Intelligent agent knowledge updating evaluation method and device, electronic equipment and storage medium

Abstract

The invention provides an agent knowledge updating evaluation method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence. An environment dimension updating urgency degree is calculated based on correlation between environment changes and knowledge elements, a user dimension personalized demand evaluation result is calculated based on a matching degree between user behaviors and the knowledge elements, and a knowledge structure importance evaluation result is calculated based on structural features of the knowledge elements in a global knowledge graph. The three types of evaluation results are subjected to weighted fusion to generate a comprehensive priority score, and then knowledge updating resources are dynamically allocated and a knowledge updating mode is determined through a multi-layer decision-making mechanism according to the comprehensive priority score. The problems that in the prior art, resource distribution is unbalanced, response lags behind, high-value knowledge is not updated in time, and updating adaptability is insufficient can be solved.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to an evaluation method and apparatus, electronic device and storage medium for updating the knowledge of an intelligent agent. Background Technology

[0002] Multi-agent collaborative knowledge update technology is a core support for artificial intelligence systems, widely applied in complex scenarios such as intelligent healthcare, financial risk control, and personalized recommendations. This technology system integrates various methods, including central server polling, event-triggered publish-subscribe, and distributed diffusion, covering the entire process of knowledge perception, feature extraction, and update decision-making, including key stages such as knowledge comparison, incremental updates, and model optimization. While centralized inference engines based on large language models and online fine-tuning techniques have become mainstream, their update strategies are still limited to fixed periods or preset events, failing to fully consider dynamic environmental changes, differences in user behavior, and the inherent connections between knowledge.

[0003] Existing knowledge update methods have significant shortcomings: First, they rely on a single priority evaluation dimension and fail to construct a comprehensive consideration mechanism that integrates environment, users, and knowledge structure, which can easily lead to uneven resource allocation or delayed response. Second, they lack the quantitative evaluation capability to dynamically perceive environmental changes, making it impossible to capture the demand for knowledge updates in a timely manner. Third, knowledge value assessment relies solely on version or time information, without combining long-term user preferences with a deep match of knowledge characteristics, which may result in delayed updates of high-value knowledge. Fourth, knowledge sharing adopts a simple copying approach, ignoring the differences between different intelligent agents in task scenarios and user needs, and failing to mitigate risks through personalized knowledge transfer methods, which can easily lead to negative transfer problems in scenarios such as medical diagnosis and financial model applications. Fifth, system parameter configurations are fixed, lacking a feedback-based closed-loop optimization mechanism, making it impossible to adjust update strategies based on actual application effects and hindering self-evolution. Summary of the Invention

[0004] This disclosure provides an evaluation method, apparatus, electronic device, and storage medium for knowledge updating of intelligent agents. Its main objective is to at least partially address one of the technical problems in related technologies.

[0005] According to a first aspect of this disclosure, an evaluation method for agent knowledge updating is provided, comprising: Based on the correlation between environmental change information and knowledge elements, the urgency assessment results for updating the environmental dimension are calculated. Based on the matching degree between user behavior information and knowledge elements, the personalized needs assessment results of user dimensions are calculated. Based on the structural characteristics of knowledge elements in the global knowledge graph, the evaluation results of the importance of knowledge structure are calculated. The urgency assessment results of the environmental dimension, the personalized needs assessment results of the user dimension, and the importance assessment results of the knowledge structure are weighted and integrated to generate a comprehensive priority score. Based on the comprehensive priority score, knowledge update resources are dynamically allocated and the knowledge update method is determined through a multi-level decision-making mechanism.

[0006] Optionally, the calculation of the urgency assessment result for updating the environmental dimension based on the correlation between environmental change information and knowledge elements includes: Encode the real-time environmental data sequence to generate an environmental state vector; The environmental attention weight is calculated through the interaction between the environmental state vector and the historical environmental latent state. The environmental state vector is linearly transformed using a learnable parameter matrix, and combined with a nonlinear activation function and normalization processing to generate an assessment result of the urgency of updating the environmental dimension.

[0007] Optionally, the calculation of personalized needs assessment results based on the matching degree between user behavior information and knowledge elements includes: The user's current feature vector is combined with the long-term preference vector to generate a user intent vector. The user intent vector is nonlinearly mapped using a multi-layer neural network, and a user attention score is generated by combining it with an activation function, serving as the result of personalized demand assessment for the user dimension.

[0008] Optionally, the calculation of the knowledge structure importance assessment result based on the structural features of knowledge elements in the global knowledge graph includes: The graph attention model is used to perform a nonlinear transformation on the knowledge node embedding vectors to calculate the attention coefficients between nodes. An aggregation function is used to perform pooling operations on the attention coefficients between the nodes to generate a global knowledge graph attention score, which serves as the result of the knowledge structure importance assessment.

[0009] Optionally, the weighted fusion of the environmental dimension update urgency assessment results, the user dimension personalized needs assessment results, and the knowledge structure importance assessment results to generate a comprehensive priority score includes: The weighting coefficients of the environmental dimension update urgency assessment results, the user dimension personalized needs assessment results, and the knowledge structure importance assessment results are dynamically adjusted through a reinforcement learning mechanism. A knowledge list to be updated is generated based on the comparison between the comprehensive priority score and the preset threshold.

[0010] Optional, also includes: After the knowledge update operation is completed, an evaluation window is entered to monitor user satisfaction indicators, task performance indicators, resource consumption indicators, and update delay indicators. The reward signal is calculated based on the monitoring indicators, and the parameters of the multi-layer decision-making mechanism are optimized through the policy network and the value network.

[0011] According to a second aspect of this disclosure, an evaluation apparatus for agent knowledge updating is provided, comprising: The first calculation unit is used to calculate the urgency assessment result of updating the environmental dimension based on the correlation between environmental change information and knowledge elements. The second calculation unit is used to calculate the personalized needs assessment results of the user dimension based on the matching degree between user behavior information and knowledge elements. The third computing unit is used to calculate the evaluation results of the importance of knowledge structure based on the structural features of knowledge elements in the global knowledge graph. The generation unit is used to weight and fuse the update urgency assessment results of the environment dimension, the personalized needs assessment results of the user dimension, and the knowledge structure importance assessment results to generate a comprehensive priority score. The determining unit is used to dynamically allocate knowledge update resources and determine the knowledge update method based on the comprehensive priority score through a multi-level decision-making mechanism.

[0012] Optionally, the first computing unit is also used for: Encode the real-time environmental data sequence to generate an environmental state vector; The environmental attention weight is calculated through the interaction between the environmental state vector and the historical environmental latent state. The environmental state vector is linearly transformed using a learnable parameter matrix, and combined with a nonlinear activation function and normalization processing to generate an assessment result of the urgency of updating the environmental dimension.

[0013] Optionally, the second computing unit is also used for: The user's current feature vector is combined with the long-term preference vector to generate a user intent vector. The user intent vector is nonlinearly mapped using a multi-layer neural network, and a user attention score is generated by combining it with an activation function, serving as the result of personalized demand assessment for the user dimension.

[0014] Optionally, the third computing unit is also used for: The graph attention model is used to perform a nonlinear transformation on the knowledge node embedding vectors to calculate the attention coefficients between nodes. An aggregation function is used to perform pooling operations on the attention coefficients between the nodes to generate a global knowledge graph attention score, which serves as the result of the knowledge structure importance assessment.

[0015] Optionally, the generating unit is also used for: The weighting coefficients of the environmental dimension update urgency assessment results, the user dimension personalized needs assessment results, and the knowledge structure importance assessment results are dynamically adjusted through a reinforcement learning mechanism. A knowledge list to be updated is generated based on the comparison between the comprehensive priority score and the preset threshold.

[0016] Optional, also includes: The optimization unit is used to monitor user satisfaction indicators, task performance indicators, resource consumption indicators, and update delay indicators during the evaluation window period after the knowledge update operation is completed; calculate reward signals based on the monitored indicators; and optimize the parameters of the multi-layer decision-making mechanism through the policy network and value network.

[0017] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect above.

[0018] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect above.

[0019] According to a fifth aspect of this disclosure, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0020] The intelligent agent knowledge update evaluation method, apparatus, electronic device, and storage medium disclosed herein calculate the urgency of updates in the environmental dimension based on the correlation between environmental changes and knowledge elements, calculate the personalized needs evaluation results in the user dimension based on the matching degree between user behavior and knowledge elements, and calculate the importance evaluation results of knowledge structure based on the structural characteristics of knowledge elements in the global knowledge graph. These three evaluation results are then weighted and fused to generate a comprehensive priority score. Based on this comprehensive priority score, a multi-layered decision-making mechanism dynamically allocates knowledge update resources and determines the knowledge update method. Therefore, this addresses the problems in existing technologies that rely solely on a single dimension to evaluate knowledge update needs, lack systematic consideration of dynamic environmental changes, personalized user needs, and the correlation of knowledge structure, and lack dynamic decision-making basis for update resource allocation and update method determination, resulting in unbalanced resource allocation, delayed response, untimely updates of high-value knowledge, and insufficient update adaptability. This achieves the technical effects of accurately quantifying knowledge update priorities, dynamically optimizing update resource allocation, ensuring priority updates for knowledge with high urgency, high user relevance, and high structural importance, and improving the targeting of knowledge updates and system collaborative efficiency.

[0021] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0022] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 A flowchart illustrating an evaluation method for intelligent agent knowledge updating provided in an embodiment of this disclosure; Figure 2 A schematic diagram of the structure of an evaluation device for intelligent agent knowledge updating provided in an embodiment of this disclosure; Figure 3 A schematic block diagram of an example electronic device provided for embodiments of this disclosure. Detailed Implementation

[0023] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0024] The following description, with reference to the accompanying drawings, outlines an evaluation method and apparatus for intelligent agent knowledge updating, an electronic device, and a storage medium according to embodiments of the present disclosure.

[0025] Figure 1 This is a flowchart illustrating an evaluation method for intelligent agent knowledge updating provided in an embodiment of this disclosure.

[0026] like Figure 1 As shown, the method includes the following steps: Step 101: Based on the correlation between environmental change information and knowledge elements, calculate the urgency assessment result of updating the environmental dimension.

[0027] In the embodiments of this disclosure, in the multi-agent collaborative knowledge update process, to determine the necessity of updating knowledge elements from an environmental perspective, environmental change information reflecting changes in the external scene or internal operating state of the system, as well as the knowledge elements in the multi-agent system to be evaluated for whether they need to be updated, are first acquired. Through a preset association analysis logic, the degree of association between environmental change information and knowledge elements is identified and quantified. This degree of association is then converted into an environmental dimension update urgency assessment result that can be used to measure the urgency of updating knowledge elements due to environmental changes. As one implementation method, the environmental change information can be feature-processed to obtain an environmental state representation. Then, an attention mechanism or similarity calculation method can be used to determine the correlation between the environmental state representation and the characteristics of the knowledge elements, thereby generating an environmental dimension update urgency assessment result.

[0028] This step enables the environmental dimension update assessment of knowledge elements to dynamically match actual environmental changes, avoiding the limitations of existing technologies that rely on fixed triggering conditions to assess environmental update needs. It provides an accurate basis for subsequent comprehensive judgment of knowledge update priorities that fits the actual needs of the environment.

[0029] Step 102: Calculate the personalized needs assessment results for the user dimension based on the matching degree between user behavior information and knowledge elements.

[0030] In the embodiments of this disclosure, in the multi-agent collaborative knowledge update process, to accurately determine whether knowledge elements meet personalized needs from a user perspective to support update decisions, user behavior information reflecting user habits, interaction preferences, or demand tendencies, as well as the knowledge elements to be evaluated in the multi-agent system, are first acquired. Through a preset matching analysis logic, the user needs reflected in the user behavior information are correlated and quantified with the content, purpose, or service scenario of the knowledge elements. The quantified matching degree is further transformed into a user-dimensional personalized demand evaluation result that characterizes the degree to which the knowledge elements satisfy the user's personalized needs. As one implementation method, user behavior information can be integrated to construct a user demand feature vector, and features of knowledge elements can be extracted to obtain knowledge feature vectors. The matching degree between the two is determined through vector similarity calculation, thereby generating a personalized demand evaluation result.

[0031] This step enables the evaluation of knowledge element updates to closely align with individual user behavior characteristics and actual needs, avoiding the problem in existing technologies where neglecting personalized differences leads to a disconnect between knowledge updates and user needs. It provides a precise user-centric basis for subsequently determining the priority of knowledge updates, thereby improving the adaptability of knowledge updates to user needs.

[0032] Step 103: Calculate the knowledge structure importance assessment results based on the structural features of knowledge elements in the global knowledge graph.

[0033] In the embodiments of this disclosure, in the multi-agent collaborative knowledge update process, to determine the core nature of knowledge elements from the perspective of knowledge structure to support update decisions, the global knowledge graph in the multi-agent system (which contains each knowledge element and the relationships between elements) and the target knowledge element to be evaluated for whether it needs to be updated are first obtained. Through a preset structural feature analysis logic, the structural attributes of the target knowledge element in the global knowledge graph (such as the number of connections between the element and other knowledge elements, connection strength, hierarchical position in the graph, or core correlation, etc.) are extracted, and these structural attributes are quantitatively analyzed. The quantitative results are then transformed into a knowledge structure importance assessment result that can characterize the fundamental position and influence scope of the target knowledge element in the entire knowledge system. As one implementation method, graph analysis algorithms can be used to identify the graph centrality, community affiliation, or critical path contribution of knowledge elements, and then the knowledge structure importance assessment result is calculated based on these structural feature values.

[0034] This step enables the evaluation of knowledge element updates to fully consider their structural value within the overall knowledge system. It avoids the problem in existing technologies where only the environmental or user dimensions are considered while neglecting the structural importance of the knowledge itself, leading to the underestimation of core knowledge. This provides a structural dimension basis that aligns with the internal logic of the knowledge system for the subsequent comprehensive determination of knowledge update priorities, ensuring that core knowledge that plays a key supporting role in the operation of the knowledge system is updated first.

[0035] Step 104: The update urgency assessment results of the environmental dimension, the personalized needs assessment results of the user dimension, and the knowledge structure importance assessment results are weighted and integrated to generate a comprehensive priority score.

[0036] In the embodiments of this disclosure, in the multi-agent collaborative knowledge update process, to comprehensively determine the overall update priority of knowledge elements by integrating the evaluation results of various dimensions, the previously obtained environmental dimension update urgency evaluation results, user dimension personalized demand evaluation results, and knowledge structure importance evaluation results are first obtained. Then, based on the specific application scenario requirements of the multi-agent system or the core objective of knowledge update, a reasonable weight allocation rule is preset (this rule is used to define the degree of influence of different dimension evaluation results on the overall update priority of knowledge elements). According to this rule, corresponding weight values ​​are assigned to the three types of evaluation results, and the weighted three types of evaluation results are fused and calculated. Finally, the calculated comprehensive value is converted into a comprehensive priority score that can comprehensively and objectively reflect the necessity and urgency of knowledge element updates. As one implementation method, the weight coefficients of each dimension can be determined by combining the system's historical update effect feedback or domain business requirements, and then the three types of evaluation results are fused through linear weighted summation to generate a comprehensive priority score.

[0037] By weighted and fused multi-dimensional evaluation results, this approach avoids the biased judgment problem caused by relying on only a single dimension to determine the priority of knowledge updates in existing technologies. It can comprehensively consider the impact of three factors—environment, users, and knowledge structure—on update requirements, providing a comprehensive and accurate basis for subsequent dynamic allocation of update resources and determination of update methods, thus ensuring the scientific and rational nature of update decisions.

[0038] Step 105: Based on the comprehensive priority score, dynamically allocate knowledge update resources and determine the knowledge update method through a multi-level decision-making mechanism.

[0039] In the embodiments of this disclosure, in the multi-agent collaborative knowledge update process, to efficiently allocate resources and select an appropriate update mode based on the overall necessity of updating knowledge elements, the previously generated comprehensive priority score of the knowledge elements is first obtained. Then, a multi-layer decision-making mechanism is initiated. This mechanism, through a phased and progressive decision-making logic, combines the comprehensive priority score with the real-time resource status of the multi-agent system (such as computational resource load, bandwidth usage, and storage availability). It first clarifies the core direction of resource allocation at the overall level, then gradually refines it to the resource quota of specific knowledge elements. Simultaneously, based on the comprehensive priority score and the inherent attributes of the knowledge elements (such as data volume and update frequency requirements), an appropriate knowledge update method (such as full update, incremental update, knowledge migration update, etc.) is determined, ultimately completing the dynamic allocation of knowledge update resources and the precise determination of the update method. As one implementation method, the multi-layer decision-making mechanism can be divided into a strategic layer (determining the general direction of resource allocation based on priority), a tactical layer (allocating specific resource quotas according to the direction), and an execution layer (selecting the update method based on the quota and priority), progressively advancing the decision-making process.

[0040] By combining a multi-layered decision-making mechanism with a comprehensive priority score, the problems of resource allocation and update method selection lacking pertinence and being easily affected by a single factor in existing technologies are avoided. This ensures that high-priority knowledge elements receive sufficient resources and appropriate update methods, while also achieving efficient utilization of system resources and improving the overall efficiency and adaptability of multi-agent collaborative knowledge updates.

[0041] The intelligent agent knowledge update evaluation method disclosed herein calculates the urgency of updates in the environmental dimension based on the correlation between environmental changes and knowledge elements, calculates the personalized needs evaluation results in the user dimension based on the matching degree between user behavior and knowledge elements, and calculates the importance evaluation results of knowledge structure based on the structural characteristics of knowledge elements in the global knowledge graph. These three evaluation results are weighted and fused to generate a comprehensive priority score. Then, based on this comprehensive priority score, knowledge update resources are dynamically allocated and knowledge update methods are determined through a multi-layer decision-making mechanism. Therefore, it can solve the problems of resource allocation imbalance, response lag, failure to update high-value knowledge in a timely manner, and insufficient update adaptability caused by existing technologies that rely only on a single dimension to evaluate knowledge update needs, lack systematic consideration of dynamic environmental changes, personalized user needs, and the correlation of knowledge structure, and lack dynamic decision-making basis for update resource allocation and update method determination. It achieves the technical effects of accurately quantifying knowledge update priority, dynamically optimizing update resource allocation, ensuring priority updates of knowledge with high urgency, high user relevance, and high structural importance, and improving the pertinence of knowledge updates and system collaborative efficiency.

[0042] As a specific implementation of this disclosure, based on the basic scheme, the calculation of the environmental dimension update urgency assessment result based on the correlation between environmental change information and knowledge elements is further defined, including: encoding the real-time environmental data sequence to generate an environmental state vector; calculating the environmental attention weight through the interaction between the environmental state vector and the historical environmental latent state; performing a linear transformation on the environmental state vector using a learnable parameter matrix, and combining it with a nonlinear activation function and normalization processing to generate the environmental dimension update urgency assessment result.

[0043] Specifically, the system collects real-time environmental data sequences during the operation of the multi-agent system. These sequences may include network traffic data and domain-related real-time monitoring data (such as market data in financial scenarios and device sensor data in medical scenarios). A Transformer encoder is used to extract and encode the real-time environmental data sequences, transforming the discrete environmental data into an environmental state vector with temporally correlated information. Simultaneously, historical environmental data is encoded using a Long Short-Term Memory (LSTM) network or a Gated Recurrent Unit (GRU) to generate historical environmental latent states. These historical environmental information represents the current state of the knowledge base (i.e., internalized knowledge). By calculating the attention interaction between real-time environmental data and historical environmental information, the system can effectively identify and extract information fragments from real-time data that are relevant to the existing knowledge state and have practical value for knowledge update decisions, thereby achieving intelligent filtering and focusing of environmental signals. This historical environmental latent state is used to store feature information about past environmental changes. Subsequently, the environmental state vector is used as the Query, and the historical environmental latent state is used as the Key to perform attention interaction calculations, obtaining environmental attention weights to quantify the impact of the correlation between the current and historical environments on knowledge update requirements. Furthermore, a learnable parameter matrix W is introduced. e Q e K e The environmental state vector is linearly transformed, and then the transformed vector is input into a tanh nonlinear activation function, limiting the output range to [-1, 1] to constrain numerical fluctuations. Finally, the activated result is normalized using softmax, mapping it to a value in the 0-1 interval. This value represents the urgency assessment result for updating the environmental dimension; a higher value indicates a greater urgency for updating knowledge elements due to changes in the current environment. This calculation process can be expressed as:

[0044] : Environmental attention weight vector, representing the importance of environmental changes to knowledge updates; The previous time step's hidden state is represented by tanh (historical environment information encoded using LSTM or GRU); tanh is the activation function, limiting the output range to [...]. [1,1]; softmax is used to normalize attention weights, ensuring that the sum is 1.

[0045] By capturing the temporal characteristics of environmental data through the Transformer encoder, calculating attention weights by combining historical environmental latent states, and then processing them through linear transformation and activation normalization, the environmental dimension update urgency assessment can fully integrate current and historical environmental information, and the results are accurately quantified. This effectively avoids misjudgment of urgency caused by single environmental data or lack of historical correlation analysis, and significantly improves the reliability and practicality of environmental dimension assessment results.

[0046] As a specific implementation of this disclosure, based on the basic scheme, the calculation of the personalized needs assessment result of the user dimension based on the matching degree of user behavior information and knowledge elements is further defined, including: performing feature cross operation on the user's current feature vector and long-term preference vector to generate a user intent vector; performing nonlinear mapping on the user intent vector through a multi-layer neural network, and combining it with an activation function to generate a user attention score as the personalized needs assessment result of the user dimension.

[0047] Specifically, two core vectors are constructed based on user behavior information: the user's current feature vector and the user's long-term preference vector. The user's current feature vector can be extracted from the user's recent (e.g., the last 7 days) behavioral data, including knowledge query keywords, interaction frequency, and dwell time, etc., and formed into a vector with unified dimensions after normalization. The user's long-term preference vector is generated by statistical modeling and feature aggregation of the user's historical behavioral data (e.g., stable interaction patterns in the last 3 months, repeatedly followed knowledge areas, long-term retention usage habits, etc.) to reflect the user's long-term stable demand tendencies. Subsequently, feature cross-operation is performed on the user's current feature vector and the user's long-term preference vector, specifically by vector concatenation (connecting the two vectors end to end to form a combined vector with superimposed dimensions) or element-wise dot product (calculating the product of corresponding elements of two vectors to generate a new feature vector), to fully integrate the user's immediate needs and long-term preference information, resulting in the user intent vector. Next, a multilayer neural network (MLP) is constructed. This network includes an input layer, at least one hidden layer, and an output layer. The dimension of the input layer matches the dimension of the user intent vector. The hidden layer uses the ReLU activation function to perform a non-linear transformation on the input features to capture the complex relationships of user needs. The output layer uses the Sigmoid activation function. The user intent vector is input into this multilayer neural network. After being processed by each layer in sequence, the output value is limited to the range [0,1]. This value is the user attention score, which is directly used as the personalized needs assessment result for the user dimension. The higher the value, the higher the degree of matching between the knowledge element and the user's current and long-term needs. This calculation process can be expressed as:

[0048] in, A user attention score is assigned to indicate the user's sensitivity to knowledge updates; is the long-term preference vector of user i (modeled through historical behavior); ⊕ is the feature cross operation (such as concatenation or dot product); MLP is a Multi-Layer Perceptron used for non-linear mapping; σ is the Sigmoid function with an output range of [0,1]; W u This is the learnable parameter matrix.

[0049] By integrating current and long-term user preferences through feature cross-operations, the one-sided representation of needs caused by relying solely on behavioral data from a single period is avoided. Furthermore, by combining the nonlinear mapping capability of multi-layer neural networks with the numerical constraints of the Sigmoid activation function, the complex characteristics of user needs can be accurately captured, making personalized needs assessment results more closely aligned with users' actual needs and effectively improving the accuracy and reliability of user-dimensional assessment results.

[0050] As a specific implementation of this disclosure, based on the basic scheme, the calculation of the knowledge structure importance assessment result based on the structural features of knowledge elements in the global knowledge graph includes: using a graph attention model to perform a nonlinear transformation on the knowledge node embedding vector to calculate the attention coefficients between nodes; and using an aggregation function to perform a pooling operation on the attention coefficients between nodes to generate a global knowledge graph attention score as the knowledge structure importance assessment result.

[0051] Specifically, the process involves obtaining the knowledge node corresponding to the knowledge element to be evaluated in the global knowledge graph, along with the initial embedding vector of that node (generated by initial encoding the attribute information and relational features of the knowledge node). A graph attention model (GAT) is introduced, and a learnable embedding transformation matrix Wg is configured. The initial embedding vector of the knowledge node is input into this matrix for nonlinear transformation, resulting in a node transformation vector with uniform dimensions and containing deep structural features. Subsequently, for any two related knowledge nodes (denoted as node j and node k), their transformation vectors are concatenated, and then a dot product is performed with the preset attention parameter vector a. The result is nonlinearly processed using the LeakyReLU activation function to highlight the differentiated features of the relationship between nodes. Finally, the result is normalized using the softmax function to obtain the attention coefficient between node j and node k, which quantifies the importance of the structural relationship between the two nodes. Furthermore, average pooling or max pooling is selected as the aggregation function to perform pooling operations on the attention coefficients between the target knowledge node and all its neighboring nodes. This integrates all structural association information of the node in the global knowledge graph, generating a global knowledge graph attention score within the range [0,1]. This score is the result of the knowledge structure importance assessment; a higher score indicates a higher degree of structural coreness of the knowledge element in the global knowledge graph. This calculation process can be expressed as:

[0052] in, v represents the attention weight between node j and its neighbor k; j ,v k W represents the embedding vectors of knowledge graph nodes j and k. g is the learnable embedding transformation matrix; a is the attention parameter vector; || is the vector concatenation operation; N(j) is the set of neighbors of node j; LeakyReLU is a non-linear activation function that allows negative output values.

[0053] Furthermore, an aggregation function such as average pooling or max pooling (Agg) is selected to perform pooling operations on the attention coefficients between the target knowledge node j and all its neighboring nodes. This integrates all structural association information of the node in the global knowledge graph, generating a global knowledge graph attention score Score_{kg} within the range [0,1]. This score is the result of the knowledge structure importance assessment; a higher score indicates a higher degree of structural coreness of the knowledge element in the global knowledge graph. The aggregation process can be represented as follows:

[0054] By accurately capturing the structural correlation strength between knowledge nodes using a graph attention model and integrating global structural information using an aggregation function, the evaluation bias caused by relying solely on local structural features is avoided. This ensures that the evaluation results of the importance of knowledge structure truly reflect the core position of knowledge elements in the overall knowledge system, providing a precise basis for prioritizing the identification of core knowledge in subsequent update decisions and effectively improving the reliability of knowledge structure dimension evaluation.

[0055] As a specific implementation of this disclosure, based on the basic scheme, the method of weightedly integrating the update urgency assessment results of the environmental dimension, the personalized demand assessment results of the user dimension, and the knowledge structure importance assessment results to generate a comprehensive priority score includes: dynamically adjusting the weight coefficients of the update urgency assessment results of the environmental dimension, the personalized demand assessment results of the user dimension, and the knowledge structure importance assessment results through a reinforcement learning mechanism; and generating a list of knowledge to be updated based on a comparison of the comprehensive priority score with a preset threshold.

[0056] Specifically, a reinforcement learning framework is constructed to achieve dynamic adjustment of weight coefficients, employing the Actor-Critic reinforcement learning algorithm. The system state space is defined as the historical update effect data of the multi-agent system (e.g., changes in user satisfaction after knowledge updates, task completion rate fluctuations), current network bandwidth utilization, and computational resource load. The action space is defined as the weight coefficients γ1 (environmental dimension update urgency assessment result), γ2 (user dimension personalized demand assessment result), and γ3 (knowledge structure importance assessment result), satisfying the constraint γ1+γ2+γ3=1, and can be adjusted according to different application scenarios. The policy network (Actor) outputs an adjustment scheme for the weight coefficients based on the current system state, while the value network (Critic) calculates the reward signal based on the actual application effect corresponding to the comprehensive priority score under this adjustment scheme (e.g., the increase in user interaction click-through rate after high-priority knowledge updates, computational resource consumption costs during the update process). The policy network parameters are iteratively optimized using gradient descent, allowing the weight coefficients to dynamically adapt to the system state and update effect. Subsequently, the dynamically adjusted γ1, γ2, and γ3 are multiplied by the corresponding environmental dimension assessment results, user dimension assessment results, and knowledge structure dimension assessment results, respectively, and summed to obtain the comprehensive priority score of each knowledge element. Two update thresholds are further preset: an emergency update threshold and a regular update threshold. The comprehensive priority score of each knowledge element is compared with these two thresholds. Knowledge elements with scores higher than the emergency update threshold are included in the "emergency update group" of the update list; those with scores between the emergency update threshold and the regular update threshold are included in the "regular update group"; and those with scores lower than the regular update threshold are temporarily excluded from the list. This process generates a clearly categorized list of knowledge elements to be updated. The final update priority score is obtained by weighted fusion of the three attention weight scores.

[0057] Among them, P (vi) γ1, γ2, γ3 are the overall attention weights of agent i; γ1, γ2, γ3 are the weight coefficients (γ1+γ2+γ3=1), which are dynamically adjusted by TRC; Agg(): the aggregated knowledge graph attention.

[0058] By dynamically adjusting the weight coefficients through reinforcement learning, the limitations of fixed weights in adapting to changes in system state are avoided, ensuring that the weights of the evaluation results in each dimension are aligned with the actual update effect and system load. Combined with preset thresholds to generate a categorized list of items to be updated, this not only accurately filters out the knowledge elements that need to be updated but also distinguishes the urgency of updates in advance, providing a clear basis for subsequent hierarchical resource allocation and effectively improving the practicality of the comprehensive priority score and the efficiency of update decisions.

[0059] As a specific implementation of this disclosure, based on the basic scheme, the embodiment of this disclosure further includes: after the knowledge update operation is completed, an evaluation window period is entered to monitor user satisfaction indicators, task performance indicators, resource consumption indicators and update delay indicators; a reward signal is calculated based on the monitored indicators, and the parameters of the multi-layer decision-making mechanism are optimized through a policy network and a value network.

[0060] Specifically, after the knowledge update operation is completed, the system automatically enters a preset evaluation window period (e.g., 24 hours, which can be flexibly adjusted according to the application scenario of the multi-agent system). During this period, the data acquisition module monitors four core indicators in real time: user satisfaction indicators (including the click rate of users on relevant knowledge after the knowledge update, page dwell time, and explicit feedback rating), task performance indicators (including the success rate and result accuracy of completing the target task based on the updated knowledge), resource consumption indicators (including the GPU computing time, network bandwidth usage, and storage resource utilization during the update process), and update latency indicators (including the average response time from the initiation of the knowledge update request to its completion). After monitoring is completed, the updated values ​​of each indicator are compared with the baseline values ​​before the update. The change in the indicator is calculated (e.g., the improvement in user satisfaction ΔU = click-through rate after update - click-through rate before update, the improvement in task performance ΔT = task success rate after update - task success rate before update). Then, according to the preset weights (e.g., α=0.3, β=0.3, γ=0.2, δ=0.2, satisfying α+β+γ+δ=1), the values ​​are substituted into the reward calculation formula R=α·ΔU +β·ΔT - γ·resource consumption indicator - δ·update delay indicator to obtain the reward signal for this update. Subsequently, based on the reward signal, the parameters of the multi-layer decision-making mechanism are optimized using the Actor-Critic reinforcement learning framework: the monitored four types of indicator data are input into the value network (Critic) as the system state s, the state value function Vπ(s) is calculated, and the temporal difference (TD) error δ=R+γVπ(s')-Vπ(s) (s' is the next state after optimization) is obtained. The weight parameters of the value network are updated by gradient descent. At the same time, the reward signal is combined with the policy gradient to adjust the parameters of the policy network (Actor) (these parameters correspond to the resource allocation ratio of the strategic layer, the budget allocation threshold of the tactical layer, and the update method selection conditions of the execution layer in the multi-layer decision-making mechanism). The above process is iteratively executed until the reward signal tends to stabilize, thus completing the parameter optimization of the multi-layer decision-making mechanism.

[0061] By comprehensively monitoring update performance indicators during the evaluation window period, we ensure that reward signals accurately reflect update quality. By combining reinforcement learning to optimize the parameters of the multi-layer decision-making mechanism, the decision logic can be dynamically adjusted according to the update performance, avoiding the problem that static parameters cannot adapt to changes in the scenario. This significantly improves the flexibility of the multi-layer decision-making mechanism and the accuracy of update decisions.

[0062] It should be noted that the embodiments of this disclosure may include multiple steps. For ease of description, these steps are numbered, but these numbers are not a limitation on the execution time slots or execution order between the steps; these steps can be implemented in any order, and the embodiments of this disclosure do not limit this.

[0063] Corresponding to the aforementioned evaluation method for agent knowledge updates, this disclosure also proposes an evaluation device for agent knowledge updates. Since the device embodiments of this disclosure correspond to the aforementioned method embodiments, details not disclosed in the device embodiments can be referred to the aforementioned method embodiments, and will not be repeated here.

[0064] Figure 2 This is a schematic diagram of the structure of an evaluation device for intelligent agent knowledge updating provided in an embodiment of this disclosure, as shown below. Figure 2 As shown, it includes: The first calculation unit 21 is used to calculate the urgency assessment result of updating the environmental dimension based on the correlation between environmental change information and knowledge elements. The second calculation unit 22 is used to calculate the personalized needs assessment results of the user dimension based on the matching degree between user behavior information and knowledge elements. The third computing unit 23 is used to calculate the evaluation results of the importance of knowledge structure based on the structural features of knowledge elements in the global knowledge graph. The generation unit 24 is used to weight and fuse the update urgency assessment results of the environment dimension, the personalized demand assessment results of the user dimension, and the knowledge structure importance assessment results to generate a comprehensive priority score. The determining unit 25 is used to dynamically allocate knowledge update resources and determine the knowledge update method based on the comprehensive priority score through a multi-level decision-making mechanism.

[0065] The intelligent agent knowledge update evaluation device disclosed herein calculates the urgency of updates in the environmental dimension based on the correlation between environmental changes and knowledge elements, calculates the personalized needs evaluation results in the user dimension based on the matching degree between user behavior and knowledge elements, and calculates the importance evaluation results of knowledge structure based on the structural characteristics of knowledge elements in the global knowledge graph. These three evaluation results are then weighted and fused to generate a comprehensive priority score. Based on this comprehensive priority score, a multi-layered decision-making mechanism dynamically allocates knowledge update resources and determines the knowledge update method. Therefore, it can solve the problems in existing technologies that rely solely on a single dimension to evaluate knowledge update needs, lack systematic consideration of dynamic environmental changes, personalized user needs, and the correlation of knowledge structure, and lack dynamic decision-making basis for update resource allocation and update method determination, resulting in unbalanced resource allocation, delayed response, untimely updates of high-value knowledge, and insufficient update adaptability. This achieves the technical effects of accurately quantifying knowledge update priorities, dynamically optimizing update resource allocation, ensuring priority updates for knowledge with high urgency, high user relevance, and high structural importance, and improving the targeting of knowledge updates and system collaborative efficiency.

[0066] Furthermore, in one possible implementation of this embodiment, the first computing unit 21 is also used for: Encode the real-time environmental data sequence to generate an environmental state vector; The environmental attention weight is calculated through the interaction between the environmental state vector and the historical environmental latent state. The environmental state vector is linearly transformed using a learnable parameter matrix, and combined with a nonlinear activation function and normalization processing to generate an assessment result of the urgency of updating the environmental dimension.

[0067] Furthermore, in one possible implementation of this embodiment, the second computing unit 22 is also used for: The user's current feature vector is combined with the long-term preference vector to generate a user intent vector. The user intent vector is nonlinearly mapped using a multi-layer neural network, and a user attention score is generated by combining it with an activation function, serving as the result of personalized demand assessment for the user dimension.

[0068] Optionally, the third computing unit 23 is also used for: The graph attention model is used to perform a nonlinear transformation on the knowledge node embedding vectors to calculate the attention coefficients between nodes. An aggregation function is used to perform pooling operations on the attention coefficients between the nodes to generate a global knowledge graph attention score, which serves as the result of the knowledge structure importance assessment.

[0069] Furthermore, in one possible implementation of this embodiment, the generation unit 24 is also used for: The weighting coefficients of the environmental dimension update urgency assessment results, the user dimension personalized needs assessment results, and the knowledge structure importance assessment results are dynamically adjusted through a reinforcement learning mechanism. A knowledge list to be updated is generated based on the comparison between the comprehensive priority score and the preset threshold.

[0070] Furthermore, in one possible implementation of this embodiment, such as Figure 2 As shown, it also includes: The optimization unit 26 is used to enter the evaluation window period after the knowledge update operation is completed, monitor user satisfaction indicators, task performance indicators, resource consumption indicators and update delay indicators; calculate reward signals based on the monitored indicators, and optimize the parameters of the multi-layer decision-making mechanism through the policy network and value network.

[0071] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus of this embodiment, and the principle is the same, so it is not limited in this embodiment.

[0072] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0073] Figure 3 A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0074] like Figure 3 As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 302 or a computer program loaded from storage unit 308 into RAM (Random Access Memory) 303. The RAM 303 may also store various programs and data required for the operation of the electronic device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An I / O (Input / Output) interface 305 is also connected to the bus 304.

[0075] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of displays, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0076] The computing unit 301 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as the evaluation method for agent knowledge updating. For example, in some embodiments, the evaluation method for agent knowledge updating may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the methods described above may be performed. Alternatively, in other embodiments, computing unit 301 may be configured to perform the aforementioned evaluation method for agent knowledge updates by any other suitable means (e.g., by means of firmware).

[0077] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0078] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0079] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0080] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0081] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0082] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service system that addresses the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0083] It's important to note that artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0084] The various numerical designations such as "first," "second," etc., used in this disclosure are merely for ease of description and are not intended to limit the scope of the embodiments of this disclosure, nor do they indicate a sequential order.

[0085] At least one of the features described in this disclosure can also be described as one or more, and multiple features can be two, three, four or more, and this disclosure does not impose any limitations. In the embodiments of this disclosure, for a technical feature, the technical features in that technical feature are distinguished by "first", "second", "third", "A", "B", "C" and "D", etc., and there is no sequential order or size order among the technical features described by "first", "second", "third", "A", "B", "C" and "D".

[0086] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0087] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An evaluation method for agent knowledge update, characterized in that, include: Based on the correlation between environmental change information and knowledge elements, the urgency assessment results for updating the environmental dimension are calculated. Based on the matching degree between user behavior information and knowledge elements, the personalized needs assessment results of user dimensions are calculated. Based on the structural characteristics of knowledge elements in the global knowledge graph, the evaluation results of the importance of knowledge structure are calculated. The urgency assessment results of the environmental dimension, the personalized needs assessment results of the user dimension, and the importance assessment results of the knowledge structure are weighted and integrated to generate a comprehensive priority score. Based on the comprehensive priority score, knowledge update resources are dynamically allocated and the knowledge update method is determined through a multi-level decision-making mechanism.

2. The method of claim 1, wherein, The assessment of the urgency of updating the environmental dimension based on the correlation between environmental change information and knowledge elements includes: Encode the real-time environmental data sequence to generate an environmental state vector; The environmental attention weight is calculated through the interaction between the environmental state vector and the historical environmental latent state. The environmental state vector is linearly transformed using a learnable parameter matrix, and combined with a nonlinear activation function and normalization processing to generate an assessment result of the urgency of updating the environmental dimension.

3. The method of claim 1, wherein, The calculation of personalized needs assessment results based on the matching degree between user behavior information and knowledge elements includes: The user's current feature vector is combined with the long-term preference vector to generate a user intent vector. The user intent vector is nonlinearly mapped using a multi-layer neural network, and a user attention score is generated by combining it with an activation function, serving as the result of personalized demand assessment for the user dimension.

4. The method according to claim 1, characterized in that, The calculation of the importance assessment result of knowledge structure based on the structural features of knowledge elements in the global knowledge graph includes: The graph attention model is used to perform a nonlinear transformation on the knowledge node embedding vectors to calculate the attention coefficients between nodes. An aggregation function is used to perform pooling operations on the attention coefficients between the nodes to generate a global knowledge graph attention score, which serves as the result of the knowledge structure importance assessment.

5. The method according to claim 1, characterized in that, The assessment results of the urgency of updates in the environmental dimension, the personalized needs assessment results in the user dimension, and the importance assessment results of the knowledge structure are weighted and integrated to generate a comprehensive priority score, including: The weighting coefficients of the environmental dimension update urgency assessment results, the user dimension personalized needs assessment results, and the knowledge structure importance assessment results are dynamically adjusted through a reinforcement learning mechanism. A knowledge list to be updated is generated based on the comparison between the comprehensive priority score and the preset threshold.

6. The method according to claim 1, characterized in that, Also includes: After the knowledge update operation is completed, an evaluation window is entered to monitor user satisfaction indicators, task performance indicators, resource consumption indicators, and update delay indicators. The reward signal is calculated based on the monitoring indicators, and the parameters of the multi-layer decision-making mechanism are optimized through the policy network and the value network.

7. An evaluation device for knowledge updating of an intelligent agent, characterized in that, include: The first calculation unit is used to calculate the urgency assessment result of updating the environmental dimension based on the correlation between environmental change information and knowledge elements. The second calculation unit is used to calculate the personalized needs assessment results of the user dimension based on the matching degree between user behavior information and knowledge elements. The third computing unit is used to calculate the evaluation results of the importance of knowledge structure based on the structural features of knowledge elements in the global knowledge graph. The generation unit is used to weight and fuse the update urgency assessment results of the environment dimension, the personalized needs assessment results of the user dimension, and the knowledge structure importance assessment results to generate a comprehensive priority score. The determining unit is used to dynamically allocate knowledge update resources and determine the knowledge update method based on the comprehensive priority score through a multi-level decision-making mechanism.

8. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.

Images