Multi-platform intelligent customer service session management method and device, medium and program product

Abstract

The invention relates to the field of AI platforms, and provides a multi-platform intelligent customer service session management method and device, a medium and a program product.The method comprises the steps that sub-accounts of a target merchant on different e-commerce platforms are aggregated for the target merchant on a session management platform, and intelligent customer service matched with platform characteristics is configured for each sub-account; the target merchant can intuitively and conveniently receive and send messages on a single platform, and the session efficiency is improved; the target merchant can switch to the inquiry of the e-commerce platform received by the sub-account icon by randomly clicking the sub-account icon, and the customer service of the target merchant does not need to switch and log in customer service accounts of a plurality of e-commerce platforms back and forth, so that the operation can be reduced, the customer service accounts of different e-commerce platforms can be concentrated on a single platform, and the user experience is improved. One manual customer service or intelligent customer service can receive inquiry of buyers of multiple e-commerce platforms on a single platform, so that the operation cost is reduced, and the session efficiency and the buyer experience are improved.

Description

Technical Field

[0001] The present application relates to the field of AI platform technology, and in particular to a multi-platform intelligent customer service session management method, device, medium and program product. Background Art

[0002] In the e-commerce retail sector, merchants operate across multiple platforms, including Taobao, Douyin, and Pinduoduo. Customer service representatives must access customer service tools for each platform. While purchasing intelligent robots for each platform can improve customer service efficiency, the workload of repeatedly entering the same product information leads to low utilization of these robots. Furthermore, a single customer service representative must simultaneously log in to the intelligent customer service systems of multiple platforms, switching back and forth between them. This involves frequent repetitive operations and results in low conversation efficiency. Summary of the Invention

[0003] This application provides a multi-platform intelligent customer service conversation management method, device, medium and program product, which can centrally receive inquiries from buyers of the same merchant on different e-commerce platforms, improve conversation efficiency, and facilitate cross-platform message management.

[0004] In a first aspect, the present application provides a multi-platform intelligent customer service session management method from the perspective of a session management platform. The method is applied to the session management platform, where a target merchant uses multiple intelligent customer service representatives on the session management platform to receive inquiries from buyers on multiple e-commerce platforms. The method includes:

[0005] In response to a login request from a target merchant on a management interface, a sub-account of the target merchant on multiple e-commerce platforms is selected for login; wherein the management interface includes a first area, a second area, and a third area. The first area displays icons of each sub-account in real time according to session update time, and each sub-account is associated with at least one intelligent customer service representative. The second area displays the session interface of the selected first sub-account, customer service performance evaluation parameters, order status, and excellent sales pitches for the product being inquired about. The third area displays dynamic product data of the product being inquired about across all e-commerce platforms, including transaction status distribution, popularity data, distribution and inventory data, and visualized product features.

[0006] Receive a first input for the first sub-account icon of the first e-commerce platform in the first area, display a first conversation interface in the second area, conduct a conversation with the first buyer according to the platform characteristics of the first e-commerce platform, and synchronously display the first dynamic product data of the first product currently inquired by the first buyer in the third area; receive a second input for the second sub-account icon of the second e-commerce platform, switch from the first conversation interface to the second conversation interface, conduct a conversation with the second buyer according to the platform characteristics of the first e-commerce platform, and update the visual content of the first area and the second area accordingly.

[0007] In a second aspect, the present application provides a session management platform having a function of implementing the multi-platform intelligent customer service session management method provided in the first aspect. The function can be implemented by hardware, or by hardware executing corresponding software. The hardware or software includes one or more modules corresponding to the above functions, and the modules can be software and / or hardware. In some embodiments, the session management platform includes an input / output module, a processing module, and a display module, the processing module controls the input / output operation of the input / output module, and controls the display operation of the display module:

[0008] The processing module is configured to select the sub-accounts of the target merchant on the plurality of e-commerce platforms to log in in response to a login request of the target merchant on the management interface; wherein the management interface includes a first area, a second area and a third area, the first area displays each sub-account icon in real time according to the session update time, and each sub-account is associated with at least one intelligent customer service; the second area displays the session interface of the selected first sub-account, and the third area displays the dynamic commodity data of the consulted commodity on the whole e-commerce platform;

[0009] After receiving a first input through the input / output module for the first sub-account icon of the first e-commerce platform in the first area, the display module is controlled to display the first session interface in the second area, to have a conversation with the first buyer according to the platform characteristics of the first e-commerce platform, and to synchronously display the first dynamic commodity data of the first commodity currently consulted by the first buyer in the third area;

[0010] The input / output module receives a second input for the second sub-account icon of the second e-commerce platform, and the display module is controlled to switch from the first session interface to the second session interface, to have a conversation with the second buyer according to the platform characteristics of the first e-commerce platform, and the first area and the second area are updated correspondingly.

[0011] In a third aspect, the present application provides a session management platform, comprising: at least one processor and a memory; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the steps in any of the multi-platform intelligent customer service session management methods provided in the first aspect and any of the embodiments of the first aspect.

[0012] In a fourth aspect, the present application provides a computer-readable storage medium having the functionality to implement the multi-platform intelligent customer service session management method provided in the first aspect above. The functionality can be implemented via hardware, or by hardware executing corresponding software implementations. The hardware or software includes one or more modules corresponding to the functionality described above, and the modules can be software and / or hardware. Specifically, the computer-readable storage medium stores a plurality of instructions, which are suitable for loading by a processor to execute the steps of any of the multi-platform intelligent customer service session management methods provided in the first aspect of the present application.

[0013] Compared with the existing technology, in the solution provided by this application, since the sub-accounts of the target merchant on different e-commerce platforms are aggregated in advance on the session management platform, and intelligent customer service matching the platform characteristics is configured for each sub-account, therefore, on the one hand, the target merchant can send and receive messages intuitively and conveniently on a single platform, thereby improving session efficiency; on the other hand, the target merchant can click on the sub-account icon at will to switch to the inquiry of the e-commerce platform received by the sub-account icon, and the target merchant's customer service does not need to switch back and forth to log in to the customer service accounts of multiple e-commerce platforms, which can reduce operations and can also concentrate the customer service accounts of different e-commerce platforms on a single platform. One manual customer service or intelligent customer service can receive inquiries from buyers of multiple e-commerce platforms on a single platform, thereby reducing operating costs and improving session efficiency and buyer experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of a session management platform provided in this application;

[0015] Figure 2 A flowchart of the intelligent customer service session configuration method in this application;

[0016] Figure 3 This is a schematic diagram of configuring intelligent customer service in this application;

[0017] Figure 4 This is a schematic diagram of the blind spot analysis based on the knowledge base hit rate heat map in this application;

[0018] Figure 5 This is a flow chart of the multi-platform intelligent customer service session management method in this application;

[0019] Figure 6a This is a schematic diagram of an interface for aggregating multi-platform conversation processing in this application;

[0020] Figure 6b This is a schematic diagram of an interface for aggregating multi-platform conversation processing in this application;

[0021] Figure 7 This is another flowchart of the multi-platform intelligent customer service session management method in this application;

[0022] Figure 8 A flowchart of a process for a conversation with a second buyer in the present application;

[0023] Figure 9 A structural diagram of a conversation management platform in the present application;

[0024] Figure 10 A structural diagram of an entity device for implementing a multi-platform intelligent customer service conversation management method in the present application. DETAILED DESCRIPTION

[0025] The terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish similar objects (for example, the first buyer and the second buyer in the present application represent buyers from different e-commerce platforms, respectively), and do not necessarily have to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or modules does not have to be limited to those steps or modules clearly listed, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The division of modules in the present application is only a logical division, and there can be another division in actual application.

[0026] The present application provides a multi-platform intelligent customer service conversation management method, device, medium and computer program, which can be used in a server or a terminal device, and specifically can be used in the management and optimization of reception services of different e-commerce platforms by merchants in the e-commerce field. In some embodiments, the present application is applied to a conversation management platform as shown in Figure 1 The reception end is used to receive the message transmission of the reception sub-account of different e-commerce platforms. The management end is used to configure each item of the intelligent customer service. Figure 1 In the present application, a plurality of intelligent customer services managed by a target merchant are included, and the intelligent customer services are connected to the sub-accounts of e-commerce platforms such as PinDuoDuo, QianNiu, and DouDian.

[0027] The present application provides a solution related to artificial intelligence (AI), nature language processing (NLP), machine learning (ML), and the like.

[0028] The following will be described in combination with Figure 2-Figure 8 The technical solutions of the embodiments of the present application are exemplified.

[0029] The session management platform aggregates buyer inquiries from across platforms for session aggregation and centralized session management. It includes a platform feature recognition engine, a spatiotemporal knowledge modeling engine compatible with multiple e-commerce platforms, and an edge computing node cluster.

[0030] The edge computing node cluster includes multiple edge computing nodes, each covering at least one type of e-commerce platform. The edge computing node cluster is configured to receive and process buyer inquiries from different e-commerce platforms. The central computing nodes in multiple edge computing centers can dynamically balance load using pre-set traffic diversion strategies.

[0031] The platform feature recognition engine is configured to recognize the platform features of the e-commerce platform to which the buyer's inquiry belongs, that is, when the intelligent customer service of the sub-account receives a buyer from a certain e-commerce platform, the platform features of the e-commerce platform can be recognized.

[0032] The spatiotemporal knowledge modeling engine is configured to federate the dynamic knowledge bases of multiple e-commerce platforms. Based on the platform feature recognition engine's federated query query, the dynamic knowledge base retrieves platform-specific knowledge that is spatiotemporally correlated with buyer inquiries, serving as a federated query center. This spatiotemporal knowledge modeling engine serves as the central execution unit for buyer interactions within the customer service system, adapting cross-platform knowledge. It collaborates with the intelligent customer service system and the federated dynamic knowledge base to achieve conversation aggregation and differentiated responses. The platform feature recognition engine and spatiotemporal knowledge modeling engine are the engines supporting the intelligent customer service system. When utilizing the federated node knowledge bases of multiple e-commerce platforms, the spatiotemporal knowledge modeling engine simultaneously provides knowledge federation aggregation and spatiotemporal pre-rehearsal capabilities.

[0033] The collaboration process among the three can be referred to Figure 2 , comprising the following steps:

[0034] 101. The edge computing node cluster receives data streams of buyer inquiries from different e-commerce platforms and forwards the data streams to the platform feature recognition engine according to a preset diversion strategy.

[0035] The preset diversion strategy is used to route messages for buyer inquiries received from different e-commerce platforms, and the preset diversion strategy includes a platform-level diversion strategy and a merchant-level reception diversion strategy.

[0036] 102. The platform feature recognition engine parses the data stream from the edge computing node cluster, identifies the platform features of the e-commerce platform to which the buyer's inquiry belongs from the data stream, and sends the platform features to the spatiotemporal knowledge modeling engine.

[0037] 103. The spatio-temporal knowledge modeling engine obtains platformized knowledge that has spatio-temporal correlation with the buyer inquiry based on the platform features and the federated query dynamic knowledge base, and sends the platformized knowledge to the intelligent customer service.

[0038] The spatio-temporal knowledge modeling engine converts the common knowledge matched in the federated knowledge base into spatio-temporal knowledge in real time, and converts the common knowledge into platform-specific knowledge that conforms to specific platform features and specific spatio-temporal scenarios. The platform-specific knowledge includes commodity activity time (such as Taobao pre-sale vs. Jingdong flash sale) and regional policy (such as Shanghai tax-free warehouse vs. Xinjiang free shipping area).

[0039] 104. The intelligent customer service generates a style answer that adapts to the platform features according to a preset differentiated response strategy and the platform features of the e-commerce platform to which the buyer inquiry belongs, and returns the style answer to the buyer through the edge computing node cluster.

[0040] The platform-specific knowledge is displayed in the conversation interface of the corresponding buyer after generating a differentiated style answer that conforms to the platform features.

[0041] In some embodiments, the spatio-temporal knowledge modeling engine is further configured to: obtain knowledge matching degree data, response delay data, and conversation transfer data in real time; wherein the knowledge matching degree data is a correlation between the buyer inquiry and the platform-specific knowledge mapped to the dynamic knowledge base in the style answer returned to the buyers of different platforms; the response delay data is the time of the intelligent customer service of each sub-account responding to different buyer inquiries; and the conversation transfer data is statistical data of the intelligent customer service of the sub-account when receiving buyer inquiries of multiple e-commerce platforms being transferred from the intelligent customer service to the manual. The knowledge base hit rate heat map is generated according to the knowledge matching degree data, the cross-platform response delay distribution graph is generated according to the response delay data, and the conversation transfer rate trend is generated according to the conversation transfer data. The first monitoring mechanism is constructed according to the knowledge base hit rate heat map, the cross-platform response delay distribution graph, and the conversation transfer data. The first monitoring mechanism includes: analyzing the blind area of the dynamic knowledge base based on the knowledge base hit rate heat map and identifying platform differences, capturing the service delay trend of the intelligent customer service associated with the target platform in real time based on the cross-platform response delay distribution graph, and dynamically balancing the cross-platform load according to the service delay trend, and analyzing the conversation transfer reason based on the conversation transfer rate trend and triggering a preset linkage event.

[0042] As can be seen, in the present application, the spatio-temporal knowledge modeling engine converts the common knowledge matched in the federated knowledge base into spatio-temporal knowledge in real time, and converts the common knowledge into platform-specific knowledge that conforms to specific platform features and specific spatio-temporal scenarios, so that the intelligent customer service can generate a native style response that conforms to the platform features, the federated query can comply with the regional policy boundary, the cross-platform consistency of the dynamic knowledge base can dynamically adapt to the promotion period, and a scenario-based interactive experience can be provided (for example, in a live streaming scenario, the spatio-temporal knowledge can be converted into a virtual anchor interactive scenario).

[0043] Figure 2 The corresponding embodiment is a platform-level configuration for the session management platform. Based on the above platform-level configuration, in order to accommodate the reception needs of multiple merchants, personalized configuration can be performed for each merchant. For example, the intelligent customer service of multiple sub-accounts of the e-commerce platform can be configured for the target merchant. Figure 3 As shown, the configuration method for aggregating cross-platform intelligent customer service sessions in this application includes:

[0044] 201. Create a reception project for aggregating multiple sub-accounts of target merchants on different e-commerce platforms.

[0045] In this reception project, the sub-accounts of the target merchant on different e-commerce platforms are bound to the management account. The sub-accounts are the reception accounts for the target merchant to open stores on various e-commerce platforms.

[0046] 202. Based on the platform characteristics of the e-commerce platform to which the sub-account belongs and the product parameters of the already distributed products, create an intelligent customer service that matches the platform characteristics and bind it to each sub-account accordingly, and associate a dynamic knowledge base of the already distributed products for the intelligent customer service of each sub-account.

[0047] Each intelligent customer service representative's capabilities may vary due to differences in their associated dynamic knowledge bases or the model structures they utilize. Some may be general-purpose, some specialized, and some may offer higher quality services for certain types of inquiries. This is not a limitation. Platform characteristics can be identified by the platform feature recognition engine when a sub-account's intelligent customer service representative receives buyers from that platform. This can be done, for example, by analyzing the native messages in the buyer's inquiries, analyzing short videos on Doudian, analyzing Taobao order associations, or analyzing Pinduoduo's group buying logic.

[0048] The intelligent customer service of each sub-account is respectively associated with a visual execution logic for responding to buyer inquiries and at least one dynamic knowledge base for the visual execution logic to call.

[0049] In this application, the dynamic knowledge base used by the intelligent customer service configured for each merchant's sub-account can be a general knowledge base (which can be pre-built by the session management platform or imported by the merchant itself) or a federated node knowledge base, without limitation. That is, the dynamic knowledge base can come from multiple e-commerce platforms, and each dynamic knowledge base can be a federated node. Alternatively, the dynamic knowledge base can be tailored for each merchant from a cross-platform distribution platform by this session management platform, and the management of various types of knowledge in the dynamic knowledge base can be entered and updated by the merchant himself, and this application does not limit this. Figure 3 As shown, Figure 3Fig. 1 shows the local situation of this session management platform, which collects various product parameters and business parameters from various e-commerce platforms, builds a reusable sub-dynamic knowledge base, and sets each sub-dynamic knowledge base as a federated node to facilitate federated queries when responding to buyer inquiries. Figure 3 Fig. 2 shows a scenario of performing federated queries based on the dynamic knowledge bases of multiple e-commerce platforms. The dynamic knowledge bases related to various product parameters and business parameters of each e-commerce platform are set as federated nodes to facilitate cross-platform federated queries. There is no need to share data between e-commerce platforms.

[0050] The dynamic knowledge base can be a federated node dynamic knowledge base. Each e-commerce platform is configured with an independent dynamic knowledge base. The dynamic knowledge base includes question templates, answers, and timestamps, and is protected by local differential privacy. A unified query language is set up for cross-platform queries. After receiving a buyer's inquiry, each e-commerce platform as a federated node generates a question-answer pair vector based on its own dynamic knowledge base. The session management platform, as the federation center, then performs confidence-weighted aggregation of the federated query results in a ciphertext space and returns the aggregated results to the buyer. Optionally, the federated dynamic knowledge base can also be incrementally updated by aligning heterogeneous product data or based on knowledge timeliness.

[0051] As can be seen, on the one hand, because the federated dynamic knowledge base is pre-configured for each intelligent customer service, knowledge sharing can be achieved in an encrypted state during real-time federated queries on buyer questions, reducing cross-platform query latency, improving privacy protection, and increasing the accuracy of retrieval of hot-selling products, while also reducing cross-platform knowledge blind spots. On the other hand, the target merchant can update the dynamic knowledge base of this session management platform to ensure that the key characteristics of the target merchant's products remain consistent across all e-commerce platforms where they are available. It is important to understand that the same product can vary across different e-commerce platforms based on platform characteristics, but the key characteristics of the product remain consistent across platforms. The intelligent customer service only outputs differentiated responses when serving buyers from different e-commerce platforms.

[0052] 203. Configure a session management interface for the reception project. The session management interface includes a first area and a second area. The first area displays icons of each sub-account in real time according to session update time, and the second area displays the session interface of the selected first sub-account.

[0053] The session management interface also includes a third area, which displays sub-account reception data (which can also be displayed in the second area). For example, this area displays the sub-account's reception data, scraped from platforms like Qianniu and Feige, primarily including the number of receptions, average response time, first response time, and satisfaction rate. When a row of data cannot fit, an expand button appears, and by default, only one row of data is displayed. Alternatively, this third area displays dynamic product data for the inquired product across all e-commerce platforms.

[0054] For reference Figure 1 The first area displays a list of buyers received by all sub-accounts that the project members can use. By clicking on an online sub-account, they can switch to the chat window of the sub-account.

[0055] 204. Configure a reception and diversion strategy for the intelligent customer service of each sub-account based on the platform characteristics and preset problem types, and configure a corresponding differentiated response strategy for the intelligent customer service of each sub-account based on the platform characteristics of the e-commerce platform.

[0056] The reception and traffic diversion strategy intelligently routes inquiries from buyers on different e-commerce platforms to each sub-account, achieving load balancing. This is because target merchants vary in their product offerings, product operation plans, and product transaction history across various e-commerce platforms. Therefore, they differentiate the reception dimensions for each sub-account on different platforms. For example, if product A is highly popular and has a large transaction volume on Platform 1, it will be assigned a higher weight in the traffic diversion strategy to support the incoming traffic.

[0057] Configure a reception diversion strategy for a project, including creating a policy name, how to divert reception sub-accounts, which project members are manual receptionists, which project members are backed by AI robots, and how to distribute traffic among multiple AI robots.

[0058] This differentiated response strategy refers to a strategy that intelligently generates responses that match buyer inquiries received by different sub-accounts based on the platform characteristics of the platform where the buyer inquiries originated. Specifically, it includes:

[0059] Platform-specific script libraries, for example, Douyin (designed short video content + live broadcast scripts), Taobao (professional customer service scripts), and Pinduoduo (group buying progress + limited-time discounts).

[0060] Interactive element adaptation strategy: automatically generate button / card messages that comply with platform specifications, and Douyin jump to the live broadcast room button.

[0061] Multi-terminal experience consistency strategy: The core information of the answers to inquiries about the same product on different platforms is consistent.

[0062] In this application, targeted merchants can customize agents with specific styles for different e-commerce platforms. The agents can adjust the tone and format of their responses based on the platform's operating style. For example, on a fashion e-commerce platform (such as Vipshop), the agent's responses can be more lively and trendy; on a professional enterprise e-commerce platform (such as JD.com), the responses can be more formal and rigorous.

[0063] The intelligent agent can also combine platform buyer profile data provided by edge computing nodes to provide personalized responses to buyers on different e-commerce platforms. By analyzing buyers' historical inquiries and purchasing behavior, the intelligent agent can understand their preferences and needs, thereby providing more precise answers and suggestions. For example, for buyers who frequently purchase sports equipment, when they inquire about related products, the intelligent agent can recommend suitable sports equipment for them and provide relevant usage tips and maintenance advice.

[0064] In some implementations, a first monitoring mechanism can also be established based on the historical response evaluation data of the intelligent customer service of each sub-account. This first monitoring mechanism includes analyzing dynamic knowledge base blind spots and identifying platform differences based on a knowledge base hit rate heat map; capturing service delay trends of the intelligent customer service associated with the target platform in real time based on a cross-platform response delay distribution map, and dynamically balancing cross-platform load based on service delay trends; and analyzing the reasons for session transfers based on session transfer rate trends and triggering preset linkage events.

[0065] Example 1: If the first knowledge base bound to the current target merchant is determined to have a blind spot based on the knowledge base hit rate heat map, then the blind spot knowledge of the target product is captured and added to the knowledge base to verify whether the hit rate has improved. (Core indicator: high-frequency question hit rate) Figure 4 shown.

[0066] Analysis process: Heat map analysis hit rate → Discover low hit blind spots → Supplement knowledge → Verify hit rate improvement. In some implementations, the following formula can be used to predict the discovery of blind spots:

[0067]

[0068] If the distance between the problem vector 𝑞 and the nearest neighbor knowledge k in the knowledge base is less than the threshold 𝜃, the domain knowledge needs to be supplemented.

[0069] Example 2: During a livestream, Doudian's intelligent customer service response latency surged. At 21:05 on May 15, 2025, monitoring revealed that Doudian's P99 latency increased from 120ms to 620ms. Using a cross-platform response latency distribution graph, we captured the service latency trends of the intelligent customer service associated with the target platform in real time and dynamically balanced the cross-platform load based on these trends, for example, triggering the expansion of the intelligent customer service nodes for Doudian sub-accounts.

[0070] It can be seen that, on the one hand, by analyzing the cross-platform response delay distribution map, dynamic cross-platform load balancing can be achieved; on the other hand, by analyzing the session transfer rate trend, service anomalies of each e-commerce platform can be quickly perceived, abnormal problems can be automatically located, and knowledge base updates and process optimization and tracking optimization measures can be triggered to form a closed loop.

[0071] In some implementations, a manual transfer prediction model is also embedded in each sub-account's intelligent customer service. This model predicts the probability of manual transfers when handling customer inquiries from the sub-account based on response latency, knowledge match, and buyer sentiment, thereby enabling early detection of risks in the intelligent customer service process. The manual transfer prediction model utilizes a multimodal fusion model. Cross-modal attention is designed within the manual transfer prediction model to address the three risk factors of response latency, knowledge match, and buyer sentiment (emphasizing the interaction between buyer sentiment and knowledge match), along with a time gating mechanism to identify n-round unresolved patterns. The model ultimately outputs the manual transfer probability.

[0072] Response delay is the time it takes for the intelligent customer service representative to respond to a buyer's inquiry (for example, 12 seconds is a timeout response).

[0073] Knowledge matching refers to the correlation between the platform-based knowledge in the dynamic knowledge base, measured by the intelligent customer service system, and the mapping of buyer inquiries and style responses. This is typically calculated using three dimensions: semantic matching (e.g., buyer inquiry intent), entity matching (e.g., product parameters), and scenario matching (e.g., after-sales policies and promotion rules). For example, knowledge matching is performed to generate a final response to a buyer's inquiry: "How can I return a scratched phone screen?"

[0074] The buyer sentiment value is obtained by analyzing the emotions in the conversation records, and is usually a negative sentiment value (such as complaints).

[0075] If the response delay is greater than a certain threshold, the knowledge match is lower than a certain threshold, and the buyer's sentiment score is lower than a certain threshold, these are all considered risk factors for triggering manual transfer. These three risk factors can be used to predict the probability of manual transfer when the intelligent customer service team receives buyer inquiries.

[0076] The linkage between these three risk factors can be derived using a sample weighting approach. For example, a weight of m is used for samples with high sentiment, n is used for samples with knowledge match <0.4, and p is used for samples with response delay >10s, where m>n>p. For example, a buyer inquires about "order paid but not shipped." The calculated response delay is 8.2s, the knowledge match is 0.65, and the buyer sentiment is [0.72, 0.68, 0.75] (text + voice + vision). The historical reception features of the intelligent customer service representative for the buyer are [0, 0, 1, 0, 0] (one of the last five manual transfers), question complexity is 0.7, conversation turns are 3, and system load is 0.68. This historical reception feature is embedded in the manual transfer prediction model. By adding cross-modal attention to the knowledge match and buyer sentiment risk factors, the final output probability is 0.87. Since 0.87 exceeds the preset threshold, the decision result is "immediately transfer to manual customer service."

[0077] It can be seen that by implanting the manual transfer prediction model, the complex interaction between response delay, knowledge matching and buyer emotions can be captured in real time during sub-account reception. It can achieve high-precision prediction while ensuring real-time response, reduce the manual transfer error rate, and improve the utilization rate of intelligent customer service.

[0078] Optionally, in some embodiments of the present application, the method further includes configuring a second monitoring mechanism and a verification mechanism for the reception project; wherein the second monitoring mechanism includes obtaining a policy monitoring component from each platform's announcement page, monitoring the product editing logs on each platform, and triggering a knowledge base update task when key policy fields change; the verification mechanism includes consistency checks on the same product's basic parameters across platforms, full release of new knowledge points after partial session verification, and prompting for manual intervention when platform rules conflict. This configuration enables the knowledge modeling engine to better adapt to cross-platform knowledge buyer interactions.

[0079] Optionally, in some embodiments of the present application, in this aggregation platform, considering the different message formats of different e-commerce platforms, in order to reduce the zero conversion processing of the message formats of different business platforms, the method further includes: configuring a protocol knowledge base and a native message format parser for each edge computing node in the edge computing node cluster, wherein the protocol knowledge base is used to store protocol metadata of the communication protocol characteristics of each e-commerce platform; and the native message format parser is used to perform semantic extraction of native messages (such as buyer inquiries) based on the protocol knowledge base. The spatiotemporal knowledge modeling engine is also configured with an AI semantic understanding layer, which is used to perform semantic understanding of the extracted semantics.

[0080] The intelligent customer service is configured to generate a style response that adapts to the platform characteristics based on the product parameters that match the native message obtained from the dynamic knowledge base received from the spatiotemporal knowledge modeling engine, and in accordance with the preset format template of the platform characteristics of the e-commerce platform to which the message belongs, that is, a response that conforms to the native format of the message source platform.

[0081] In some implementations, the protocol metadata includes:

[0082] The first type of storage structure includes message structure definition, signature rules and session mechanism, such as Taobao;

[0083] The second type of storage structure includes message structure description files and live broadcast special effects mappings, such as those used in TikTok. The message structure description files (Protocol Buffers, or protobuf) are structured data serialization protocols, and protobuf description files describe platform-specific message characteristics. Based on these protobuf description files and live broadcast special effects mappings, intelligent customer service can retrieve special effects parameters, invoke special effects tools to render dynamic effects, and synthesize message cards to display to buyers, enabling multi-scenario reuse.

[0084] The third type of storage structure includes group buying rules and real-time activity strategies, such as the Pinduoduo platform.

[0085] It can be seen that by setting up corresponding differentiated message storage structures for different e-commerce platforms according to platform characteristics in this session management platform, zero conversion processing of message formats for different business platforms can be achieved, and cross-platform compatibility and low-latency transmission can be guaranteed.

[0086] Optionally, in some embodiments of the present application, if the number of fields unique to an e-commerce platform explodes (such as the newly added "live broadcast lucky bag" function field on Douyin) or model version conflicts lead to data confusion (coexistence of new and old order status fields on Taobao), then it can be optimized based on dynamic knowledge management. Specifically, the embodiments of the present application also include: configuring a preloading strategy, which is to inject the analysis results of hot data obtained in real time into the answers to the corresponding product activities in advance; configuring dynamic fields for the third type of storage structure, which are extension fields, model version numbers and discarded field lists.

[0087] Accordingly, after conducting a conversation with the second buyer according to the platform characteristics of the first e-commerce platform, the method further includes: dynamically allocating intelligent customer service or manual transfer to the selected sub-account based on the buyer's portrait, inquiry type, customer service performance, and customer service idle status; obtaining real-time hot spot data analysis results; injecting the real-time hot spot analysis results into the answer to the product activity of the first product; injecting the answer to the product activity of the first product into the dynamic field in the message storage structure; and sending the activity consultation answer obtained based on the message storage structure to the second conversation interface.

[0088] It can be seen that by preloading strategies and configuring dynamic fields, answers related to promotional activities can be injected into the dynamic fields of the third storage structure in advance based on real-time public opinion analysis, thereby improving conversation efficiency and reception service effects.

[0089] Optionally, in some embodiments of the present application, considering that some e-commerce platforms may prohibit reverse engineering, in order to achieve zero-protocol parsing, spatiotemporal knowledge modeling and deployment of dedicated processing containers can also be performed. Specifically, the dynamic knowledge base is derived from the knowledge bases of different e-commerce platforms, and the knowledge bases of different e-commerce platforms are all federated nodes; the method also includes:

[0090] Based on federated nodes, cross-supply chain consulting and cross-platform after-sales processes are established. The spatiotemporal knowledge modeling engine acquires encrypted knowledge vectors from multiple e-commerce platforms, fuses them using homomorphic encryption, and updates global model parameters using a federated averaging algorithm. These knowledge vectors include short video feature vectors and product parameter vectors. Dedicated processing containers are deployed for each e-commerce platform, each of which uses the platform's interface files to parse buyer messages. These dedicated processing containers serve as edge computing units and can establish unified data standards and interfaces to enable data sharing and collaboration between edge computing nodes, intelligent agents, and quantum state computing modules.

[0091] It can be seen that through this method, it is possible to parse buyer messages from different e-commerce platforms using the 0 protocol without reverse engineering the business platform, and maintain the integrity of the buyer messages.

[0092] Optionally, in some embodiments of this application, customized agents can be deployed based on the platform characteristics and buyer needs (including merchants and buyers) of different e-commerce platforms. A quantum state computing module is integrated into this session management platform based on a quantum simulator for algorithm optimization and testing.

[0093] Optionally, in some embodiments of the present application, federated knowledge queries may suffer from knowledge distillation loss (resulting in loss of detail, such as the high attenuation rate of micro-expression information in TikTok short videos), cold start difficulties (high error rate for new knowledge points), or message coordination delays between the session management platform and the e-commerce platform to which the message actually belongs. A consultation response solution can also be configured to predict knowledge hotspots (such as long-tail products) in the future time period to reduce the error rate. The present application method also includes:

[0094] 301. Build a cross-platform spatiotemporal coordinate system and mark the spatiotemporal entropy value of each knowledge point in the dynamic knowledge base.

[0095] The spatiotemporal knowledge modeling engine first generates a multi-dimensional semantic vector, then constructs a buyer interaction feature vector = [click-through rate x, dwell time y, transfer rate z], and finally introduces the spatiotemporal dimension t to obtain a cross-platform spatiotemporal coordinate system: (𝑥,𝑦,𝑧,𝑡) = GeoHash(𝐼𝑃) ⊕ timestamp

[0096] For example, during the Lunar New Year shopping festival, a hot-selling consultation scenario about a certain brand of nut gift box suddenly emerged. The nut gift box suddenly became popular on TikTok, and the number of inquiries surged by 500%. The cross-platform space-time coordinate system detected a surge in the space-time entropy value of the Shenzhen area, marked the Shenzhen area as a red starburst, and enhanced the gravitational field of the Shenzhen area, thereby making Taobao's inventory data, Pinduoduo's similar product information, and historical after-sales problem solutions close to the nut gift box in the nebula-shaped knowledge density distribution.

[0097] 302. Focus high-frequency knowledge points in the dynamic knowledge base based on the knowledge quality index and space-time curvature to form a nebula-shaped knowledge density distribution. The knowledge quality index is obtained based on access frequency, association strength, and timeliness. Optionally, a calculation formula for the knowledge quality index is as follows: Formula 1:

[0098] 𝑀=log(𝐹+1)×√S×𝑒 −𝑇 / 𝜏 Formula 1

[0099] Among them, access frequency (F), association strength (S), timeliness (T), and τ are the knowledge half-life.

[0100] The spacetime curvature is used to distort the distribution trajectory of a second knowledge point within the nebula-shaped knowledge density distribution based on the gravitational field generated by the mass of the first knowledge point. This quantifies the potential evolution of buyer behavior patterns and product popularity dynamics (e.g., the diffusion of hot spots). The second knowledge point is an adjacent knowledge point to the first knowledge point within the nebula-shaped knowledge density distribution. This spacetime curvature includes temporal curvature, spatial curvature, and behavioral curvature. Temporal curvature represents the acceleration of buyer attention over time; spatial curvature represents the degree of curvature of the diffusion path of a hot spot; and behavioral curvature represents the drastic degree of shift in buyer interaction patterns.

[0101] For example, if a 30-minute prediction of a surge in traffic for a particular knowledge point (e.g., detecting an unusual surge in traffic in a livestreaming room) is made, a corresponding intelligent customer service resource scheduling plan can be automatically triggered to accommodate the target merchant on that livestreaming platform. Another example, on Douyin, interest in down jackets among buyers in the Yangtze River Delta increases by 15% per hour, three times the rate of buyers in North China. By monitoring interest in down jackets, a peak in interest curvature was detected between 8:00 PM and 10:30 PM. This allows prediction of the spread of interest in the down jacket over the next 2-3 hours and identification of unusual fluctuations (for example, identifying a sudden change in space-time curvature in a region could indicate the risk of fraudulent orders).

[0102] The nebula-like knowledge density distribution refers to the nebula-like structure formed by product-related knowledge according to popularity, correlation density and timeliness. For example, it includes core areas, key areas and dark areas:

[0103] The core area (5%) consists of the highest-quality knowledge points, including product specifications and promotional rules for popular products. Examples include real-time inventory and pricing for popular Taobao products, key selling points currently being discussed in Douyin livestreams, and countdowns and remaining spots for Pinduoduo group purchases. The important area (15%) consists of less important knowledge points, including matching recommendations and frequently asked questions by buyers. These can be knowledge graphs, such as illustrated guides for pairing down jackets with sweaters, solutions to payment issues in mobile group purchases, and lipstick swatches and skin tone matching recommendations. The dark knowledge area (80%) consists of freely disseminated long-tail knowledge, unlabeled but potentially influential knowledge, such as historical data and solutions for handling extreme cases. Examples include complaint handling records for a specific product from two years ago, solutions to logistics anomalies in specific regions, and customized needs for niche buyer groups.

[0104] The core area monitors changes to basic product information in real time. The important area regularly updates recommended combinations (once every hour) and intelligently filters frequently asked questions. For example, if a buyer stays for more than 30 seconds, the associated knowledge is automatically enhanced (e.g., a pop-up ingredient test report is displayed). The dark knowledge area is used to activate historical data through gravitational field disturbances of spacetime curvature to respond to sudden public opinion. For example, detecting abnormal fluctuations in the dark knowledge area (such as those that match pre-set scalper fraud characteristics) triggers a defense mechanism.

[0105] 303. Generate a spatiotemporal heat map based on knowledge points from different e-commerce platforms. The spatiotemporal heat map includes knowledge overlap and conflict areas, knowledge blind spots, and new knowledge focuses based on the time dimension.

[0106] 304. Based on the nebula-shaped knowledge density distribution and the spatiotemporal heat map to simulate the evolution of knowledge density, an operation plan for the product to be operated is generated.

[0107] For example, the operation plan may include price or inventory updates for products (such as real-time group purchase data to enhance the sense of urgency), hourly updates of recommended combinations, and the use of historical data to respond to sudden public opinion (such as identifying scalpers, extreme price fluctuations, or analysis of past group purchases).

[0108] In some implementations, for the live streaming knowledge storm scenario:

[0109] The product to be operated is a product with peak traffic in the live broadcast room (i.e., when the traffic of a product's live broadcast room soars), and the knowledge quality index is dynamically obtained based on the real-time number of viewers, interaction rate, and conversion rate; step 304 includes:

[0110] 304-1: Simulates a response to a buyer's inquiry intent and associates product-related data that matches the buyer's inquiry intent. For example, if a buyer asks "Is sunscreen SPF60+ suitable for sensitive skin?", the system automatically associates the product ingredient list, laboratory test reports, and comparison data for similar products.

[0111] 304-2: Based on the spatio-temporal curvature, the commodity-related data is aggregated to the knowledge focus, and a plurality of intelligent dialogues are generated according to the platform characteristics of the first e-commerce platform using a diffusion model. For example, a picture-text mixed arrangement dialogue is generated using a diffusion model, the information presentation order is adjusted in combination with a real-time heat map, a knowledge hot area is identified, the style is "live goods", and the prompt words are "highlight" and "use urgent tone".

[0112] 304-3: Based on the spatio-temporal heat map analysis, the buyer behavior trajectory adjusts the dialogue characteristics of each intelligent dialogue, and each intelligent dialogue is rearranged. For example, according to the buyer behavior trajectory, the information density can be adjusted for the buyer position (browsing stage), and the dialogue characteristics of different browsing stages are obtained as shown in the following table:

[0113] stage gravitational strength Characteristics of speech Dark Knowledge Area Low Generalized recommendations (e.g., "This sunscreen is very popular recently") Important areas middle Parameter comparison (e.g., "SPF50+ is 30% higher than similar products") core area high Urgency-enhancing sales pitch (e.g., "Only 200 orders left")

[0114] In some other embodiments, for a cross-platform policy fluctuation scenario, platform 1 suddenly adjusts the 10 billion subsidy rules of a certain brand of mobile phone, resulting in price fluctuations in platforms 2 and 3, and the number of buyer inquiries increases several times. Considering that buyer questions also show regional differences and platform differences, the buyer geographic distribution of platform 2, the live room real-time barrage of platform 3, and the order flow of platform 1 can be collected in real time, and a three-dimensional heat cloud map can be generated based on the buyer geographic distribution, the live room real-time barrage, and the order flow. The heat map features of platforms 1, 2 and 3 are extracted based on the three-dimensional heat cloud map, and the spatio-temporal dimension is introduced to optimize the dialogue of the mobile phone on each platform in space and time, to reconstruct the dialogue, and then rearrange the reconstructed dialogue based on the heat weighting algorithm. For example, in the Yangtze River Delta region, strengthen price protection; for the Northeast region, increase logistics commitment; for the live room, alleviate inventory anxiety. It can be seen that by configuring the consultation response scheme for the knowledge hotspots (such as long-tail commodities) in the future period, the error rate of long-tail commodities can be reduced, and various price protection, return policy and other problems caused by the lag of customer service dialogue update on each platform can be reduced.

[0115] Optionally, in some embodiments, the artificial transfer process can also be optimized, and the preset shunt strategy includes a cross-platform load balancing strategy configured for each edge computing node in the edge computing node cluster, for mutual support and transfer of artificial customer service of the target merchant on different e-commerce platforms.

[0116] Optionally, in some embodiments of the present application, the method further includes: configuring an event update strategy; obtaining reception data of each sub-account, obtaining a speech set from the reception data, the speech set including excellent reception speech with reception quality higher than a preset quality, and reception speech generated based on historical reception conversation answers; when the speech set is detected to have changed based on the event update strategy, synchronizing the changed speech set to each sub-account. It is understandable that before synchronizing the speech, adaptive adjustments can also be made based on the differences in platform characteristics to adapt to the speech of the platform to be synchronized. It can be seen that by configuring the event update strategy, dynamic synchronization of excellent speech across sub-accounts can be achieved on this intelligent customer service aggregation platform, which can further improve reception services.

[0117] Optionally, in some embodiments, cross-platform knowledge transfer can be configured, and the method further includes: obtaining the hit rate and activity of a single knowledge point and summing them; and obtaining the knowledge importance weight of each e-commerce platform based on the hit rate and activity of the single knowledge point and the sum. The knowledge importance weight can be adaptively and dynamically adjusted based on buyer feedback. It can be seen that by setting up federated knowledge collaborative adjustment, the feedback closed-loop mechanism of knowledge importance weight can be strengthened, thereby improving the portability of cross-platform knowledge. While reducing the limitations of a single platform knowledge base, it is possible to synchronize the activity rules of each platform, so that the answers of the intelligent customer service bound to the target merchant using this session management platform are more in line with the context of the current e-commerce platform.

[0118] Optionally, in some embodiments, Figure 5 The following is a flow chart of a multi-platform intelligent customer service session management method. This method is applied to a session management platform, where a target merchant uses multiple intelligent customer service representatives to receive inquiries from buyers on multiple e-commerce platforms. This embodiment of the application includes:

[0119] 401. In response to the target merchant's login request on the management interface, select the target merchant's sub-account on multiple e-commerce platforms to log in. The management interface includes the first area, the second area and the third area. Figure 6a , showing the conversation interface between platform 1 buyers and intelligent customer service. The first area displays the icons of each sub-account in real time according to the conversation update time. Each sub-account is associated with at least one intelligent customer service.

[0120] The second area displays the session interface of the selected first sub-account, and can also display the customer service performance evaluation parameters, order status, and excellent dialogue of the consulted goods. It can be understood that each time the sub-account receives a new consultation, the pre-set excellent dialogue of the goods will be collected according to the current consultation of the buyer and synchronized to the second area in real time. When the sub-account receives consultation of other goods, the excellent dialogue of the consulted goods will be updated in the second area. The excellent dialogue of the goods includes, but is not limited to, consultation of product parameters, consultation of product activities, consultation of product after-sales, etc. The excellent dialogue can be a pre-set dialogue, or dynamically generated according to the pre-set dialogue and real-time session record. These excellent dialogues can be sent to the session interface by dragging, or the target excellent dialogue can be matched according to the real-time session record and highlighted for the artificial customer service to choose, or displayed in the session interface (the buyer cannot see) to prompt the artificial customer service to send to the session interface by clicking to confirm.

[0121] The third area displays the dynamic goods data of the consulted goods on the entire e-commerce platform. The dynamic goods data includes the transaction status distribution, heat data, goods distribution and inventory data, and visualized goods features of the consulted goods. The visualized goods features include the selling point parameters of the consulted goods, such as long standby, intelligent dialogue, folding large screen, etc., without the need for artificial customer service to click to the goods detail page for response. The dynamic goods data can be used by intelligent customer service or artificial customer service to generate style answers in real time according to the buyer's intention, e-commerce platform features, and session progress with the buyer.

[0122] 402. Receive a first input for a first sub-account icon of a first e-commerce platform in the first area, display a first session interface in the second area, conduct a session with a first buyer, and synchronously display first dynamic goods data of a first goods currently consulted by the first buyer in the third area.

[0123] The session can be conducted according to the platform features of the first e-commerce platform and the first buyer.

[0124] 403. Receive a second input for a second sub-account icon of a second e-commerce platform, switch from the first session interface to a second session interface, conduct a session with a second buyer, and update the visualized content in the first area and the second area.

[0125] Updating the visualized content means updating the display order of the sub-account icon list in the first area according to the time sequence of different buyer messages, lighting the sub-account icon of unread messages, and dynamically switching the first dynamic goods data of the first goods previously displayed to the second dynamic goods data of the second goods according to the second goods consulted by the second buyer. (The first goods and the second goods can be the same or different)

[0126] It can be understood that if the whole process is an intelligent customer service reception, and the manual customer service of the target merchant does not selectively browse each conversation, the entire management interface will change the display order of the sub-account icons in the first area due to message updates, and does not necessarily switch the conversation interface corresponding to the latest message at the same time.

[0127] The second buyer can be conversed according to the platform characteristics of the first e-commerce platform, that is, when the intelligent customer service generates an answer, the strategy of intelligently generating an answer matching the buyer's inquiry can be according to the platform characteristics. Specifically, the corresponding answer can be generated based on the platform-specific dialogue library, or the platform-specific dialogue library can be used to automatically generate buttons that conform to the platform specification in the answer. The answer can be parsed by the first e-commerce platform and displayed to the second buyer. It should be noted that if the first buyer and the second buyer inquire about the same product, the intelligent customer service receiving can ensure that the core information of the answers output when inquiring about the same product on the first and second e-commerce platforms is consistent according to the multi-end experience consistency strategy.

[0128] It can be seen that by aggregating the sub-accounts of the target merchant on different e-commerce platforms in the conversation management platform and configuring intelligent customer service matching the platform characteristics for each sub-account, on the one hand, the cross-platform conversation management mechanism enables the target merchant to intuitively and conveniently send and receive messages on a single platform, improving conversation efficiency; on the other hand, the target merchant can click on the sub-account icon to switch to the inquiry on the e-commerce platform received by the sub-account icon, and the customer service of the target merchant does not need to switch between the customer service accounts of multiple e-commerce platforms, which can reduce the operation and can also concentrate the customer service accounts of different e-commerce platforms in a single platform. A manual customer service or intelligent customer service can receive inquiries from buyers of multiple e-commerce platforms on a single platform, thereby reducing operating costs, improving conversation efficiency and buyer experience.

[0129] Optionally, in some embodiments of the present application, considering that the same buyer may inquire about the same problem multiple times on different e-commerce platforms, in order to more comprehensively understand, objectively analyze the inquiry purpose and quickly respond, semantic association or spatio-temporal association can also be performed. Specifically, the following operations S1-S2 are included:

[0130] S1: Identify associated conversations through semantic association or spatio-temporal association.

[0131] Among them, the associated conversation refers to multiple inquiry conversations initiated by the same buyer on different e-commerce platforms for the same problem or associated problem. The associated problem refers to a problem that has a greater degree of association with the current problem.

[0132] In some embodiments, the following cross-platform semantic alignment model can be used:

[0133]

[0134] Where $Q_i$ is the text question and $M_i$ is the associated multimedia information (screenshot / video clip)

[0135] For example, a buyer's Douyin inquiry video: "Why hasn't this dress arrived yet?" + Logistics screenshot

[0136] Taobao inquiry text: "Logistics for order number 12345 is not updated"

[0137] The cross-platform semantic alignment model outputs: $\cos(\vec{v}{TikTok}, \vec{v}{Taobao}) = 0.92$

[0138] Semantic similarity calculation can be used to determine historical conversations with semantic associations.

[0139]

[0140] Among them, α, β, and γ are all weights, α>β, α>γ.

[0141] Spatiotemporal associations can be identified using a joint spatiotemporal embedding approach. For example, a buyer's location "wx4er5" → Taobao warehouse "wx4er9" → determined to be in the same city. After identifying the associated sessions, the buyer's behavior trajectory can be modeled to facilitate identification of the buyer's true inquiry intent chain, as follows: watching a product video on Douyin (30 minutes) → searching for the same product on Taobao (2 hours) → comparing prices on Pinduoduo (4 hours) → consulting in a Douyin livestream. Modeling this buyer's behavior trajectory allows prediction of the platform the buyer might consult next.

[0142] S2: The at least two associated sub-accounts are merged into two conversation windows in the second area and the conversation history is displayed simultaneously. For example, Douyin inquiries and Taobao orders from the same buyer can be automatically linked, or customer service replies to Taobao orders can be sent a Douyin product link simultaneously, achieving cross-platform conversation message aggregation.

[0143] In some implementations, the interaction timeline of the first product on the entire e-commerce platform can be visualized in the second area, marking the key event nodes of the first product. The interaction timeline can visualize the buyer's cross-platform behavior trajectory. Merging at least two associated sub-accounts into two session windows in the second area to synchronously display the session records can refer to Figure 6b The reception interface of customer service (including any combination of Xiaoshu and intelligent customer service) simultaneously displays the conversation interfaces of platform 1 buyers and platform 2 buyers with intelligent customer service.

[0144] In the present application, the two session windows include a first session window of a sub-account of a current session and a second session window of a second sub-account that has a semantic association or a spatiotemporal association with the first sub-account.

[0145] The second area includes a first sub-area and a second sub-area. The first conversation window is displayed in the first sub-area, and the second conversation window is displayed in the second sub-area. There can be at least one second sub-account. If there is one, the first conversation window and the second conversation window are displayed side by side. If there are at least two, the first conversation window occupies the same size of the second area, and multiple second conversation windows are displayed in the second sub-area. The multiple second conversation windows in the second sub-area can be dynamically displayed using any operation such as sliding, dragging, or scrolling in a preset direction to display the global content of a single second conversation window. This application is not limited to this.

[0146] For example, a buyer tracks after-sales issues with home appliances. The buyer's behavior trajectory is:

[0147] Douyin consultation: "What to do if the air conditioner is cooling poorly" (including troubleshooting video)

[0148] Taobao customer service: "The air conditioner with order number XXX needs repair"

[0149] Pinduoduo: "Same model air conditioner installation problem"

[0150] The response process of the intelligent customer service associated with the sub-account currently receiving the buyer is as follows:

[0151] Semantic association recognition: fault description similarity 0.88

[0152] Time and space correlation: the interval between three consultations is less than 48 hours, and the location distance is less than 3km

[0153] Automatically merge conversation records and prompt intelligent customer service to prioritize local repair orders.

[0154] Optionally, the first prompt information may be highlighted in the second area, and in response to the target merchant's first input to the first prompt information, at least two associated sub-accounts may be merged into two session windows in the second area to synchronously display the session records.

[0155] As can be seen, on the one hand, whether human or intelligent customer service representatives handle inquiries from multiple e-commerce platforms on the same session management platform, the related sessions can be automatically linked and displayed simultaneously on the platform, facilitating comparison and effective responses. On the other hand, by identifying related sessions, buyer behavior trajectories can be modeled, facilitating the identification of the true intent chain of buyer inquiries. This allows for the visualization of cross-platform buyer behavior in a virtual space, ultimately improving the intelligent customer service's response and resolution speed, and reducing the rate of repeated inquiries from buyers.

[0156] Optionally, in some embodiments of the present application, the answers of each intelligent customer service are generated based on one of the first type of storage structure, the second type of storage structure, or the third type of storage structure. The method further includes: generating answers based on the first type of storage structure according to the platform characteristics of the first e-commerce platform, the first type of storage structure including message structure definition, signature rules, and session mechanism; generating answers based on the second type of storage structure according to the platform characteristics of the second e-commerce platform, the second type of storage structure including protobuf description files and live broadcast special effects mapping; generating answers based on the third type of storage structure according to the platform characteristics of the third e-commerce platform, the third type of storage structure including group buying rules and instant activity strategies.

[0157] Generally speaking, the first type of storage structure can be preset to be dedicated to the first e-commerce platform, the second type of storage structure can be preset to be dedicated to the second e-commerce platform, and the third type of storage structure can be preset to be dedicated to the third e-commerce platform. It should be noted that if some e-commerce platforms belong to different implementation forms of the same product of the same operator, they can share the same type of storage structure to generate corresponding answers. For example, if an operator launches a shopping APP and the shopping applet at the same time, it should be regarded as using the same type of storage structure. For another example, the service providers of different e-commerce platforms may develop similar APPs and buyer groups, and they can also reuse the storage structure of similar e-commerce platforms to generate answers. There is no limitation on this.

[0158] It can be seen that according to the pre-configured differentiated answer service strategy, when actually receiving buyers from different e-commerce platforms, corresponding answers can be generated based on the platform characteristics of the e-commerce platform to which the buyer belongs and the preset storage structure of the platform characteristics, thereby achieving differentiated responses, rather than using a unified speech style, which leads to a situation where the conversation effect is poor due to the consistent style.

[0159] Optionally, in some embodiments of the present application, the method further includes: searching for potential collaborative historical information from all e-commerce platforms based on the second buyer's account, order contact information, and inquiry content, wherein the potential collaborative historical information is historical inquiries of the second buyer for similar inquiries on non-second e-commerce platforms; the similar inquiries include identical inquiries; obtaining first key information from the potential collaborative historical information; referencing or merging the first key information into the second session interface, and sending an answer referencing the first key information to the second session interface.

[0160] Specifically, the intelligent agent can be responsible for cross-platform message linkage and similar message aggregation. When a buyer initiates relevant inquiries on multiple platforms, the intelligent agent identifies the buyer's identity through a unified identity authentication mechanism and associates and integrates messages from different e-commerce platforms. For example, if a buyer inquires about the price and inventory of the same product on e-commerce platform A and e-commerce platform B, the intelligent agent can combine and process the two messages and reply based on the information from both e-commerce platforms, avoiding repeated inquiries from the buyer. As can be seen, this approach can aggregate the same inquiries from buyers across platforms, facilitating timely and comprehensive understanding.

[0161] Optionally, in some embodiments of the present application, considering the disordered and redundant multi-platform conversation data and the different expressions of different buyers on different platforms, the present application can also sort out key information of interest to merchants from the data, facilitating effective tracking, diagnosis and intelligent management. Specifically, the method further comprises:

[0162] obtaining target historical data, the target historical data being all excellent scripts about the second product in the present aggregation platform, and conversation records of all e-commerce platforms; obtaining first type data and second type data from the target historical data and sending them to a server to which the target merchant belongs, the first type data being hotness data of the second product from buyers, and the second type data being after-sales status data of the second product; obtaining a hot spot analysis result of the second product on all e-commerce platforms in real time according to the first type data, injecting the hot spot analysis result into a dynamic expansion field in the answer of the product activity, to generate a target answer and display it on the second conversation interface.

[0163] The first type data is used by the server to analyze a global sales distribution heat map of the second product on all e-commerce platforms, and to generate a differentiated stocking strategy for all e-commerce platforms based on the global sales distribution heat map and the characteristics of the buyer groups of the e-commerce platforms. The second type data is used by the server to analyze the global after-sales distribution status of the second product on all e-commerce platforms, and to generate an after-sales strategy accordingly.

[0164] Optionally, in some embodiments of the present application, each e-commerce platform is a federal node, and the conversation management platform is a federal center node; the method further comprises: receiving a first message from a second buyer, the first message being used to inquire about discount information of a hot-selling product; mapping the first message into a second quantum state; triggering a federal query based on the second quantum state, and distributing an encryption request to each e-commerce platform; receiving discount information from each e-commerce platform, and performing quantum aggregation on each discount information to obtain a target discount scheme; after decrypting the target discount scheme, sending it to the second conversation interface.

[0165] Specifically, quantum computing is used to classify and cluster massive amounts of buyer inquiries, identifying optimal classification solutions. Quantum computing can help intelligent agents select the optimal response strategy. For example, when handling complex buyer complaints, quantum computing can quickly assess the effectiveness and risks of different solutions, providing the agent with the best decision-making advice.

[0166] In some implementations, quantum state computing can help intelligent agents select the optimal response strategy. When handling complex buyer complaints, quantum state computing can quickly evaluate the effectiveness and risks of different solutions and provide the intelligent agent with the best decision-making recommendations. Figure 7 As shown, the embodiment of the present application includes:

[0167] 501: Obtain buyer complaint data for the second product and create a model.

[0168] Buyer complaint data includes buyer complaint text (structured data after semantic analysis, such as complaint type, involved products / services, emotional intensity, etc.), historical complaint case library (including solutions, processing results, buyer feedback scores), and platform rules and constraints (such as return and exchange policies, compensation limits, and response time requirements).

[0169] 502: Extract complaint features from the buyer complaint data and map the complaint features into a third quantum state.

[0170] Complaint features include complaint type, sentiment intensity, and a historical case library. For example, complaint type can be mapped to a binary code (e.g., electronics = 00, logistics = 01). Sentiment intensity can be mapped to a continuous variable (using quantum amplitude encoding, e.g., |ψ> = α|0> + β|1>, with α²=0.7 indicating high sentiment). Furthermore, the historical case library can be mapped to a quantum superposition state (multiple cases stored in parallel).

[0171] 503: Generate multiple candidate solutions based on the third quantum state and platform characteristics.

[0172] Among them, the candidate solutions can be a predefined rule base or dynamically generated by combining the characteristics of the e-commerce platform and the buyer portrait. The two can be selected alone or used in combination without any limitation.

[0173] 504: Represent the multiple candidate solutions in quantum states to obtain a fourth quantum state.

[0174] For example, |ψ_solution> = Σ c_i |solution_i>, where c_i is the weight coefficient. Buyer complaint data and multiple candidate solutions can be encoded into quantum states.

[0175] 505: Generate a target superposition state of the third quantum state and the fourth quantum state through quantum gate operation.

[0176] 506: Estimate the matching degree of the fourth quantum state of each candidate solution with the third quantum state through quantum phase estimation, amplify the amplitude of the first candidate solution and suppress the amplitude of the second candidate solution through quantum interference operation.

[0177] Among them, the first candidate solution is the candidate solution with a matching degree higher than the first threshold value, and the second candidate solution is the candidate solution with a matching degree lower than the second threshold value. The first threshold value is greater than the second threshold value.

[0178] 506: Simulate the buyer feedback distribution after the execution of the first candidate solution by using the quantum Monte Carlo method.

[0179] Among them, the buyer feedback distribution is the satisfaction score and the probability of secondary complaints. Quantum annealing algorithm can also be used to optimize the dilemma of jumping out of local optimal solution in the multi-complaint concurrent processing scenario.

[0180] 507: Encode the risk parameter as a quantum state noise model and dynamically adjust the evaluation weight.

[0181] The risk parameter is the compensation amount and legal risk.

[0182] 508: Entangle the buyer complaint data and the buyer historical behavior data, predict the change of the buyer retention rate, and perform quantum state correlation analysis through the third quantum state and the fourth quantum state to identify the correlation risk factors.

[0183] For example, one of the correlation risk factors that leads to a decrease in repeat purchase rate is to identify the reason for the logistics delay.

[0184] 509: Map the quantum state amplitude distribution of the first candidate solution and the second candidate solution to a quantum state heat map, and label the key influencing factors in the quantum state heat map.

[0185] For example, the weight proportion of logistics delay is 60%. Among them, the key influencing factors are obtained according to the correlation risk factors identified for each quantum state in step 508. As can be seen, through the visual display of the quantum state, the entire reception service can be tracked through the quantum interference path in real time during the reception, thereby generating an inquiry decision causal chain. For example, the inquiry decision causal chain is “high emotion → priority pacification → compensation coupon”.

[0186] 510: Visualize the quantum state heat map in the second area.

[0187] For example, taking the inquiry of the buyer about the preferential activities of the brand X, model X00p mobile phone (such as the lowest price + gift) as an example, the preferential activity parameters submitted by each e-commerce platform to the conversation management platform include: discount rate (di, such as 9 times of the price of the Plus member of Jingdong → 0.9), inventory (si, such as 1000 units of X00p inventory),

[0188] Buyer matching degree (ui, similarity based on historical consultations, range [0,1]), activity validity period (ti, unit: hours).

[0189] First, the session management platform encodes each activity parameter into a quantum state, which is as follows:

[0190] JD.com: d=0.85, s=500, u=0.92, t=72

[0191] Taobao: d=0.88, s=300, u=0.88, t=48

[0192] Pinduoduo: d=0.75, s=1000, u=0.85, t=96

[0193] Secondly, the session management platform uses the quantum annealing algorithm to superimpose the discount rate code |di>, inventory quantity code |si>, buyer matching code |ui> and validity period code |ti> to obtain the target discount plan |ψi>=|di>⊗∣si>⊗∣ui>⊗∣ti>, and finally returns the target discount plan to the buyer terminal.

[0194] For example, after optimizing campaign parameters, the optimal solution (i.e., the target discount plan) is Taobao (d = 0.88) + Pinduoduo (s = 1000). If federated verification confirms that the total inventory is 1300 ≤ 1500, then buyer coverage is met and Taobao's weight is increased.

[0195] As can be seen, quantum superposition enables the parallel evaluation of multiple candidate solutions. Quantum interference can prevent the output of locally optimal candidate solutions. By leveraging edge computing and reducing latency, the synergy of quantum annealing and federated learning enables real-time optimization of cross-platform promotions. This involves simultaneously encoding multi-dimensional promotion parameters such as discount, inventory, and buyer matching. After local annealing calculations on each e-commerce platform, the session management platform aggregates the results to obtain the global optimal solution. Based on buyer feedback, the weights of the promotion parameters for each platform in the target promotion are dynamically adjusted, thus achieving real-time optimization of cross-platform promotions.

[0196] Optionally, in some embodiments of the present application, the second product is a long-tail product, or the candidate product recommended by the second sub-account to the second buyer is a long-tail product. Figure 8 As shown, the conversation with the second buyer according to the platform characteristics of the first e-commerce platform includes:

[0197] Based on the second buyer's inquiry intent, associating product-related data that matches the inquiry intent; clustering the product-related data toward a knowledge focus based on spatiotemporal curvature, and using a diffusion model based on the platform characteristics of the first e-commerce platform to generate multiple intelligent sales pitches for the second product on the second e-commerce platform;

[0198] The multiple intelligent speech techniques are reconstructed in time and space; the speech features of each reconstructed intelligent speech technique are adjusted based on the buyer behavior trajectory and heat weighting algorithm analyzed in the time and space heat map, and each intelligent speech technique is rearranged and updated to the second area.

[0199] Spatiotemporal reconstruction involves incorporating time (e.g., pre-sales, flash sales) and trajectory (e.g., regional promotional policies: XX duty-free warehouse, free shipping in XX region) features into intelligent sales. This allows the intelligent sales to match buyers' current needs in real time and attract buyers to close transactions. For example, the East China edge node cluster receives a request from a buyer on the Taobao platform and pre-extracts basic platform features: {Platform: Taobao, Promotion Period: Double 11}. The platform feature recognition engine enhances these features and transmits them to the spatiotemporal knowledge modeling engine, which extracts the original knowledge (product: XX mobile phone, price guarantee: true). This knowledge is then transformed into spatiotemporal information, resulting in "Orders purchased between 11.11 and 11.11 are eligible for price guarantee and enjoy a 7-day automatic price guarantee." This information is then output to the intelligent customer service representative, who generates a differentiated response based on the Taobao platform features: "Dear, XX mobile phones purchased during Double 11 enjoy a 7-day automatic price guarantee. Click [My Price Guarantee] to apply." It can be seen that by configuring a consultation response plan that predicts knowledge hotspots (such as long-tail products) in the future time period, it is possible to reduce the misanswer rate for long-tail products and alleviate various problems such as price assurance and return policies caused by the lagging updates of customer service scripts on various platforms.

[0200] Figures 1 to 8 Any technical feature mentioned in any embodiment corresponding to any one of the items is also applicable to the embodiments of this application. Figures 9 and 10 The corresponding embodiments will not be described in detail later.

[0201] The above describes a multi-platform intelligent customer service session management method in an embodiment of the present application. The following describes a session management platform for executing the multi-platform intelligent customer service session management method. Figure 9 ,like Figure 9 The structure diagram of a session management platform 80 is shown, which can be applied to cross-e-commerce platform intelligent customer service reception scenarios, and improve the efficiency of conversation by aggregating the intelligent customer service reception services of multiple platforms. Figures 1-8The steps in the multi-platform intelligent customer service session management method in any corresponding embodiment. The session management platform 80 may include an input and output module 801, a display module 802 and a processing module 803. The processing module 803 controls the input and output operations of the input and output module 801 and controls the display operation of the display module 802. The functional implementation of each module can be referred to Figures 1-8 The operations performed in any corresponding embodiment are not described in detail here.

[0202] In some embodiments, the processing module 803 is configured to respond to a login request from a target merchant on a management interface and select a sub-account of the target merchant on multiple e-commerce platforms for login; wherein the management interface includes a first area, a second area, and a third area, wherein the first area displays icons of each sub-account in real time according to session update time, and each sub-account is associated with at least one intelligent customer service representative; the second area displays a session interface of the selected first sub-account, and the third area displays dynamic product data of the inquired product across all e-commerce platforms;

[0203] After receiving a first input for a first sub-account icon of a first e-commerce platform in a first area through the input / output module 801, the display module 802 is controlled to display a first conversation interface in a second area, conduct a conversation with the first buyer according to the platform characteristics of the first e-commerce platform, and synchronously display first dynamic product data of the first product currently being inquired by the first buyer in a third area;

[0204] The input-output module 801 receives a second input for a second sub-account icon of the second e-commerce platform, controls the display module 802 to switch from the first conversation interface to the second conversation interface, conducts a conversation with the second buyer according to the platform characteristics of the first e-commerce platform, and updates the visual content of the first area and the second area accordingly.

[0205] Other functions of the input and output module 801, the display module 802 and the processing module 803 can be found in Figures 1-8 It can be seen that by adopting this solution, it is possible to centrally receive inquiries from buyers of the same merchant on different e-commerce platforms, improve conversation efficiency and provide an intuitive and convenient conversation experience, and facilitate cross-platform message management.

[0206] The above describes the session management platform 80 for executing the multi-platform intelligent customer service session management method in the embodiment of the present application from the perspective of modular functional entities. The following describes the session management platform 80 for executing the multi-platform intelligent customer service session management method in the embodiment of the present application from the perspective of hardware processing. It should be noted that in the embodiment of the present application Figure 9The physical device corresponding to the input / output module 801 in the illustrated embodiment may be an input / output unit, a transceiver, a radio frequency circuit, a communication module, an output interface, etc.; the physical device corresponding to the display module 802 may be a display screen; and the physical device corresponding to the processing module 803 may be a processor. Figure 9 The session management platform 80 shown may have Figure 10 The structure shown, when Figure 9 The session management platform 80 shown has Figure 10 When the structure shown is Figure 10 The processor and transceiver in the embodiment can realize the same or similar functions as the input / output module 801, the display module 802 and the processing module 803 provided in the aforementioned apparatus embodiment corresponding to the session management platform 80. Figure 10 The memory in the storage processor executes the computer program that needs to be called when the multi-platform intelligent customer service session management method is executed.

[0207] In the above embodiments, the description of each embodiment has its own focus. For parts not described in detail in one embodiment, reference can be made to the relevant descriptions of other embodiments. Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, devices, and modules can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0208] The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, the functional modules in each embodiment of the present application may be integrated into a processing module, or each module may exist physically alone, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may be stored in a computer-readable storage medium.

[0209] In the above embodiments, all or part of the embodiments may be implemented using software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments may be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program is loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part.

[0210] The above is a detailed introduction to the technical solution provided by the present application. Specific examples are used in the present application to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those skilled in the art, according to the ideas of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A multi-platform intelligent customer service session management method, characterized in that: The method is applied to a session management platform, where a target merchant uses multiple intelligent customer service representatives to receive inquiries from buyers on multiple e-commerce platforms. The method includes: In response to a login request from a target merchant on a management interface, a sub-account of the target merchant on multiple e-commerce platforms is selected for login; wherein the management interface includes a first area, a second area, and a third area, wherein the first area displays icons of each sub-account in real time according to session update time, and each sub-account is associated with at least one intelligent customer service representative; the second area displays a session interface of the selected first sub-account, and the third area displays dynamic product data of the inquired product across all e-commerce platforms; receiving a first input for a first sub-account icon of a first e-commerce platform in a first area, displaying a first conversation interface in a second area, conducting a conversation with a first buyer according to platform features of the first e-commerce platform, and synchronously displaying first dynamic product data of a first product currently being inquired by the first buyer in a third area; Upon receiving a second input for a second sub-account icon of a second e-commerce platform, the first area switches from the first conversation interface to the second conversation interface, and a conversation is conducted with the second buyer according to the platform characteristics of the first e-commerce platform. The first area and the second area update visual content accordingly.

2. The method according to claim 1, characterized in that The method further comprises: Identify related sessions through semantic association or spatiotemporal association. Related sessions refer to multiple consultation sessions initiated by the same buyer on different e-commerce platforms regarding the same or related issues. The at least two associated sub-accounts are merged into two session windows in the second area and the session records are displayed synchronously.

3. The method according to claim 2, characterized in that The answer of each intelligent customer service is generated based on one of the first type of storage structure generating answer, the second type of storage structure generating answer, or the third type of storage structure generating answer; the method further includes: generating an answer based on a first type of storage structure according to platform characteristics of the first e-commerce platform, wherein the first type of storage structure includes a message structure definition, a signature rule, and a session mechanism; generating an answer based on a second type of storage structure according to platform characteristics of the second e-commerce platform, wherein the second type of storage structure includes a message structure description file and a live broadcast special effect mapping; According to the platform characteristics of the third e-commerce platform, an answer is generated based on the third type of storage structure, and the third type of storage structure includes group buying rules and instant activity strategies.

4. The method according to claim 3, characterized in that The method further comprises: Searching for potential collaboration history information across all e-commerce platforms based on the second buyer's account, order contact information, and inquiry content. The potential collaboration history information is historical inquiries by the second buyer for the same inquiry on non-second e-commerce platforms. Acquiring first key information from the potential collaborative history information; The first key information is referenced or incorporated into the second conversation interface, and an answer referencing the first key information is sent to the second conversation interface.

5. The method according to claim 4, characterized in that Each e-commerce platform is a federated node, and the session management platform is a federated central node; the method further includes: receiving a first message from a second buyer, wherein the first message is for inquiring about discount information of a hot-selling product; Mapping the first message into a second quantum state; Triggering a federated query based on the second quantum state and distributing the encrypted request to each e-commerce platform; Receive discount information from various e-commerce platforms and perform quantum aggregation on each discount information to obtain the target discount plan; After decrypting the target preferential plan, it is sent to the second session interface.

6. The method according to claim 5, characterized in that The communicating with the second buyer according to the platform characteristics of the first e-commerce platform includes: Based on the inquiry intention of the second buyer, associating product-related data that matches the inquiry intention; Aggregating the product-related data toward a knowledge focus based on spatiotemporal curvature, and generating multiple intelligent sales pitches for the second product on the second e-commerce platform using a diffusion model according to the platform characteristics of the first e-commerce platform; Performing spatiotemporal reconstruction on the multiple intelligent speech techniques; The buyer behavior trajectory based on the spatiotemporal heat map analysis and the heat weighting algorithm are used to adjust the speech features of each reconstructed intelligent speech, and each intelligent speech is rearranged and updated to the second area.

7. The method according to claim 6, characterized in that The method further comprises: Obtain target historical data, where the target historical data is all excellent sales pitches for the second product on the aggregation platform and conversation records across all e-commerce platforms; Obtaining first and second types of data from the target historical data and sending them to a server belonging to the target merchant, wherein the first type of data is the buyer's enthusiasm for the second product, and the second type of data is the after-sales status data of the second product; Obtaining, in real time, a hotspot analysis result of the second product across the entire e-commerce platform based on the first type of data, and injecting the hotspot analysis result into a dynamically extended field in the answer to the product activity to generate a target answer and display it on the second conversation interface; The first type of data is used by the server to analyze a global sales distribution heat map of the second product across the entire e-commerce platform, and to generate a differentiated distribution strategy across the entire e-commerce platform based on the global sales distribution heat map and characteristics of the buyer group of the e-commerce platform; The second type of data is used by the server to analyze the global after-sales distribution status of the second product on the entire e-commerce platform and generate a corresponding after-sales strategy.

8. The method according to any one of claims 1 to 7, characterized in that The method further comprises: Obtain buyer complaint data for the second product and build a model; Extracting complaint features of the buyer's complaint data, and mapping the complaint features into a third quantum state; generating a plurality of candidate solutions based on the third quantum state and platform characteristics; Representing the plurality of candidate solutions in quantum states to obtain a fourth quantum state; The third quantum state and the fourth quantum state generate a target superposition state through a quantum gate operation; The matching degree between the fourth quantum state and the third quantum state of each candidate solution is estimated through quantum phase, and the amplitude of the first candidate solution is amplified and the amplitude of the second candidate solution is suppressed through quantum interference operation; Use the Quantum Monte Carlo method to simulate the distribution of buyer feedback after the first candidate solution is implemented; Encode risk parameters into quantum state noise models and dynamically adjust assessment weights; Entangling the buyer complaint data and the buyer's historical behavior data to predict changes in buyer retention rates, and performing quantum state correlation analysis through the third and fourth quantum states to identify correlation risk factors; Mapping the quantum state amplitude distributions of the first candidate solution and the second candidate solution to a quantum state heat map, and marking the key influencing factors obtained based on the correlation risk factors in the quantum state heat map; The quantum state heat map is visualized and displayed in the second area.

9. A session management platform, characterized in that: The session management platform includes: at least one processor and memory; The memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the method according to any one of claims 1 to 8.

10. A computer program product comprising instructions, wherein the computer program product comprises program instructions, and when the program instructions are run on a computer or a processor, the computer or the processor is caused to perform the method according to any one of claims 1 to 8.

Images