Intelligent customer service session aggregation performance adjusting method, platform and program product

Abstract

The invention relates to the field of AI platforms, and provides an intelligent customer service session aggregation performance adjustment method, platform and program product, and the method comprises the steps: obtaining a target merchant on an e-commerce session management platform, aggregating the sub-accounts of the target merchant on different e-commerce platforms for the target merchant, and configuring an intelligent customer service matched with the platform features. Through cross-platform aggregation, user inquiry of a target merchant in different e-commerce channels is received, the session efficiency is improved, and cross-platform message management is facilitated; a knowledge base blind area is analyzed based on a knowledge base hit rate thermodynamic diagram, platform differences are identified, a service delay trend of an intelligent customer service of a certain platform is captured in real time based on a cross-platform response delay distribution diagram to dynamically balance cross-platform loads, and a corresponding linkage strategy is executed through session transfer rate trend analysis. And the performance optimization of the intelligent customer service reception service is automatically triggered.

Description

Technical Field

[0001] This application relates to the field of AI platform technology, and in particular to a method, platform and program product for adjusting the performance of intelligent customer service conversation aggregation. Background Technology

[0002] In the e-commerce retail sector, merchants open stores on multiple platforms such as Taobao, Douyin, and Pinduoduo. They typically purchase Artificial Intelligence (AI) customer service for each platform to improve customer service efficiency. However, due to differences in platform characteristics and user preferences, there is currently no method to monitor and dynamically optimize the performance of these AI customer service systems when listing the same or similar products on different platforms, resulting in low conversation efficiency. Summary of the Invention

[0003] This application provides a method, platform, and program product for performance tuning of intelligent customer service conversation aggregation, which can centrally handle user inquiries from the same merchant on different e-commerce channels, improve conversation efficiency, and optimize the performance of cross-platform customer service message management.

[0004] Firstly, this application provides a method for adjusting the performance of intelligent customer service conversation aggregation from the perspective of an e-commerce conversation management platform. The method is applied to an e-commerce conversation management platform, whereby a target merchant uses multiple intelligent customer service representatives to handle user inquiries from multiple e-commerce platforms. The method includes:

[0005] The system acquires real-time reception data from multiple intelligent customer service systems; it obtains a cross-platform response latency distribution map, a session transfer rate trend, and a knowledge base hit rate heatmap based on the real-time customer service reception data; if the response latency of the intelligent customer service system of the target merchant's target e-commerce platform meets preset latency conditions during a specific period based on the cross-platform response latency distribution map analysis, it captures the service status of the intelligent customer service system corresponding to each e-commerce platform in real time, determines the latency trend of the target e-commerce platform based on the service status, and dynamically adjusts the capacity or updates the instance based on the latency trend; if the analysis shows that the reason for the surge in session transfer rate is at least one of knowledge gaps, user emotions, or complex processes, it executes the corresponding preset linkage event based on the corresponding reason; if the knowledge base hit rate heatmap determines that the first knowledge base bound to the target merchant has blind spots, it captures the blind spot knowledge of the target product and supplements it to the first knowledge base, and verifies whether the hit rate has improved.

[0006] Secondly, this application provides a method for performance tuning of intelligent customer service conversation aggregation from the perspective of an e-commerce platform. The method is applied to an e-commerce conversation management system, which includes an e-commerce conversation management platform, a first e-commerce platform, and a second e-commerce platform; both the first and second e-commerce platforms are federated nodes. The method includes:

[0007] The e-commerce session management platform performs conflict detection on the product activity parameters of the first product. When a knowledge conflict is detected between the first e-commerce platform and the second e-commerce platform, a federal arbitration mechanism is triggered.

[0008] The first e-commerce platform generates a quantum basis selection control sequence and a basis selection sequence. Based on the quantum basis selection control sequence and the basis selection sequence, it encodes the first activity parameter of the first product into a first quantum state and sends the first quantum state to the second e-commerce platform through the e-commerce session management platform. The basis selection sequence is a deterministic basis vector sequence generated by the quantum basis selection control sequence based on the first activity parameter.

[0009] The second e-commerce platform uses a preset shared key to decrypt and obtain the plaintext quantum basis selection control sequence and basis selection sequence. It then extracts the first activity parameter from the basis selection sequence through basis vector alignment and updates the first activity parameter to the dynamic knowledge base of the first product on the second e-commerce platform.

[0010] In some embodiments, the method further includes:

[0011] The first e-commerce platform uses the first quantum private key to perform a quantum signature on the first original product activity parameters of the first product, obtains the first signature data and the encrypted first original product activity parameters, and sends the first signature data and the encrypted first original product activity parameters to the e-commerce session management platform;

[0012] The e-commerce session management platform decomposes the hash value in the first signature data to obtain the first quantum private key, and sends the first quantum private key to the second e-commerce platform through quantum teleportation.

[0013] The second e-commerce platform uses a quantum public key to verify the validity of the first signature data.

[0014] Thirdly, this application provides an e-commerce session management platform with functions corresponding to the intelligent customer service configuration method provided in the first aspect above. These functions can be implemented via hardware or by executing corresponding software through hardware. The hardware or software includes one or more modules corresponding to the aforementioned functions; these modules can be software and / or hardware.

[0015] In some implementations, the e-commerce session management platform includes:

[0016] The input / output module is used to acquire real-time reception data from the multiple intelligent customer service representatives.

[0017] The processing module is used to obtain a cross-platform response delay distribution map, a session transfer rate trend, and a knowledge base hit rate heatmap based on the real-time customer service reception data obtained by the input / output module.

[0018] The display module is used to display the cross-platform response latency distribution map, the session transfer rate trend, and the knowledge base hit rate heatmap.

[0019] The processing module is also used for: if the response latency of the intelligent customer service of the target e-commerce platform of the target merchant meets the preset latency conditions in a specific period of time based on the cross-platform response latency distribution map analysis, then the input / output module captures the service status of the corresponding intelligent customer service of each e-commerce platform in real time, determines the latency trend of the target e-commerce platform based on the service status, and dynamically adjusts the capacity or updates the instance based on the latency trend; and if the analysis shows that the reason for the surge in the session transfer rate is at least one of knowledge gaps, user emotions, or complex processes, then the corresponding preset linkage event is executed according to the corresponding reason; if the knowledge base hit rate heatmap determines that there are blind spots in the first knowledge base bound to the target merchant, then the input / output module captures the blind spot knowledge of the target product and supplements it to the first knowledge base, and verifies whether the hit rate has improved.

[0020] Fourthly, this application provides an intelligent customer service session configuration device, which includes 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 intelligent customer service session aggregation performance adjustment methods provided in the first and second aspects above.

[0021] Fifthly, this application provides a computer-readable storage medium having the function of implementing the intelligent customer service session aggregation performance adjustment method corresponding to the first aspect described above. 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 function, and the modules can be software and / or hardware. Specifically, the computer-readable storage medium stores multiple instructions, which are adapted for loading by a processor to execute the steps in any of the intelligent customer service session aggregation performance adjustment methods provided in the first and second aspects of this application.

[0022] Compared to existing technologies, the solution provided in this application enables target merchants to send and receive messages intuitively and conveniently on a single platform based on a cross-platform session management mechanism, thereby improving session efficiency. On the other hand, it analyzes knowledge base blind spots and identifies platform differences by using a knowledge base hit rate heatmap, captures the service latency trend of a certain platform's intelligent customer service in real time based on a cross-platform response latency distribution map to dynamically balance cross-platform load, and monitors intelligent customer service sessions in real time based on session transfer rate trend analysis. Based on the monitoring data, it executes corresponding linkage strategies (the purpose of which is to optimize the intelligent customer service reception service), thereby achieving rapid perception of the service status of each business channel, automatically triggering intelligent customer service reception service optimization, and tracking optimization measures to form a closed loop. Attached Figure Description

[0023] Figure 1a This is a schematic diagram of an intelligent customer service operation platform provided in this application;

[0024] Figure 1b This is a schematic diagram of a configuration process for the session management platform in this application;

[0025] Figure 2 This is a flowchart illustrating a performance tuning method for the customer service aggregation platform in this application.

[0026] Figure 3a This is a schematic diagram illustrating the configuration of intelligent customer service in this application;

[0027] Figure 3b A schematic diagram illustrating the configuration of the federated dynamic knowledge base in this application;

[0028] Figure 3c This is a schematic diagram illustrating the service latency trend captured by the cross-platform response latency distribution map in this application;

[0029] Figure 3d This is a schematic diagram of the blind spot analysis based on the knowledge base hit rate heatmap in this application;

[0030] Figure 4 This is another schematic diagram of the performance tuning method for the customer service aggregation platform in this application;

[0031] Figure 5 This is another schematic diagram of the performance tuning method for the customer service aggregation platform in this application;

[0032] Figure 6 A schematic diagram of the structure of an e-commerce session management platform in this application;

[0034] Figure 7 This is a schematic diagram of the physical device implementing the intelligent customer service conversation aggregation performance adjustment method in this application. Detailed Implementation

[0035] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects (e.g., in the embodiments of this application, "first buyer" and "second buyer" respectively refer to buyers from different e-commerce platforms), and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0036] This application provides a method, related apparatus, and storage medium for adjusting the performance of intelligent customer service conversation aggregation, which can be used on servers or terminal devices. Specifically, it can be used by merchants in the e-commerce field to manage and optimize their customer service on different e-commerce platforms.

[0037] In some implementations, this solution is applied to, for example... Figure 1a When referring to the intelligent customer service operation platform shown (which is equivalent to the e-commerce conversation management platform, customer service message aggregation platform, and intelligent customer service aggregation platform, and is not distinguished from them), the intelligent customer service operation platform includes a reception end and a management end. The reception end is used to handle the sending and receiving of messages from reception sub-accounts of different e-commerce platforms. The management end is used for configuring various aspects of the intelligent customer service. Figure 1a This includes multiple intelligent customer service systems managed by the target merchants, which connect to multiple sub-accounts on multiple e-commerce platforms.

[0038] The e-commerce session management platform involved in this application can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The e-commerce session management platform involved in this application can be a smartphone, tablet computer, computer, smart speaker, personal digital assistant, etc., but is not limited to these.

[0039] The following combination Figures 1a-7 The technical solutions of the embodiments of this application will be described by way of example.

[0040] like Figure 1b The diagram illustrates a configuration flow for a session management platform. This platform aggregates user queries from across platforms for session aggregation and centralized session management. The platform includes a platform feature recognition engine, a spatiotemporal knowledge modeling engine compatible with multiple e-commerce platforms, and an edge computing node cluster.

[0041] An edge computing node cluster comprises 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 user queries from different e-commerce platforms. The central computing nodes in multiple edge computing centers can achieve dynamic load balancing through preset traffic distribution strategies.

[0042] The platform feature recognition engine is configured to identify the platform features of the e-commerce platform to which the user's inquiry belongs. In other words, when the intelligent customer service of a sub-account serves a user of a certain e-commerce platform, it can identify the platform features of that e-commerce platform.

[0043] The spatiotemporal knowledge modeling engine is configured as a federated dynamic knowledge base integrating multiple e-commerce platforms. Based on platform feature recognition engines, it performs federated queries on the dynamic knowledge base to obtain platform-specific knowledge that is spatiotemporally related to user queries, serving as the federated query center. In essence, this spatiotemporal knowledge modeling engine acts as the central unit for user interaction execution within the customer service system, adapting to cross-platform knowledge. It collaborates with the intelligent customer service system and the federated dynamic knowledge base to achieve session aggregation and differentiated responses.

[0044] The platform feature recognition engine and spatiotemporal knowledge modeling engine support the intelligent customer service system. When using a federated node knowledge base from multiple e-commerce platforms, the spatiotemporal knowledge modeling engine simultaneously possesses knowledge federation aggregation and spatiotemporal pre-simulation functions.

[0045] The collaboration process among the three parties can be referenced. Figure 1b It includes the following steps:

[0046] 101. The edge computing node cluster receives data streams from user inquiries from different e-commerce platforms and forwards the data streams to the platform feature recognition engine according to a preset diversion strategy.

[0047] Among them, the preset routing strategy is used to route messages for user inquiries from different e-commerce platforms. The preset routing strategy includes platform-level routing strategy and merchant-level reception routing strategy.

[0048] 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 user query belongs from the data stream, and sends the platform features to the spatiotemporal knowledge modeling engine.

[0049] 103. The spatiotemporal knowledge modeling engine, based on the platform features and the dynamic knowledge base of user query federated queries, acquires platform-based knowledge that is spatiotemporally related to user queries and sends it to the intelligent customer service.

[0050] This platform-based knowledge includes product activity times (such as Taobao pre-sales vs. JD.com flash sales) and regional policies (such as Shanghai duty-free warehouses vs. Xinjiang free shipping zones).

[0051] 104. The intelligent customer service system, based on the platform-based knowledge received from the spatiotemporal knowledge modeling engine, generates a style-adapted answer that matches the platform characteristics according to the preset differentiated response strategy and the platform characteristics of the e-commerce platform to which the user's inquiry belongs, and returns it to the user through the edge computing node cluster.

[0052] After generating differentiated style answers that conform to the platform's characteristics based on the platform's knowledge, the answers are displayed in the corresponding user's conversation interface.

[0053] In some embodiments, the spatiotemporal knowledge modeling engine is further configured to:

[0054] Real-time acquisition of knowledge matching data, response latency data, and session transfer data (all three are derived from historical response evaluation data of the intelligent customer service of each sub-account); among them, knowledge matching data measures the relevance of user inquiries and style responses returned to users on different platforms to platform-based knowledge in a dynamic knowledge base; response latency data is the time taken for the intelligent customer service of each sub-account to respond to inquiries from different users; session transfer data is statistical data on the transition from intelligent customer service to human agent when the intelligent customer service of a sub-account handles inquiries from users on multiple e-commerce platforms.

[0055] A knowledge base hit rate heatmap is generated based on knowledge matching data; a cross-platform response latency distribution map is generated based on response latency data; and a session transfer rate trend is generated based on session transfer data. A first monitoring mechanism is constructed based on the knowledge base hit rate heatmap, the cross-platform response latency distribution map, and the session transfer data. The first monitoring mechanism includes: analyzing dynamic knowledge base blind spots and identifying platform differences based on the knowledge base hit rate heatmap; capturing the service latency trend of the intelligent customer service associated with the target platform in real time based on the cross-platform response latency distribution map and dynamically balancing cross-platform load based on the service latency trend; and analyzing the reasons for session transfer based on the session transfer rate trend and triggering preset linkage events.

[0056] As can be seen, the above platform-level configuration provides support for the personalized configuration of the target merchant's reception items across the entire platform in subsequent embodiments. Through this platform-level configuration, when actually receiving user inquiries, the spatiotemporal knowledge modeling engine transforms general knowledge in the federated knowledge base into proprietary platform-specific knowledge that conforms to specific platform characteristics and spatiotemporal scenarios in real time. This enables intelligent customer service to generate native style responses that conform to platform characteristics, ensures that federated queries comply with regional policy boundaries, achieves cross-platform consistency of the dynamic knowledge base to dynamically adapt to promotional cycles, and provides a scenario-based interactive experience.

[0057] Figure 1bThe 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 configurations can be made for each individual merchant. Taking the configuration of intelligent customer service for multiple e-commerce platform sub-accounts for a target merchant as an example, the session configuration of intelligent customer service in the e-commerce session management platform includes the following method for aggregating cross-platform intelligent customer service sessions in this embodiment:

[0058] 201. Create a reception project to aggregate multiple sub-accounts of target merchants on different e-commerce platforms.

[0059] This reception project binds the target merchant's sub-accounts on different e-commerce platforms to the management account. The sub-accounts are the reception accounts used by the target merchant to open stores on each e-commerce platform. The reception project is an independent unit, including project members (i.e., human customer service, without distinction between the two), reception sub-accounts on different e-commerce platforms, intelligent customer service, reception triage strategy, and data dashboard.

[0060] 202. Based on the platform characteristics of the e-commerce platform to which the sub-account belongs and the product parameters of the products already listed, create intelligent customer service that matches the platform characteristics and bind it to each sub-account accordingly, and associate the intelligent customer service of each sub-account with the dynamic knowledge base of the products already listed.

[0061] Due to differences in the associated dynamic knowledge base or the model structure used, the capability parameters of each intelligent customer service representative will vary. Some are general-purpose, some are specialized, and some are better at handling certain types of inquiries; there are no limitations on this. Platform characteristics can be identified by the platform feature recognition engine when the intelligent customer service of a sub-account serves users of a certain e-commerce platform. For example, it can be done by analyzing the native messages in user inquiries, such as parsing short videos from Douyin stores, analyzing the correlation of Taobao orders, or analyzing the group-buying logic of Pinduoduo, in order to identify platform characteristics.

[0062] In the reception project, each sub-account's intelligent customer service is associated with at least one dynamic knowledge base of the listed products, as well as the visual execution logic for responding to user inquiries. This dynamic knowledge base is used by the visual execution logic. The intelligent customer service is bound to the reception project, and human customer service representatives who are project members can see the dynamic knowledge base of their own project on the knowledge management function page, and can update and review the relevant knowledge of the intelligent customer service to which the project belongs.

[0063] In this application, the dynamic knowledge base used by the intelligent customer service configured for each merchant's sub-account can be a regular knowledge base (pre-built by the conversation management platform or imported by the merchant) or a federated node knowledge base; there are no limitations on this. 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 customized by this conversation management platform specifically for various merchants from cross-platform distribution, and the management of various types of knowledge in the dynamic knowledge base can be entered and updated by the merchants themselves; this application does not limit this. Figure 3a As shown, Figure 3a Fig1 represents the local configuration of this session management platform. It collects various product parameters and business parameters from different e-commerce platforms to build a reusable sub-dynamic knowledge base. Each sub-dynamic knowledge base is set as a federated node to facilitate federated queries when responding to user inquiries. Figure 3a Fig2 illustrates a scenario where federated queries are performed based on dynamic knowledge bases from multiple e-commerce platforms. Each e-commerce platform sets its dynamic knowledge bases, including various product parameters and business parameters, as federated nodes, facilitating cross-platform federated queries without requiring data sharing between the e-commerce platforms.

[0064] A dynamic knowledge base can be a dynamic knowledge base for federated nodes, such as... Figure 3b As shown, each e-commerce platform is configured with an independent dynamic knowledge base. Each platform's dynamic knowledge base includes question templates, answers, and timestamps, protected by local differential privacy. A unified query language for cross-platform queries can be pre-configured. Upon receiving a user query, each e-commerce platform, acting as a federation node, generates a question-answer pair vector based on its own dynamic knowledge base. This vector is then transmitted via quantum teleportation to the session management platform, which acts as the federation center. The session management platform then performs confidence-weighted aggregation of the federated query results in encrypted space and returns the aggregated result to the user.

[0065] It is evident that, on the one hand, by configuring a federated dynamic knowledge base for each intelligent customer service representative, knowledge sharing can be achieved under encrypted conditions, reducing cross-platform query latency, improving privacy protection, and increasing the accuracy of searching for best-selling products, while also reducing cross-platform knowledge blind spots. On the other hand, target merchants can update their dynamic knowledge base on the current session management platform, ensuring consistency in the key characteristics of their products across all e-commerce platforms where they have already listed their products. While the same product may vary across different e-commerce platforms due to platform characteristics, the key characteristics of the product remain consistent across platforms; the intelligent customer service will only output differentiated responses when dealing with users from different e-commerce platforms.

[0066] 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 the icons of each sub-account in real time according to the session update time, and the second area displays the session interface of the selected first sub-account.

[0067] The conversation management interface also includes a third area, which displays sub-account reception data (which can also be displayed in the second area). For example, sub-account reception data shows data scraped from platforms like Qianniu and Feige, primarily including reception volume, average response time, first response time, and satisfaction rate. When a single line is insufficient, an expand button appears; by default, only one line of data is displayed. Alternatively, this third area can display dynamic product data for the inquired product across all e-commerce platforms.

[0068] For reference Figure 1a The first area displays a list of users served by all sub-accounts available to project members (i.e., human customer service representatives). Clicking on a sub-account online will switch to that sub-account's chat window. (See also the chat windows for when the target merchant serves users from multiple platforms.) Figure 1a )exhibit.

[0069] 204. Configure reception and diversion strategies for the intelligent customer service of each sub-account based on platform characteristics and preset question types, and configure corresponding differentiated response strategies for the intelligent customer service of each sub-account based on the platform characteristics of the e-commerce platform.

[0070] The traffic routing strategy refers to an intelligent routing approach for user inquiries received by each sub-account from different e-commerce platforms, used to achieve load balancing. This is because the target merchant's existing products, product operation plans, and transaction data vary across different e-commerce platforms. Therefore, the target merchant will differentiate the traffic flow for each sub-account on different e-commerce platforms. For example, if product A has high popularity and transaction volume on platform 1, then more weight needs to be allocated in the traffic routing strategy to support the traffic received.

[0071] Configure a reception traffic distribution strategy for a specific reception project, including how reception sub-accounts are distributed, which project members are handled by humans, which project members are responsible for AI bots, and how traffic is allocated among multiple AI bots. Specifically, you can create a strategy name, specify the project to which it belongs, and select the traffic distribution strategy.

[0072] In this embodiment, a conversation management platform aggregates the sub-accounts of target merchants across different e-commerce platforms. Each sub-account is configured with intelligent customer service tailored to platform characteristics, along with traffic routing strategies, differentiated response strategies, and a primary monitoring mechanism. A human referral prediction model is also embedded in the intelligent customer service of each sub-account. Therefore, on one hand, the cross-platform conversation management mechanism, by aggregating customer service sub-accounts from multiple e-commerce platforms (i.e., business channels) on a single platform and binding them to intelligent customer service based on a dynamic knowledge base and visualized processes, allows target merchants to send and receive messages intuitively and conveniently on a single platform, improving conversation efficiency.

[0073] In some implementations, a human call transfer prediction model can also be embedded in the intelligent customer service of each sub-account.

[0074] The human referral prediction model is used to predict the probability of a sub-account being manually transferred when handling user issues, based on response latency, knowledge matching degree, and user sentiment value, in order to proactively detect the risks of intelligent customer service. The human referral prediction model employs a multimodal fusion model. In this model, cross-modal attention (emphasizing the interaction between user sentiment value and knowledge matching degree) is designed for the three risk factors of response latency, knowledge matching degree, and user sentiment value. A time gating mechanism is designed to identify unresolved patterns over n rounds, ultimately outputting the probability of a human referral.

[0075] Response latency is the time it takes for the intelligent customer service to respond to a user's inquiry (e.g., 12 seconds is a timeout response).

[0076] Knowledge matching degree refers to the relevance of user inquiries and style responses mapped to platform-based knowledge in a dynamic knowledge base within an intelligent customer service system. It is typically calculated using three dimensions: semantic level (e.g., user inquiry intent), entity level (e.g., product parameters), and scenario level (e.g., after-sales policies, activity rules). For example, knowledge matching is performed on a user inquiry, "How do I return or exchange a newly purchased mobile phone with scratches on the screen?", to generate the final answer.

[0077] User sentiment scores are derived from analyzing conversation records and are typically negative sentiment scores (e.g., complaints).

[0078] If the response delay exceeds a certain threshold, the knowledge matching degree falls below a certain threshold, or the user's sentiment score falls below a certain threshold, these are all considered risk factors for triggering a human transfer. These three risk factors can be used to predict the probability of a human transfer when the intelligent customer service system handles user inquiries.

[0079] The interaction between these three risk factors can be obtained using a sample weighting method. For example, high sentiment scores can be weighted × m, knowledge matching scores < 0.4 can be weighted × n, and response delays > 10s can be weighted × p, where m > n > p. For example, a user inquires about "orders paid for but not shipped." The calculated response delay is 8.2s, knowledge matching score is 0.65, and user sentiment score is [0.72, 0.68, 0.75] (text + voice + visual). The historical reception characteristics of the intelligent customer service for the user are found to be [0, 0, 1, 0, 0] (1 out of the last 5 times was transferred to a human), question complexity is 0.7, number of conversation rounds is 3, and system load is 0.68. This historical reception characteristic is embedded into the human transfer prediction model. Because cross-modal attention is added to the two risk factors of knowledge matching score and user sentiment score, the final output probability is 0.87. Since 0.87 is greater than the preset threshold, the decision result is "immediately transfer to human customer service."

[0080] It is evident that by embedding a human transfer prediction model, the complex interactions between response delay, knowledge matching degree, and user emotions can be captured in real time during sub-account reception. This can achieve high-precision prediction while ensuring real-time response, reduce the human error rate, and improve the utilization rate of intelligent customer service.

[0081] In this embodiment, a first monitoring mechanism is configured on the e-commerce session management platform. This mechanism analyzes knowledge base blind spots and identifies platform differences based on a knowledge base hit rate heatmap. It also captures the service latency trend of a platform's intelligent customer service in real time based on a cross-platform response latency distribution map to dynamically balance cross-platform load. Furthermore, it monitors intelligent customer service sessions in real time by analyzing session transfer rate trends. Based on the monitoring data, corresponding linkage strategies are executed (the purpose of which is to optimize the intelligent customer service's reception service). This enables rapid perception of the service status of each business channel, automatically triggers intelligent customer service reception service optimization, and tracks optimization measures to form a closed loop. On the other hand, by embedding a manual transfer prediction model into the intelligent customer service, the manual transfer rate can be reduced, balancing session efficiency and user experience.

[0082] Optionally, in some embodiments of this application, the method further includes configuring a second monitoring mechanism and a verification mechanism. The second monitoring mechanism includes obtaining policy monitoring components from the announcement pages of each platform, monitoring product editing logs on each platform, and triggering a knowledge base update task when key policy fields change; the verification mechanism includes consistency checks of the same product's basic parameters across platforms, full release of new knowledge points after verification in some sessions, and prompting manual intervention when platform rules conflict.

[0083] Optionally, in some embodiments of this application, considering the different message formats of different e-commerce platforms, in order to reduce the need for zero-conversion processing of message formats for different business platforms, this application further includes:

[0084] Each edge computing node in the edge computing node cluster is configured with a protocol knowledge base and a native message format parser. The protocol knowledge base is used to store protocol metadata of the communication protocol characteristics of each e-commerce platform. The native message format parser is used to extract semantics from native messages (such as user queries) based on the protocol knowledge base.

[0085] The intelligent customer service is configured to generate a style response that matches the platform characteristics, i.e., a response that conforms to the original format of the platform to which the message originates, based on the product parameters obtained from the dynamic knowledge base received from the spatiotemporal knowledge modeling engine and the product parameters that match the original message.

[0086] For example, setting up platform-specific script libraries: Douyin (designing short video content + live stream scripts), Taobao (professional customer service scripts), and Pinduoduo (group buying progress + limited-time discounts). And adapting platform-specific interactive elements: automatically generating button / card messages that conform to platform specifications (such as the Douyin live stream jump button). Ensuring consistency in core information when answering inquiries about the same product on different e-commerce platforms.

[0087] In some implementations, the protocol metadata includes the following three types of storage structures:

[0088] The first type of storage structure includes message structure definition, signature rules, and session mechanism, such as the Taobao platform.

[0089] The second type of storage structure includes protobuf description files and live streaming effect mappings. For example, on the Douyin platform, protobuf is a structured data serialization protocol, and the protobuf description file describes the message characteristics of a specific platform. Based on this protobuf description file and live streaming effect mapping, effect parameters can be obtained, effect tools can be called to render dynamic effects, and message cards can be synthesized and displayed to the user, enabling reuse across multiple scenarios.

[0090] The third type of storage structure includes group-buying rules and real-time activity strategies, such as those found on the Pinduoduo platform.

[0091] As can be seen, by setting up different message storage structures for different e-commerce platforms according to their characteristics in this session management platform, zero-conversion processing of message formats for different business platforms can be achieved, while ensuring cross-platform compatibility and low-latency transmission.

[0092] Optionally, in some embodiments of this application, considering that some e-commerce platforms may prohibit reverse engineering, unified knowledge modeling and the deployment of dedicated processing containers can be performed to achieve zero-protocol parsing. Specifically, the dynamic knowledge base originates from the knowledge bases of different e-commerce platforms, and the knowledge bases of different e-commerce platforms are all federated nodes; the method further includes:

[0093] Establish cross-supply chain consulting and cross-platform after-sales processes based on federated nodes;

[0094] The unified knowledge modeling engine obtains encrypted knowledge vectors from multiple e-commerce platforms, fuses the knowledge vectors using homomorphic encryption, and updates the global model parameters based on a federated averaging algorithm; the knowledge vectors include short video feature vectors and product parameter vectors.

[0095] Each e-commerce platform is assigned a dedicated processing container, which is used to call the SDK of its respective e-commerce platform to parse user messages. These dedicated processing containers are edge computing units.

[0096] The smart contract based on the blockchain consensus mechanism defines a data sharing mechanism between platforms. The data sharing mechanism allows the first e-commerce platform to share its activity policies with the second e-commerce platform for limited access, and the shared data is homomorphically encrypted.

[0097] In some implementations, a federated node topology can be adopted, where each e-commerce platform and e-commerce session management platform serves as a federated node, and shared protocols are configured among the federated nodes. Each e-commerce platform acts as an independent node, maintaining its local dynamic knowledge base. Data sharing rules are defined based on smart contracts; for example, Pinduoduo's activity rules can be accessed to a limited extent by Douyin nodes. This federated node topology can also employ a differential synchronization algorithm to implement a knowledge update synchronization mechanism, broadcasting knowledge increments to the federated network hourly, reducing data transmission volume by 90%. Correspondingly, a federated knowledge graph is constructed based on the dynamic knowledge bases of each e-commerce platform to achieve cross-platform alignment, specifically by combining semantic vectors with the similarity of product identifiers for matching. Therefore, this approach enables zero-protocol parsing of user messages from different e-commerce platforms without reverse engineering the business platforms, while maintaining the integrity of the user messages.

[0098] like Figure 2 As shown, the intelligent customer service conversation aggregation performance tuning method in this application is applied to an e-commerce conversation management platform. The target merchant uses multiple intelligent customer service representatives on the e-commerce conversation management platform to handle user inquiries from multiple e-commerce platforms. The target merchant can provide reception services using a hybrid approach of human customer service and intelligent customer service on the e-commerce conversation management platform. Embodiments of this application include:

[0099] 201. Obtain real-time reception data from the multiple intelligent customer service systems.

[0100] 202. Based on the real-time customer service reception data, obtain a cross-platform response delay distribution map, a session transfer rate trend, and a knowledge base hit rate heatmap.

[0101] 203. If the response latency of the intelligent customer service of the target merchant's target e-commerce platform meets the preset latency conditions during a specific period based on the cross-platform response latency distribution map analysis, then the service status of the intelligent customer service of each e-commerce platform is captured in real time, the latency trend of the target e-commerce platform is determined based on the service status, and the capacity is dynamically adjusted or the instance is updated based on the latency trend.

[0102] Example 1: If the response latency of a target merchant on Pinduoduo suddenly increases from 200ms to 700ms (preset threshold 500ms), an automatic check of the CPU utilization of the ECS instance on that node will be triggered. Then, if CPU > 70%, vertical scaling will be performed; if CPU < 50%, horizontal scaling will be performed (adding 2 container instances).

[0103] Example 2: During a live-streaming sales event, the response latency of Douyin Store's intelligent customer service surged. On May 15, 2025, at 21:05, it was observed that the latency of Douyin Store P99 increased from 120ms to 620ms. Figure 3c As shown, the system captures the service latency trend of the intelligent customer service associated with the target platform in real time based on the cross-platform response latency distribution map and dynamically balances the cross-platform load according to the service latency trend, such as triggering the expansion of intelligent customer service nodes of Douyin sub-accounts.

[0104] 204. If the analysis shows that the surge in session switching rate is due to at least one of the following: lack of knowledge, user emotions, or complex processes, then execute the corresponding preset linkage event according to the corresponding reason.

[0105] In some implementations, the dynamic knowledge base originates from the knowledge bases of different e-commerce platforms, and the knowledge bases of different e-commerce platforms are all federated nodes; if the reason for the surge in session switching rate is analyzed to be at least one of knowledge gaps, user emotions, or complex processes, then a corresponding preset linkage event is executed according to the corresponding reason, including:

[0106] If the surge in session switching rate is attributed to knowledge gaps, a knowledge base update is triggered, including one of the following: obtaining the latest activity parameters of the first product on the first e-commerce platform and linking these parameters with the first product on the first e-commerce platform based on a product linkage mechanism; or obtaining the inventory status of the first product on the first e-commerce platform and linking this inventory status with the live-streaming activity information of the second product on the second e-commerce platform based on a product linkage mechanism. The product linkage mechanism is a differential synchronization algorithm that periodically obtains the product knowledge increments from each federated node and broadcasts these increments to the federated network.

[0107] It is evident that the product-based linkage mechanism can overcome the limitations of the knowledge base and break down data silos across platforms.

[0108] If the analysis shows that the surge in conversation switching rate is due to process complexity, then add a quick operation button to the visualized process for the product association dimension of the first product; if the analysis shows that the surge in conversation switching rate is due to negative user emotions, then switch to reassuring language and assign specific customer service representatives with ability parameters higher than the preset threshold to intervene.

[0109] 205. If the first knowledge base bound to the target merchant is determined to have blind spots based on the knowledge base hit rate heatmap, then the blind spot knowledge of the target product is captured and added to the first knowledge base, and the hit rate is verified to improve.

[0110] Example 1, such as Figure 3dAs shown, if the knowledge hit rate is analyzed based on the knowledge base hit rate heatmap, and it is determined that there are blind spots in the first knowledge base bound to the current target merchant, then the blind spot knowledge of the target product is captured and added to the knowledge base. Finally, it is verified whether the hit rate has improved. The core indicator is the hit rate of high-frequency questions.

[0111] For scenarios with low hit rates in areas where knowledge gaps exist:

[0112] For example, the hit rate for "ingredient safety" questions in the mother and baby category is only 43%. This is because the knowledge base only stores mandatory national standards and lacks information on brand-specific ingredients. However, users actually ask questions about specific ingredients such as "lactoferrin content," leading to the low hit rate. Possible solutions include: extracting parameter tables from product detail pages, adding brand-specific testing reports, and building an ingredient knowledge graph.

[0113] Example 2: When the hit rate for the "mobile phone" category is found to be below 60%, data from the product details page is automatically retrieved.

[0114] The lower hit rate due to platform differences is a concern. For example, the hit rate for "video tutorials" on Douyin is 78%, while the hit rate for similar questions on Taobao is only 32%. This is because Douyin users are accustomed to solving problems by watching short videos, and the existing knowledge base only stores text descriptions and has not integrated short video resources.

[0115] Example 3: Regarding supplementary knowledge, QA knowledge can be automatically extracted from the chat history between human customer service and buyers on a regular basis. The QA knowledge is compared with the existing knowledge in the dynamic knowledge base. If it is new knowledge, it will be sent to human customer service for review. Once the review is approved, it will be entered into the database with one click.

[0116] The following formula can be used to actively detect blind spots: If the feature distance between the problem vector 𝑞 and the nearest neighbor knowledge k in the knowledge base is less than the threshold 𝜃, it indicates that knowledge in this domain needs to be supplemented.

[0117] As can be seen, by analyzing the above three scenarios, the reception status of the target merchant's intelligent customer service can be captured in a timely manner and corresponding linkage events can be executed. On the one hand, by analyzing the cross-platform response latency distribution map, dynamic balancing of cross-platform load can be achieved; on the other hand, by analyzing the session transfer rate trend, service anomalies in various business channels can be quickly detected, anomalies can be automatically located, and intelligent customer service reception service optimization, knowledge base updates and process optimization, and tracking optimization measures can be triggered to form a closed loop, thereby improving session efficiency.

[0118] Optionally, in some embodiments of this application, a platform-differentiated adaptation strategy can be set to improve the cross-platform coverage of the knowledge base. Specifically, corresponding knowledge points can be dynamically generated based on platform characteristics, or the language style of knowledge points from a platform can be transferred based on platform characteristics (i.e., dynamically generating corresponding knowledge presentation methods based on the platform characteristics of the e-commerce platform and transferring the language style) to supplement knowledge gaps. See Table 2 below:

[0119]

[0120] Table 2

[0121] Optionally, in some embodiments of this application, each e-commerce platform and the e-commerce session management platform are federated nodes, responding to user inquiries through federated queries of a federated dynamic knowledge graph. Considering that in federated query scenarios, each e-commerce platform needs to perform signature verification, resulting in a large computational load for signature data verification, especially in high-concurrency inquiry scenarios, this application, in order to ensure that the cross-platform customer service system can obtain secure and consistent product activity parameters in real time and resist collusion attacks, achieves a single signature by aggregating signature data. Specifically, the method further includes steps 301-304:

[0122] 301. Each e-commerce platform uses a quantum private key to generate signature data Si for the activity parameters of the first product locally. For example, Si = QCpriv(di,si,ti).

[0123] 302. The e-commerce session management platform receives activity parameter data pairs of the first product from multiple e-commerce platforms.

[0124] The activity data for each e-commerce platform includes signature data and encrypted raw activity parameters.

[0125] 303. The e-commerce session management platform uses the Shor algorithm to decompose the hash value in each signature data. After confirming that the original product activity parameters have not been tampered with by the hash value, it aggregates the signature data from each e-commerce platform to obtain aggregated signature data.

[0126] For example, aggregate signature data S agg =S JD.com ⊕S Taobao ⊕S Pinduoduo. Taking a cross-platform iPhone 15 inquiry between JD.com and Pinduoduo as an example. A user inquires on JD.com whether the iPhone 15 supports trade-ins. JD.com customer service needs to verify whether there is a conflict with Pinduoduo's promotional rules, and then aggregate the resulting S... agg Include:

[0127] JD.com's trade-in subsidy rate is 0.15%, while Pinduoduo's minimum trade-in price is ¥3500.

[0128] 304. The e-commerce session management platform distributes aggregated signature data to various e-commerce platforms via quantum teleportation. For example, the aforementioned S... agg Distributed to JD.com's customer service system, and used in the e-commerce conversation management platform for S agg The verification was performed, and the verification result is as follows:

[0129] Local rules: JD.com subsidy rate 0.12, minimum buyback price ¥3200

[0130] Global rules: JD.com subsidy rate 0.15, Pinduoduo buyback price ¥3500

[0131] The semantic difference between local and global rules was calculated to be d = 0.08 < 0.15 (preset threshold). Therefore, this S... agg Serving buyers. For the intelligent customer service of the e-commerce conversation management platform, it can respond to the buyer it serves: "Current maximum subsidy ¥450 (0.15 from JD.com + ¥3500 buyback from Taobao)".

[0132] For example, firstly, the first and second signature data of a first product are obtained from a first e-commerce platform and a second e-commerce platform, respectively. Then, the hash values ​​in the first and second signature data are decomposed using Shor's algorithm to confirm that the original product activity parameters have not been tampered with. Finally, the signature data from the first and second e-commerce platforms are merged into a joint signature data and distributed to both e-commerce platforms.

[0133] It is evident that by aggregating and distributing signatures, it is possible to ensure that cross-platform customer service systems can obtain secure and consistent product activity parameters in real time, ensuring that all participants reach a consensus on global rules and resisting collusion attacks; by aggregating signature data to achieve a single signature, the amount of verification computation is reduced.

[0134] Optionally, in some embodiments of this application, when querying best-selling products based on federated queries, there may be instances where the signature data and activity parameters of the best-selling product on different e-commerce platforms do not conflict. In such cases, real-time consistency maintenance of cross-platform activity parameters can be achieved through the collaboration of quantum signature aggregation, conflict detection, and federated arbitration. Specifically, as... Figure 5 As shown, this application also includes:

[0135] 401. The e-commerce session management platform performs conflict detection on at least one of the activity parameters and signature data of the first product.

[0136] The purpose of conflict detection is to identify inconsistencies between rules submitted across multiple platforms, such as price discrepancies, time conflicts, and inventory discrepancies. This conflict detection can be triggered when the e-commerce session management platform receives an inquiry from a user on any e-commerce platform regarding the top-selling product, or it can be triggered by the e-commerce session management platform based on high-frequency questions from historical customer service data across different e-commerce platforms. This application does not limit this to either approach.

[0137] In some implementations, the following formula can be used for comparison:

[0138] Δd=∣di−dj∣, price difference threshold εd =0.05 (5%)

[0139] Δs=∣si−sj∣, inventory difference threshold εs =20%

[0140] Δt = max(ti,tj) − min(ti,tj), time difference threshold εt = 24h

[0141] 302. If the difference between any activity parameters exceeds the threshold and the semantic similarity is less than the preset similarity, or if the aggregated signature verification fails, the e-commerce session management platform determines that there is a knowledge conflict between the e-commerce platforms.

[0142] The Quantum NLP engine is used to parse the activity rule text and calculate semantic similarity: Sim=1−21(JSD(Pi,Pj)+JSD(Pj,Pi)), where Pi and Pj are the rule probability distributions of e-commerce platform i and j, respectively.

[0143] If Sim < 0.85 or the difference of any parameter exceeds the threshold, the federal arbitration mechanism is triggered.

[0144] In some implementations, the present application may employ the following quantum conflict detection algorithm:

[0145] H conflict =α⋅Δd⊗σx+β⋅Δs⊗σy+γ⋅Δt⊗σz

[0146] P conflict =∣⟨ψ∣H conflict |ψ>|2, when Pconflict A conflict is determined when the value is greater than 0.9.

[0147] For example, JD.com and Taobao have d=0.85 vs d=0.88 → Δd=0.03<0.05 (no conflict); JD.com and Pinduoduo have t=72h vs t=96h → Δt=24h (triggered conflict).

[0148] 403. The e-commerce session management platform triggers a federal arbitration mechanism to quantify the economic losses of conflicts between e-commerce platforms, distributes conflict arbitration requests to each e-commerce platform through quantum keys, and obtains corresponding activity parameters from each e-commerce platform.

[0149] The conflict arbitration request includes the conflict type, the activity parameters with knowledge conflict, the difference in activity parameters, and the degree of semantic difference. For example:

[0150] Conflict type: "Time conflict"

[0151] Platform A rule: {"t":72h, "d":0.85},

[0152] Platform B rule: {"t":96h, "d":0.75},

[0153] It can be seen that the time difference between rules A and B on platform A is 24 hours, and the semantic difference is 0.12.

[0154] 404. The e-commerce session management platform weights the activity parameters of each e-commerce platform to obtain weighted activity parameters, and allocates the inventory of the first product on each e-commerce platform according to the dynamic weight of the e-commerce platform.

[0155] For ease of understanding, let's take the conflict of activities of the X brand mobile phone 12 as an example.

[0156] 1. The current status of the activity parameters is as follows:

[0157] JD.com: ¥500 off + free AirPods (500 units in stock)

[0158] Taobao: 88VIP members get an extra discount, price is ¥470 (300 units in stock)

[0159] Pinduoduo: ¥450 subsidy (1000 units in stock)

[0160] 2. The activity conflict detection process includes:

[0161] First, compare the price difference and the inventory difference:

[0162] Price difference: JD.com vs. Taobao = ¥20 (Δ_d = 0.023 < 0.05)

[0163] Inventory difference: JD.com vs Pinduoduo = 50% (Δ_s = 0.5 > 0.2 → conflict)

[0164] Secondly, semantic analysis of the keywords is performed:

[0165] JD.com rule keywords: ["direct price reduction", "free gifts"]

[0166] Pinduoduo's rule keywords: ["subsidies", "no free gifts"]

[0167] The calculated semantic similarity Sim = 0.72 < 0.85 (preset threshold)

[0168] Finally, based on the above comparison of price differences and inventory differences, and semantic analysis of keywords, the results of the federal arbitration regarding the conflict of interest in the X brand mobile phone 12 are as follows:

[0169] Using the weighted average method: Final price = 0.4 × 500 + 0.3 × 480 + 0.3 × 450 = 479 yuan

[0170] Inventory allocation: can be allocated according to the weight of e-commerce platforms, such as 300 units for JD.com and 700 units for Pinduoduo.

[0171] In some implementations, in scenarios involving knowledge conflicts, after triggering the federal arbitration mechanism, real-time consistency of cross-platform activity parameters can be achieved through cross-platform synchronization of dynamic knowledge bases. Specifically, the e-commerce platforms with knowledge conflicts include a first e-commerce platform and a second e-commerce platform. After triggering the federal arbitration mechanism, this application further includes:

[0172] 405. The first e-commerce platform generates a quantum basis selection control sequence and a basis selection sequence, and encodes the first activity parameter of the first product into a first quantum state based on the quantum basis selection control sequence and the basis selection sequence.

[0173] The quantum basis selection control sequence, generated by the first e-commerce platform, is a random bit string consisting of 0s and 1s used to determine the basis vector selection rules for quantum state encoding. The purpose of this quantum basis selection control sequence is to: achieve non-orthogonal encoding of quantum states through random basis selection, thus intercepting replay attacks; and for the first e-commerce platform to map product activity parameters (such as discount rates and activity rules) to the first quantum state, generating a new quantum basis selection control sequence in each round of federated training to prevent replay attacks (e.g., preventing attackers from reusing historical quantum states of product activity parameters). Examples are shown in Table 1 below.

[0174]

[0175] Table 1

[0176] The basis selection sequence is a deterministic basis vector sequence generated based on the first activity parameter (i.e., the original data of the activity parameter, generated by a quantum random number generator). The purpose of this basis selection sequence is to encode the product activity parameters (such as discount rates and superposition conditions) into the first quantum state. The randomness of this basis selection sequence prevents eavesdroppers from cracking the rules through statistical analysis, thus achieving anti-interference. The basis selection sequence can be deduced from the quantum state measurement results, thereby verifying data integrity.

[0177] For example, Platform A synchronizes the "iPhoPX Limited-Time Offer" activity rules to Platform B. The activity rules are: a direct discount of ¥500, free AirPods, and a limited time offer of 3 days.

[0178] First, perform quantum parameter mapping:

[0179] Quantum basis selection control sequence b_A = [0,1,0,1,0](@ref)⊗ Product ID hash value

[0180] Basis selection sequence x_A

[0181] x_A = Hadamard(0)⊗CNOT(1)⊗Hadamard(0)⊗...⊗ProductHash

[0182] Then, the quantum state is generated: |ψ>=∑b_A^i|x_A^i>⊗|iPhone15_info>

[0183] Where b_A[i] is the basis selection control sequence for the i-th position, b_A[i]∈0,1; Hadamard(0) is the horizontal basis, and CNOT(1) is the diagonal basis.

[0184] 406. The first e-commerce platform sends the first quantum state to the second e-commerce platform through the e-commerce session management platform.

[0185] 407. The second e-commerce platform uses a preset shared key to decrypt the first quantum state and obtain the plaintext quantum basis selection control sequence and basis selection sequence.

[0186] 408. The second e-commerce platform extracts the first activity parameter from the basis selection sequence through basis vector comparison, and updates the first activity parameter to the dynamic knowledge base of the first product on the second e-commerce platform.

[0187] Understandably, considering that the calculation rules for activity parameters on different e-commerce platforms may be incompatible, a unified rule-based question bank can be constructed, and the quantum ontology alignment algorithm can be used to eliminate cross-platform differences in activity parameters.

[0188] It is evident that, on the one hand, by aggregating quantum signature data from different e-commerce platforms, a single signature can be generated, thereby reducing cross-platform signature latency. On the other hand, by combining quantum activity parameter comparison with semantic analysis, hybrid conflict detection can be achieved, and in conjunction with a federated arbitration mechanism, economic losses can be quantified, the authenticity of activity parameters can be guaranteed, and the activity prices and inventory of each e-commerce platform can be balanced to maximize profits. After conflict detection, by aligning the dynamic knowledge bases of the same product on each e-commerce platform, quantum-safe federated synchronization of the cross-platform dynamic knowledge base can be achieved, which can adapt to the iterative situation of product knowledge or product activity parameters and achieve network-wide consistency. That is, the embodiments of this application achieve quantum-safe federated synchronization of product parameters and activity parameters through the physical characteristics of quantum basis vectors, transforming the parameter aggregation process of traditional federated learning into a basis vector encoding and decoding process of quantum states, thereby achieving consistency of the cross-platform dynamic knowledge base through basis vector comparison.

[0189] In some implementations, security enhancement mechanisms can be employed during the knowledge base synchronization process. The method further includes: the first e-commerce platform adding random redundant basis vectors to the basis selection sequence (to defend against basis vector analysis attacks), dynamically adjusting the length of the quantum basis selection control sequence according to the sensitivity of the first product, and performing quantum signature on the basis selection sequence using a first quantum private key. (This is used to verify the authenticity of the activity rules.)

[0190] Optionally, in some embodiments of this application, hot-selling activities can also be identified based on quantum optimization algorithms. Specifically, such as... Figure 5 As shown, embodiments of this application also include:

[0191] 501. The e-commerce session management platform receives the first message from the first buyer's terminal.

[0192] The first message is used to inquire about discount information for best-selling products.

[0193] 502. The e-commerce session management platform maps the first message to a second quantum state.

[0194] 503. The e-commerce session management platform triggers a federated query based on the second quantum state and distributes encrypted requests to each e-commerce platform.

[0195] 504. The e-commerce session management platform receives promotional information from various e-commerce platforms and performs quantum aggregation on the promotional information to obtain the target promotional plan.

[0196] 505. After decrypting the target discount scheme, the e-commerce session management platform returns it to the buyer's terminal.

[0197] For example, if a user inquires about promotional activities (such as lowest price + free gifts) for a mobile phone of brand X and model X00p, the promotional activity parameters submitted by each e-commerce platform to the e-commerce session management platform are as follows:

[0198] Discount rate (di, e.g., JD Plus member price 10% off → 0.9)

[0199] Inventory quantity (si, e.g., X00p inventory 1000 units).

[0200] Buyer matching score (ui, similarity based on historical inquiries, range [0,1])

[0201] Activity validity period (ti, unit: hours)

[0202] First, the e-commerce session management platform encodes each activity parameter into a quantum state, as shown below:

[0203] JD.com: d=0.85, s=500, u=0.92, t=72

[0204] Taobao: d=0.88, s=300, u=0.88, t=48

[0205] Pinduoduo: d=0.75, s=1000, u=0.85, t=96

[0206] Secondly, the e-commerce session management platform uses the quantum annealing algorithm to superimpose the discount rate code |di>, inventory code |si>, user matching degree code |ui>, and validity period code |ti> to obtain the target discount scheme |ψi>= |di>⊗∣si>⊗∣ui>⊗∣ti>, and finally returns the target discount scheme to the buyer's terminal.

[0207] For example, optimizing the activity parameters yields the optimal solution (i.e., the target discount scheme): selecting the combination of Taobao (d=0.88) + Pinduoduo (s=1000). Federated validation determines that the total inventory is 1300 ≤ 1500, thus achieving the target buyer coverage and improving Taobao's ranking.

[0208] It is evident that the synergy between quantum annealing and federated learning enables real-time optimization of cross-platform promotional activities. This involves simultaneously encoding multi-dimensional activity parameters such as discounts, inventory, and buyer matching; performing local annealing calculations on each e-commerce platform; and then aggregating these parameters through the e-commerce session management platform to obtain the globally optimal solution. Based on buyer feedback, the weights of the activity parameters for each platform in the target promotional plan are dynamically adjusted, thereby achieving real-time optimization of cross-platform promotional activities.

[0209] Optionally, in some embodiments of this application, considering that federated knowledge queries may suffer from knowledge distillation loss (leading to loss of details, such as the high attenuation rate of micro-expression information in Douyin short videos), cold start dilemma (high false answer rate for new knowledge points), or message collaboration latency between this e-commerce session management platform and the actual e-commerce platform to which the message belongs, a consultation response scheme that predicts knowledge hotspots (such as long-tail products) in future time periods can also be configured to reduce the false answer rate. This method also includes:

[0210] 601. Construct a cross-platform spatiotemporal coordinate system and label the spatiotemporal entropy value of each knowledge point in the dynamic knowledge base.

[0211] The spatiotemporal knowledge modeling engine first generates multi-dimensional semantic vectors, then constructs user interaction feature vectors = [click rate x, dwell time y, conversion rate to human agent z]. Finally, the spatiotemporal dimension t is added to obtain a cross-platform spatiotemporal coordinate system: (𝑥,𝑦,𝑧,𝑡) = GeoHash(𝐼𝑃)⊕timestamp

[0212] For example, during the Spring Festival shopping season, a sudden surge in inquiries about a popular gift box appeared. This nut gift box suddenly became a hit on Douyin, with inquiries increasing by 500%. The cross-platform spatiotemporal coordinate system detected a surge in the spatiotemporal entropy value of the Shenzhen region, marked the Shenzhen region as a red starburst, and enhanced the gravitational field of the Shenzhen region. As a result, Taobao inventory data, Pinduoduo's similar product information, and historical after-sales problem solutions were brought closer to the nut gift box in the nebula-like knowledge density distribution.

[0213] 602. Based on knowledge quality indicators and spatiotemporal curvature, focus on high-frequency knowledge points in the dynamic knowledge base to form a nebula-like knowledge density distribution.

[0214] The knowledge quality index is derived based on access frequency, relevance strength, and timeliness. Optionally, one formula for calculating the knowledge quality index is as follows: 𝑀=log(𝐹+1)×√S×𝑒 −𝑇 / 𝜏 Where F is the access frequency, S is the association strength, T is the timeliness, and τ is the knowledge half-life, which is usually set to 30 days.

[0215] Spatiotemporal curvature is used to distort the distribution trajectory of a second knowledge point within a nebula-like knowledge density distribution, based on the gravitational field generated by the mass of a first knowledge point. This quantifies the potential evolutionary patterns of user behavior patterns and product popularity dynamics (e.g., the diffusion of hotspot areas). The second knowledge point is an adjacent knowledge point of the first knowledge point within the nebula-like knowledge density distribution. This spatiotemporal curvature includes temporal curvature, spatial curvature, and behavioral curvature. Temporal curvature characterizes the acceleration of user attention changes over time; spatial curvature characterizes the degree of path curvature in the diffusion of hotspot areas; and behavioral curvature characterizes the drastic nature of changes in user interaction patterns.

[0216] For example, if a surge in traffic for a certain knowledge point is predicted 30 minutes in advance (such as detecting an abnormal surge in traffic in a live stream), the corresponding intelligent customer service resource scheduling plan for the target merchant on the live stream platform can be automatically triggered.

[0217] For example, taking Douyin (TikTok) as an example, the attention given to down jackets by users in the Yangtze River Delta region increases by 15% per hour, three times the growth rate of users in North China. By detecting the attention given to this down jacket, a peak in the curvature of attention was detected between 8:00 PM and 10:30 PM. Therefore, it is possible to predict the diffusion path of the down jacket's popularity in the next 2-3 hours and identify abnormal fluctuations (for example, identifying a sudden change in the spatiotemporal curvature in a certain area would indicate the risk of fraudulent orders).

[0218] Nebula-like knowledge density distribution refers to a nebula-like structure formed by classifying product-related knowledge according to its popularity, relevance, and timeliness.

[0219] Core Area: 5% of the highest quality knowledge points, including product parameters and best-selling promotion rules, such as real-time inventory and prices of best-selling products on Taobao, the core selling points currently being explained in Douyin live streams, and the countdown and remaining slots for Pinduoduo group buying.

[0220] Important Zone: 15% of the less important knowledge points, including matching recommendations and common user questions. This can be presented as a knowledge graph, such as illustrated guides on matching down jackets with sweaters, solutions to payment issues in mobile group buying, and lipstick swatches and skin tone matching suggestions.

[0221] Dark knowledge zone: 80% of long-tail knowledge is freely diffused, which is "dark knowledge" that is not labeled but actually affects the distribution of knowledge, such as historical data and extreme case handling solutions. For example, complaint handling records of a certain product two years ago, solutions for logistics anomalies in specific regions, and customized needs of niche user groups.

[0222] The core area is used for real-time monitoring of changes in basic product information; the important area is used for periodic (1 hour) updates of recommended combinations and intelligent filtering of frequently asked user questions. For example, when a user stays for more than 30 seconds, the system automatically enhances the attraction of related knowledge (such as popping up an ingredient analysis report); the dark knowledge area is used to awaken historical data through the gravitational field perturbation of spacetime curvature to cope with sudden public opinion events. For example, a defense mechanism is triggered when abnormal fluctuations are detected in the dark knowledge area (such as those matching preset scalper fraud characteristics).

[0223] 603. Generate spatiotemporal heat maps based on knowledge points from different e-commerce platforms.

[0224] The spatiotemporal heatmap includes knowledge overlap and conflict areas, knowledge blind spots, and new knowledge focal points based on the time dimension.

[0225] 604. Based on the nebula-like knowledge density distribution and the spatiotemporal heat map simulation of the evolution of knowledge density, an operation plan for the product to be operated is generated.

[0226] For example, the operational plan may include product price or inventory updates (e.g., real-time group-buying data to enhance the sense of urgency), hourly updates of recommended combinations, and leveraging historical data to respond to sudden public opinion events (e.g., identifying scalpers, extreme price fluctuations, or analyzing past group-buying events).

[0227] In some implementation methods, for the knowledge storm scenario of live-streaming e-commerce:

[0228] The product to be operated is the product with peak traffic in the live stream. The knowledge quality index is dynamically calculated based on the real-time number of viewers, interaction rate, and conversion rate. Step 604 includes:

[0229] 604-1: Simulates the response to user queries and associates product-related data that matches the user's query intent.

[0230] For example, when a user asks "Is sunscreen SPF60+ suitable for sensitive skin?", the system automatically links to the product ingredient list, laboratory test reports, and comparative data of similar products.

[0231] 604-2: Based on spatiotemporal curvature, the product-related data is aggregated towards the knowledge focus, and multiple intelligent scripts are generated using a diffusion model. For example, a diffusion model is used to generate text-image mixed scripts, and the information presentation order is adjusted by combining real-time heatmaps, knowledge hotspots are identified, the style is "live-streaming e-commerce", and the prompt words are "highlight" and "use urgent tone".

[0232] 604-3: Adjusting the verbal characteristics of each intelligent script based on user behavior trajectories analyzed by spatiotemporal heatmaps, and rearranging each intelligent script. For example, based on user behavior trajectories, the information density can be adjusted for the user's location (browsing stage) to obtain the verbal characteristics for different browsing stages, as shown in Table 3 below:

[0233]

[0234] Table 3

[0235] In other implementations, for cross-platform policy fluctuation scenarios, if Platform 1 suddenly adjusts the subsidy rules for a certain brand of mobile phone, causing price fluctuations on Platforms 2 and 3, and a surge in user inquiries, a three-fold increase can be observed. Considering that user issues may also exhibit regional and platform differences, real-time data can be collected on Platform 2 (user geographic distribution), Platform 3 (real-time live stream comments), and Platform 1 (order flow). Based on this data, a 3D heatmap is generated. Heatmap features from Platforms 1, 2, and 3 are extracted from the 3D heatmap, and a spatiotemporal dimension is introduced to optimize the messaging for the mobile phone on each platform. This involves reconstructing the messaging and then rearranging it based on a heat-weighted algorithm. For example, in the Yangtze River Delta region, price protection is strengthened; in Northeast China, logistics commitments are increased; and in live streams, inventory anxiety is alleviated.

[0236] As can be seen, in this embodiment of the application, by configuring a consultation response scheme that predicts knowledge hotspots (such as long-tail products) in the future, the false response rate of long-tail products can be reduced, and various problems such as price protection and return policies caused by the lag in updating customer service scripts on various platforms can be reduced.

[0237] Optionally, in some embodiments of this application, the method further includes:

[0238] The system acquires reception data from each sub-account and extracts a set of scripts from this data. This set includes excellent reception scripts with reception quality exceeding a preset quality, as well as reception scripts generated based on historical reception conversation responses. After detecting changes in the script set based on a preset event update strategy, the changed script set is synchronized to each sub-account. It is understandable that before synchronizing the scripts, adaptive adjustments can be made to the scripts to be synchronized based on platform differences to ensure compatibility with the platforms to be synchronized. Therefore, by using an event update strategy, the system achieves dynamic synchronization of excellent scripts across sub-accounts within this e-commerce conversation management platform, thereby optimizing the reception services of intelligent customer service across all e-commerce platforms.

[0239] Figure 1 to Figure 5 Any technical feature mentioned in the embodiments corresponding to any one of the above also applies to the embodiments of this application. Figures 6 to 7 The corresponding implementation examples will not be repeated hereafter.

[0240] The above describes the intelligent customer service session aggregation performance tuning method in this application. The following describes the e-commerce session management platform that implements the above intelligent customer service session aggregation performance tuning method.

[0241] See Figure 6 ,like Figure 6The diagram shows the structure of an e-commerce conversation management platform 70, which can be applied to cross-e-commerce platform intelligent customer service scenarios. By aggregating intelligent customer service services from multiple platforms, it improves conversation efficiency. The e-commerce conversation management platform 70 in this embodiment can achieve the functionality corresponding to Figure 1 above. Figure 5 The steps in the intelligent customer service conversation aggregation performance adjustment method executed by the e-commerce conversation management platform 70 in any corresponding embodiment. The functions implemented by the e-commerce conversation management platform 70 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. The e-commerce conversation management platform 70 may include an input / output module 701, a processing module 702, and a display module 703. The functional implementation of the input / output module 701, the processing module 702, and the display module 703 can be referred to Figure 1- Figure 5 The operations performed in any of the corresponding embodiments will not be described in detail here.

[0242] In some implementations, the input / output module 701 can be used to acquire real-time reception data of the plurality of intelligent customer service representatives;

[0243] Processing module 703 can be used to obtain cross-platform response delay distribution map, session transfer rate trend and knowledge base hit rate heat map based on the real-time customer service reception data;

[0244] Display module 703 is used to display cross-platform response latency distribution map, session transfer rate trend and knowledge base hit rate heatmap;

[0245] The processing module 703 is also used for: if the response latency of the intelligent customer service of the target e-commerce platform of the target merchant meets the preset latency conditions in a specific period of time based on the cross-platform response latency distribution map analysis, then the input / output module 701 captures the service status of the intelligent customer service of each e-commerce platform in real time, determines the latency trend of the target e-commerce platform based on the service status, and dynamically adjusts the capacity or updates the instance based on the latency trend; if the analysis shows that the reason for the surge in the session transfer rate is at least one of knowledge gaps, user emotions, or complex processes, then the corresponding preset linkage event is executed according to the corresponding reason; if the first knowledge base bound to the target merchant is determined to have blind spots based on the knowledge base hit rate heatmap, then the input / output module 701 captures the blind spot knowledge of the target product and supplements it to the first knowledge base, and verifies whether the hit rate has improved.

[0246] In this embodiment of the application, by adopting this solution, it is possible to centrally handle user inquiries from the same merchant on different e-commerce channels, improve session efficiency, facilitate cross-platform message management, and optimize platform performance.

[0247] The e-commerce session management platform 70 implementing the intelligent customer service session aggregation performance adjustment method in this application embodiment has been described above from the perspective of modular functional entities. The following description focuses on the hardware processing perspective of the e-commerce session management platform 70 implementing the intelligent customer service session aggregation performance adjustment method in this application embodiment. It should be noted that in this application embodiment... Figure 6 In the embodiments shown, the physical device corresponding to the input / output module 701 can be an input / output unit, a transceiver, a radio frequency circuit, a communication module, and an output interface, etc.; the physical device corresponding to the display module 703 can be a display screen; and the physical device corresponding to the processing module 702 can be a processor. Figure 6 The e-commerce session management platform 70 shown can have the following functions: Figure 7 The structure shown, when Figure 6 The e-commerce session management platform 70 shown has the following features: Figure 7 When the structure shown is used, Figure 7 The processor and transceiver in the device can perform the same or similar functions as the input / output module 701, processing module 702, and display module 703 provided in the aforementioned device embodiment corresponding to the e-commerce session management platform 70. Figure 7 The memory storage processor in the memory needs to call the computer program when executing the above-mentioned intelligent customer service session aggregation performance tuning method.

[0248] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, apparatuses, or modules, and may be electrical, mechanical, or other forms.

[0249] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0250] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0251] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. A computer program product includes one or more computer instructions. When the computer program is loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another, for example, to another computer, server, or data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium.

[0252] The technical solutions provided in this application have been described in detail above. Specific examples have been used in this application to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the embodiments of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for adjusting the performance of intelligent customer service conversation aggregation, characterized in that, The method is applied to an e-commerce session management platform, where the target merchant uses multiple intelligent customer service representatives on the platform to handle user inquiries from multiple e-commerce platforms; the method includes: Obtain real-time reception data from the multiple intelligent customer service systems; Based on the real-time customer service reception data, a cross-platform response delay distribution map, a session transfer rate trend, and a knowledge base hit rate heatmap were obtained. If the response latency of the intelligent customer service of the target merchant's target e-commerce platform meets the preset latency conditions during a specific period based on the cross-platform response latency distribution map analysis, then the service status of the intelligent customer service of each e-commerce platform is captured in real time, the latency trend of the target e-commerce platform is determined based on the service status, and the capacity is dynamically adjusted or the instance is updated based on the latency trend. If the analysis shows that the surge in session switching rate is due to at least one of the following: lack of knowledge, user emotions, or complex processes, then the corresponding preset linkage event will be executed according to the corresponding reason. If the first knowledge base bound to the target merchant is found to have blind spots based on the knowledge base hit rate heatmap, then the blind spot knowledge of the target product is captured and added to the first knowledge base, and the hit rate is verified to improve.

2. The method according to claim 1, characterized in that, The dynamic knowledge base originates from the knowledge bases of different e-commerce platforms, and these knowledge bases are all federated nodes. If the reason for the surge in session switching rate is analyzed to be at least one of knowledge gaps, user emotions, or complex processes, then corresponding preset linkage events are executed according to the corresponding reason, including: If the surge in session switching rate is attributed to a lack of knowledge, a knowledge base update will be triggered, including one of the following: Obtain the latest activity parameters of the first product on the first e-commerce platform, and link the latest activity parameters with the first product on the first e-commerce platform based on the product linkage mechanism; Alternatively, obtain the inventory status of the first product on the first e-commerce platform, and link the inventory status with the live streaming information of the second product on the second e-commerce platform based on the product linkage mechanism. The product linkage mechanism is a differential synchronization algorithm that periodically acquires the product knowledge increment of each federation node and broadcasts the product knowledge increment to the federation network. If the analysis shows that the surge in session conversion rate is due to process complexity, then add a quick operation button to the visualized process for the product association dimension of the first product. If the analysis indicates that the surge in conversation switching rate is due to negative user emotions, then switch to reassuring language and assign specific customer service representatives with capabilities exceeding a preset threshold to intervene.

3. The method according to claim 1 or 2, characterized in that, Each e-commerce platform and e-commerce session management platform is a federated node, and the method further includes: Receive signature data for the first product from multiple e-commerce platforms; After decomposing the hash values ​​in each signature data and confirming that the original product activity parameters have not been tampered with by using the hash values, the signature data from each e-commerce platform is aggregated to obtain aggregated signature data. The aggregated signature data is distributed to various e-commerce platforms via quantum teleportation.

4. The method according to claim 3, characterized in that, The method further includes: Conflict detection is performed on at least one of the activity parameters and signature data of the first product. If the difference between any activity parameters exceeds the threshold and the semantic similarity is less than the preset similarity, or if the aggregated signature verification fails, then it is determined that there is a knowledge conflict between the e-commerce platforms. A federal arbitration mechanism is triggered to quantify the economic losses of conflicts between e-commerce platforms. Conflict arbitration requests are distributed to each e-commerce platform via quantum keys. Weighted activity parameters are obtained by weighting the activity parameters of each e-commerce platform. The inventory of the first product on each e-commerce platform is allocated according to the dynamic weights of the e-commerce platforms. The conflict arbitration request includes the conflict type, the activity parameters with knowledge conflicts, and the semantic difference degree of the activity parameter difference.

5. The method according to claim 4, characterized in that, If there is a knowledge conflict between the first product on the first e-commerce platform and the second e-commerce platform, after triggering the federal arbitration mechanism, the method further includes: The first e-commerce platform generates a quantum basis selection control sequence and a basis selection sequence, encodes the first activity parameter of the first product into a first quantum state according to the quantum basis selection control sequence and the basis selection sequence, and sends the first quantum state to the second e-commerce platform through the e-commerce session management platform; wherein, the basis selection sequence is a deterministic basis vector sequence generated by the quantum basis selection control sequence based on the first activity parameter; The second e-commerce platform uses a preset shared key to decrypt and obtain the plaintext quantum basis selection control sequence and basis selection sequence. It extracts the first activity parameter from the basis selection sequence through basis vector alignment and updates the first activity parameter to the dynamic knowledge base of the first product on the second e-commerce platform.

6. The method according to claim 5, characterized in that, The method further includes: The first e-commerce platform adds random redundant basis vectors to the basis selection sequence, dynamically adjusts the length of the quantum basis selection control sequence according to the sensitivity of the first product, and uses the first quantum private key to perform quantum signature on the basis selection sequence.

7. The method according to claim 3, characterized in that, The method further includes: Receive the first message from the first buyer's terminal, the first message being used to inquire about discount information for hot-selling products; Map the first message to a second quantum state; Based on the second quantum state, a federated query is triggered, and encrypted requests are distributed to various e-commerce platforms; We receive promotional information from various e-commerce platforms and perform quantum aggregation on these promotional information to obtain the target promotional plan. After decrypting the target discount offer, it is returned to the buyer's terminal.

8. A method for adjusting the performance of intelligent customer service conversation aggregation, characterized in that, The method is applied to an e-commerce session management system, which includes an e-commerce session management platform, a first e-commerce platform, and a second e-commerce platform. Both the first and second e-commerce platforms are federated nodes, and the method includes: The e-commerce session management platform performs conflict detection on the product activity parameters of the first product. When a knowledge conflict is detected between the first e-commerce platform and the second e-commerce platform, a federal arbitration mechanism is triggered. The first e-commerce platform generates a quantum basis selection control sequence and a basis selection sequence. Based on the quantum basis selection control sequence and the basis selection sequence, it encodes the first activity parameter of the first product into a first quantum state and sends the first quantum state to the second e-commerce platform through the e-commerce session management platform. The basis selection sequence is a deterministic basis vector sequence generated by the quantum basis selection control sequence based on the first activity parameter. The second e-commerce platform uses a preset shared key to decrypt and obtain the plaintext quantum basis selection control sequence and basis selection sequence. It then extracts the first activity parameter from the basis selection sequence through basis vector alignment and updates the first activity parameter to the dynamic knowledge base of the first product on the second e-commerce platform.

9. An e-commerce conversation management platform, characterized in that, The e-commerce session management platform includes: At least one processor and memory; The memory is used to store computer programs, and the processor is used to invoke the computer programs stored in the memory to execute the method as described in any one of claims 1-8.

10. A computer program product comprising instructions, the computer program product including program instructions that, when executed on a computer or processor, cause the computer or processor to perform the method as claimed in any one of claims 1 to 8.

Images