket. Figure 7 As shown, the second decision agent (ie, the second decision submodule) is mainly used for executing the process information summary flow and the manual assisted execution flow.

[0069] Optionally, in the process information summary process, the second judgment submodule is mainly used to: obtain the negotiation process information and judgment process information of the first user and the second user based on the request context information, and the judgment process information may be relevant information for the first judgment submodule to judge the service request. The second judgment submodule can summarize the negotiation process information and the judgment process information to obtain the event summary information of the service request. The second judgment submodule can identify the judgment exception type based on the event summary information, and the judgment exception type includes: judgment blocking type or system exception type. In the manual assisted execution process, for the judgment blocking type, the second judgment submodule can identify the judgment card point information, and submit the judgment card point information and the complaint work order to the manual customer service for processing. If it is a system exception type, the second judgment submodule can identify the document exception information in the complaint work order, correct the document exception information in the complaint work order, and submit the corrected document and the complaint work order to the manual customer service for processing. Further optionally, the second judgment submodule can also obtain the processing solution provided by the manual customer service for the complaint ticket based on the judgment card point information, and can mark the processing solution as a card point solution. The card point solution and the judgment card point information can be used as training data to optimize the first judgment submodule.

[0070] like Figure 7 As shown, in the process information summary process, the second judgment agent can obtain the short-term memory of a single complaint case, including negotiation process information and judgment process information. The second judgment agent can summarize the short-term memory, obtain event summary information, and perform scene diversion (i.e., identify the abnormal type) based on the event summary information. If it is a judgment blocking type, the second judgment agent can identify the judgment card point information, hand it over to the manual customer service to produce a solution, and mark the card point solution. If the system is abnormal, the second judgment agent can identify the document abnormal information, and perform self-service diagnosis based on the document abnormal information to recommend available correction tools. After determining the correction tool, the second judgment agent can call the correction tool to correct the document abnormal information, and jump to the judgment blocking branch based on the corrected document. Optionally, as Figure 7 As shown, after the accuracy of the correction tool for correcting the document anomaly information exceeds a certain threshold, the document corrected by the correction tool can be deposited into the first decision agent. In some embodiments, the first decision agent can use the corrected document to re-execute the intelligent decision process.

[0071] In the above-mentioned embodiments of the present application, the main intelligent processing module and the sub-intelligent processing module can combine the semantic understanding, prediction, and logical judgment capabilities of the language model to efficiently respond to user service requests in different scenarios, especially user complaints in complex scenarios. For example, in complex complaint scenarios (such as those involving multi-party negotiation or evidence analysis), the intelligent agent can automatically generate solutions and promote their execution through multi-round dialogue, sentiment analysis, evidence extraction, and other technologies, greatly reducing reliance on manual customer service.

[0072] It should be noted that the execution entity of each step of the method provided in the above embodiment can be the same device, or the method can be executed by different devices. For example, the execution entity of steps 101 to 104 can be device A; for another example, the execution entity of steps 101 and 102 can be device A, and the execution entity of step 103 can be device B; and so on.

[0073] In addition, some of the processes described in the above embodiments and the accompanying drawings include multiple operations that appear in a specific order, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0074] It should be noted that the first user information (including but not limited to the first user's device information, the first user's personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the first user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for the first user to choose to authorize or refuse.

[0075] Figure 8 The structure diagram of an electronic device provided by an exemplary embodiment of the present application is shown, and the electronic device is applicable to the request processing method provided by the above embodiment. Figure 8 As shown, the electronic device includes: a memory 801 , a processor 802 and a communication component 803 .

[0076] Memory 801 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method operating on the electronic device, data structures, contact data, phone book data, messages, images, videos, etc. In this embodiment, the computer programs stored in memory 801 may include programs related to multiple intelligent processing modules in the request processing system, including: a main intelligent processing module and multiple sub-intelligent processing modules that cooperate with the main intelligent processing module.

[0077] Processor 802 is coupled to memory 801 and is used to execute a computer program related to the main intelligent processing module in memory 801, so as to: respond to a service request of a first user and obtain the request context information of the first user; plan a processing plan for the service request based on the request context information to obtain a target processing plan, wherein the target processing plan includes at least one processing link; schedule the sub-intelligent processing modules corresponding to each of the at least one processing link to process the service request based on the request context information; among the at least one processing link, the sub-intelligent processing modules corresponding to at least some of the processing links are used to call the target model to process the service request based on the request context information.

[0078] Optionally, when the processor 802 plans the processing plan for the service request based on the request context information and obtains the target processing plan, it is specifically used to: search the target knowledge base based on the request context information to obtain the target knowledge information required to process the service request; perform feature extraction on the request context information to obtain the target request feature corresponding to the request context information; plan the processing plan for the service request based on the target knowledge information and the target request feature to obtain the target processing plan.

[0079] Optionally, when the processor 802 performs feature extraction on the request context information and obtains the target request feature corresponding to the request context information, it is specifically used to: perform feature extraction on the request context information from at least one dimension to obtain the target request feature, and the at least one dimension includes: at least one of: order dimension, membership dimension and logistics dimension.

[0080] Optionally, when the processor 802 plans the processing plan for the service request based on the target knowledge information and the target request characteristics to obtain the target processing plan, it is specifically used to: construct a target prompt word based on the target knowledge information and the target request characteristics; and call the target language model based on the target prompt word to plan the processing plan for the service request to obtain the target processing plan.

[0081] Optionally, when the processor 802 plans the processing plan for the service request based on the target knowledge information and the target request characteristics and obtains the target processing plan, it is specifically used to: determine a request processing workflow containing multiple process nodes, and the jump relationship between some process nodes in the request processing workflow is triggered by specific jump conditions; match the process nodes and jump conditions in the request processing workflow based on the target knowledge information and the target request characteristics to obtain a target work link adapted to the service request; and determine the target processing plan based on the target work link.

[0082] Optionally, the request context information includes complaint information of the first user against the second user; when the processor 802 schedules the sub-intelligent processing modules corresponding to each of the at least one processing links to process the service request according to the request context information, it is specifically used to: schedule the initiation sub-module corresponding to the complaint initiation link to determine whether the service request meets the complaint access requirements based on the request context information, and, when it meets the complaint access requirements, create a complaint work order corresponding to the service request; schedule the negotiation sub-module corresponding to the negotiation link to determine whether the service request meets the negotiation access requirements based on the request context information, and, when it meets the negotiation access requirements, determine the target negotiation plan by conducting at least one round of dialogue with the first user and / or the second user; schedule the first judgment sub-module corresponding to the intelligent judgment link to obtain the information required for judgment based on the request context information, and determine the target judgment plan based on the information required for judgment; schedule the second judgment sub-module corresponding to the manual assisted judgment link to identify the exception type based on the request context information, and submit the exception information corresponding to the exception type to the manual customer service to assist the manual customer service in processing the complaint work order.

[0083] Optionally, when the processor 802 schedules the initiating submodule corresponding to the complaint initiation link and determines whether the service request meets the complaint access requirements based on the request context information, it is specifically used to: schedule the initiating submodule to obtain the complaint form features and complaint access features corresponding to the service request based on the request context information; search the target knowledge base for the complaint form collection rules and complaint access rules corresponding to the service request based on the request context information; if the complaint form features meet the requirements of the form collection rules and the complaint access features meet the requirements of the complaint access rules, then create the complaint work order.

[0084] Optionally, if the complaint form features do not meet the requirements of the form collection rules, the processor 802 determines the target form information to be collected; performs at least one round of dialogue with the first user to collect the target form information; and, after collecting the target form information, re-executes the step of obtaining the complaint form features.

[0085] Optionally, when the processor 802 determines whether the service request meets the negotiation access requirements based on the request context information when scheduling the negotiation sub-module corresponding to the negotiation link, it is specifically used to: schedule the negotiation sub-module, obtain real-time conversation information between the first user and the second user and / or obtain historical negotiation information between the first user and the second user based on the request context information; call the target language model based on the real-time conversation information and / or historical negotiation information, so that the target language model determines whether the service request meets the negotiation access requirements.

[0086] Optionally, when determining a target negotiation scheme based on at least one round of conversation with the first user and / or the second user, the processor 802 is specifically configured to: identify the negotiation intention of the second user through at least one round of conversation with the second user; if the second user agrees to negotiate, identify the target demand information of the first user through at least one round of conversation with the first user; determine at least one recommended negotiation scheme based on the target demand information of the first user; provide the at least one recommended negotiation scheme to the second user so that the second user selects a target negotiation scheme from the at least one recommended negotiation scheme; conduct at least one round of conversation with the first user based on the target negotiation scheme to identify the negotiation intention of the first user; if the first user agrees to negotiate, determine that the target negotiation scheme is a consensus scheme.

[0087] Optionally, when the processor 802 schedules the first judgment submodule corresponding to the intelligent judgment link, obtains the information required for judgment according to the request context information, and determines the target judgment scheme based on the information required for judgment, it is specifically used to: schedule the first judgment submodule to obtain the target demand information of the first user based on the real-time conversation information between the first user and the second user and the request context information; collect the information required for judgment based on the target demand information and the target judgment knowledge; call the target language model to determine the judgment scheme and the confidence of the judgment scheme based on the information required for the judgment; if the confidence of the judgment scheme is lower than the set threshold, submit the complaint ticket to manual customer service for processing; if the confidence of the judgment scheme is greater than or equal to the set threshold, use the judgment scheme as the target judgment scheme.

[0088] Optionally, when the processor 802 obtains the target demand information of the first user based on the request context information, it is specifically used to: call the target language model to determine whether the target demand information of the first user can be identified based on the real-time conversation information between the first user and the second user and the request context information; if not, obtain the supplementary demand information of the first user by calling or leaving a message to the first user; if so, determine the target demand information of the first user based on the real-time conversation information, the request context information and the supplementary demand information.

[0089] Optionally, when the processor 802 obtains the information required for the judgment based on the demand information and the target judgment knowledge, it is specifically used to: call the target language model to determine whether it has complete information required for the judgment based on the demand information and the target judgment knowledge; if not, supplement the information required for the judgment through at least one round of dialogue with the first user.

[0090] Optionally, when the processor 802 schedules the second judgment submodule corresponding to the manual assisted judgment link, according to the request context information, it is specifically used to: schedule the second judgment submodule to obtain the negotiation process information and judgment process information between the first user and the second user according to the request context information; summarize the negotiation process information and the judgment process information to obtain the event summary information of the service request; identify the judgment exception type according to the event summary information, and the judgment exception type includes: judgment blocking type or system exception type; if it is the judgment blocking type, identify the judgment card point information, and submit the judgment card point information and the complaint work order to the manual customer service for processing; if it is the system exception type, correct the document exception information in the complaint work order, and submit the corrected document and the complaint work order to the manual customer service for processing.

[0091] Optionally, the processor 802 is further configured to: obtain a processing solution provided by the manual customer service for the complaint work order based on the judgment card point information; and mark the processing solution as a card point solution.

[0092] Optionally, when the processor 802 responds to the service request of the first user and obtains the request context information of the first user, it is specifically used to: respond to the service request of the first user, conduct at least one round of dialogue with the first user to obtain the request context information based on the at least one round of dialogue; or, obtain the request context information of the first user from at least one complaint channel based on the identification information of the first user.

[0093] Further, if Figure 8As shown, the electronic device also includes other components such as a power supply component 804 , a display component 805 , and an audio component 806 . Figure 8 Only some components are shown schematically, which does not mean that the electronic device only includes Figure 8 Components shown. Figure 8 Components in the dotted box are optional components, not mandatory components, and may depend on the product form of the electronic device. The electronic device of this embodiment can be implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone or an IOT device, or a server device such as a conventional server, a cloud server or a server array. If the electronic device of this embodiment is implemented as a terminal device such as a desktop computer, a laptop computer, a smart phone, etc., it may include Figure 8 If the electronic device of this embodiment is implemented as a conventional server, cloud server or server array and other server-side devices, it may not include Figure 8 Components within the dotted box.

[0094] Among them, the memory 801 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0095] Among them, the communication component 803 is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as 2G (such as Global System for Mobile Communications (GSM)), 3G (such as Wideband Code Division Multiple Access (WCDMA), 4G (such as Long Term Evolution (LTE)), 4G+ (such as upgraded version of Long Term Evolution (LTE-Advanced, LTE-A)), or 5G (fifth generation mobile communication technology), or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel.

[0096] The power supply component 804 is used to provide power to various components of the device where the power supply component is located. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply component is located.

[0097] The display assembly includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from a first user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor can not only sense the boundaries of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

[0098] The audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC), and when the device where the audio component is located is in an operating mode, such as call mode, recording mode, and voice recognition mode, the microphone is configured to receive external audio signals. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0099] In this embodiment, when a user makes a service request, the main intelligent processing module can obtain the user's request context information and automatically plan a target processing solution for processing the service request based on the request context information. The main intelligent processing module can schedule the sub-intelligent processing modules corresponding to at least one processing link in the target processing solution to process the service request based on the request context information. Based on this embodiment, the main intelligent processing module can flexibly schedule the sub-intelligent processing modules based on the request context information, thereby dynamically assigning the processing tasks of the service request to the appropriate sub-intelligent processing modules, reducing dependence on manual customer service. Some sub-intelligent processing modules can use the target language model to process uncertain and ambiguous information, and can make reasonable decisions based on incomplete or uncertain data. The main intelligent processing module and the sub-intelligent processing modules work together to efficiently respond to user service requests in different scenarios. Furthermore, this is conducive to increasing the proportion of automated processing of user complaints in the service platform and significantly reducing the labor costs brought by manual customer service.

[0100] Accordingly, an embodiment of the present application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to implement the steps in the above method embodiment. The computer-readable storage medium includes volatile or non-volatile or a combination thereof, and may be removable or non-removable. Examples of computer-readable storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, magnetic cassette, tape disk storage or other magnetic storage device or any other non-transmission medium.

[0101] Accordingly, an embodiment of the present application further provides a computer program product, which includes a computer program or instructions, and when the computer program or instructions are executed by a processor, the processor is enabled to implement the steps in the above-mentioned method embodiment. It should be understood that each process or a combination of multiple processes in the above-mentioned method flow can be implemented by a computer program or instruction. In addition, these computer programs or instructions can be applied to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device, so that the processor of the general-purpose computer, the special-purpose computer, the embedded processor or other programmable data processing device can be implemented as a device for implementing the corresponding functions in the above-mentioned method embodiment.

[0102] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, product, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, product, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not preclude the presence of additional identical elements in the process, method, product, or apparatus that includes the element.

[0103] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A request processing method, characterized in that: A main intelligent processing module is applied to a request processing system, wherein the request processing system further comprises a plurality of sub-intelligent processing modules cooperating with the main intelligent processing module, and the method comprises: Responding to a service request from a first user, and obtaining request context information of the first user; Planning a processing solution for the service request according to the request context information to obtain a target processing solution, wherein the target processing solution includes at least one processing step; Schedule the sub-intelligent processing modules corresponding to each of the at least one processing link to process the service request according to the request context information; in the at least one processing link, the sub-intelligent processing modules corresponding to at least some of the processing links are used to call the target model to process the service request according to the request context information.

2. The method according to claim 1, characterized in that Planning a processing solution for the service request based on the request context information to obtain a target processing solution includes: Searching the target knowledge base according to the request context information to obtain target knowledge information required to process the service request; Extracting features from the request context information to obtain target request features corresponding to the request context information; A processing solution for the service request is planned according to the target knowledge information and the target request characteristics to obtain a target processing solution.

3. The method according to claim 2, characterized in that Performing feature extraction on the request context information to obtain target request features corresponding to the request context information includes: Feature extraction is performed on the request context information from at least one dimension to obtain the target request feature, where the at least one dimension includes at least one of an order dimension, a membership dimension, and a logistics dimension.

4. The method according to claim 2, characterized in that Planning a processing solution for the service request based on the target knowledge information and the target request characteristics to obtain a target processing solution, including: constructing a target prompt word according to the target knowledge information and the target request feature; According to the target prompt word, a target language model is called to plan a processing solution for the service request to obtain the target processing solution.

5. The method according to claim 2, characterized in that Planning a processing solution for the service request based on the target knowledge information and the target request characteristics to obtain a target processing solution, including: Determining a request processing workflow comprising a plurality of process nodes, wherein jump relationships between some of the process nodes in the request processing workflow are triggered by specific jump conditions; Matching process nodes and jump conditions in the request processing workflow according to the target knowledge information and the target request characteristics to obtain a target work link adapted to the service request; The target processing solution is determined according to the target working link.

6. The method according to claim 1, characterized in that The request context information includes complaint information of the first user against the second user; Scheduling the sub-intelligent processing modules corresponding to the at least one processing link to process the service request according to the request context information includes: The initiation submodule corresponding to the scheduling complaint initiation link determines whether the service request meets the complaint access requirements based on the request context information, and creates a complaint work order corresponding to the service request when the complaint access requirements are met; a negotiation submodule corresponding to the scheduling negotiation phase, determining whether the service request meets the negotiation access requirements based on the request context information, and, if the service request meets the negotiation access requirements, determining a target negotiation solution by conducting at least one round of dialogue with the first user and / or the second user; The first decision submodule corresponding to the scheduling intelligent decision link obtains information required for decision according to the request context information, and determines a target decision solution based on the information required for decision; The second judgment submodule corresponding to the manual assisted judgment link is dispatched to identify the exception type according to the request context information, and submit the exception information corresponding to the exception type to the manual customer service to assist the manual customer service in processing the complaint work order.

7. The method according to claim 6, characterized in that The initiation submodule corresponding to the scheduling complaint initiation link determines whether the service request meets the complaint access requirements based on the request context information, including: Scheduling the initiating submodule to obtain complaint form features and complaint access features corresponding to the service request according to the request context information; According to the request context information, searching the target knowledge base for complaint form collection rules and complaint access rules corresponding to the service request; If the complaint form features meet the requirements of the form collection rules, and the complaint access features meet the requirements of the complaint access rules, then the complaint work order is created.

8. The method according to claim 7, characterized in that Also includes: If the complaint form features do not meet the requirements of the form collection rules, determining the target form information to be collected; Perform at least one round of dialogue with the first user to collect the target form information; as well as, After collecting the target form information, re-execute the step of obtaining the complaint form features.

9. The method according to claim 6, characterized in that The negotiation submodule corresponding to the scheduling negotiation link determines whether the service request meets the negotiation access requirements based on the request context information, including: Scheduling the negotiation submodule to obtain real-time conversation information between the first user and the second user and / or obtaining historical negotiation information between the first user and the second user based on the request context information; The target language model is called according to the real-time dialogue information and / or historical negotiation information, so that the target language model determines whether the service request meets the negotiation admission requirement.

10. The method according to claim 7, characterized in that Determining a target negotiation plan based on at least one round of dialogue with the first user and / or the second user includes: identifying the negotiation intention of the second user through at least one round of dialogue with the second user; If the second user agrees to negotiate, identifying the target demand information of the first user by conducting at least one round of dialogue with the first user; Determining at least one recommended negotiation solution based on the target demand information of the first user; providing the at least one recommended negotiation solution to the second user, so that the second user selects a target negotiation solution from the at least one recommended negotiation solution; Conducting at least one round of dialogue with the first user according to the target negotiation plan to identify the negotiation intention of the first user; If the first user agrees to the negotiation, the target negotiation solution is determined to be a consensus solution.

11. The method according to claim 6, characterized in that The first decision submodule corresponding to the intelligent decision link is dispatched to obtain information required for decision according to the request context information, and determine a target decision solution based on the information required for decision, including: Scheduling the first judgment submodule to obtain target demand information of the first user based on the real-time conversation information between the first user and the second user and the request context information; Collect information required for judgment based on the target demand information and target judgment knowledge; Based on the information required for the judgment, calling the target language model to determine a judgment scheme and a confidence level of the judgment scheme; If the confidence level of the decision is lower than the set threshold, the complaint ticket will be submitted to manual customer service for processing; If the confidence of the decision scheme is greater than or equal to the set threshold, the decision scheme is used as the target decision scheme.

12. The method according to claim 11, characterized in that Acquiring target demand information of the first user according to the request context information includes: Based on the real-time conversation information between the first user and the second user and the request context information, calling the target language model to determine whether the target request information of the first user can be recognized; If not, obtaining the first user's additional request information by calling or leaving a message to the first user; If so, the target demand information of the first user is determined according to the real-time conversation information, the request context information and the supplemented demand information.

13. The method according to claim 11, characterized in that Based on the appeal information and target judgment knowledge, obtain the information required for judgment, including: Based on the appeal information and target judgment knowledge, calling the target language model to determine whether it has complete information required for judgment; If not, the information required for the decision is supplemented through at least one round of dialogue with the first user.

14. The method according to claim 6, characterized in that Scheduling the second decision submodule corresponding to the manual assisted decision link, based on the request context information, includes: Scheduling the second decision submodule to obtain negotiation process information and decision process information between the first user and the second user according to the request context information; Summarizing the negotiation process information and the decision process information to obtain event summary information of the service request; Identify, according to the event summary information, a judgment abnormality type, wherein the judgment abnormality type includes: a judgment blocking type or a system abnormality type; If the judgment is a blocking type, identifying the judgment card point information, and submitting the judgment card point information and the complaint work ticket to the manual customer service for processing; If it is the system exception type, the document exception information in the complaint work order is corrected, and the corrected document and the complaint work order are submitted to the manual customer service for processing.

15. The method according to claim 14, characterized in that Also includes: Obtaining the processing solution provided by the manual customer service for the complaint ticket based on the judgment checkpoint information; The processing solution is marked as a card point solution.

16. The method according to any one of claims 1 to 15, characterized in that Responding to the service request of the first user and obtaining the request context information of the first user includes: responding to a service request from a first user, conducting at least one round of dialogue with the first user to obtain the request context information according to the at least one round of dialogue; or According to the identification information of the first user, request context information of the first user is obtained from at least one complaint channel.

17. An electronic device, characterized in that: include: memory and processor; The memory is used to store one or more computer instructions; The processor is configured to execute the one or more computer instructions to perform the steps of the method according to any one of claims 1 to 8.

18. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 can be implemented.

19. A computer program product, characterized in that include: A computer program / instruction, which, when executed by a processor, can implement the steps of the method according to any one of claims 1 to 8.

Images