A knowledge graph-based intelligent library reading promotion management method

Abstract

The application relates to the technical field of promotion management, and discloses a wisdom library reading promotion management method based on a knowledge graph, which comprises the following steps: S1, acquiring library resources, user behaviors and scene data to construct a knowledge graph, providing a semantic association basis for intelligent recommendation and intelligent matching of book resources and user demands; S2, dynamically adjusting knowledge association weights to improve the accuracy, self-adaptability and intelligent decision of promotion strategies of the system; and S3, analyzing multi-dimensional behavior data of users, mining implicit interests of the users, depicting interest portraits of the users and dynamically updating the interest portraits in real time. Through the parallelism and randomness of quantum technology, the response capability of the strategy to sudden hot spots and changes in user demands is improved, the reading promotion strategy continuously maintains high efficiency in a dynamic environment, the quality and effect of reading promotion are continuously improved through continuous optimization of the promotion strategy, and the wisdom library is developed in the direction of being more intelligent and more self-adaptive.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of promotion management, and in particular to a smart library reading promotion management method based on a knowledge graph. BACKGROUND

[0002] In the field of smart libraries, existing reading promotion management methods are usually based on traditional machine learning or rule engines, and construct interest models by analyzing user explicit behavior data, and make resource recommendations in combination with the basic associations of the knowledge graph. Although this method can achieve basic reading promotion functions, it has obvious limitations in dealing with user implicit interest mining, dynamic interest evolution, and complex scene adaptation. Firstly, it is difficult to effectively capture user behavior associations across platforms and space-time, and it is difficult to deeply mine implicit needs such as cross-disciplinary interests, resulting in recommended content being limited to the fields corresponding to user explicit behavior, and lacking the ability to explore potential interests. Secondly, in the resource allocation optimization process, traditional algorithms are inefficient in dealing with high-dimensional associations of the knowledge graph, and lack a dynamic feedback mechanism, making it difficult to adapt to real-time changes in user needs and promotion scenarios. Thirdly, the evaluation system and strategy updating mechanism are relatively static, and cannot update the knowledge graph and algorithm parameters in real time, and the fusion ability of multi-modal feedback data is insufficient, resulting in poor adaptability and stability of the promotion strategy in complex environments. SUMMARY

[0003] In view of the problems existing in the prior art, the purpose of the present application is to provide a smart library reading promotion management method based on a knowledge graph to solve the problems raised in the background art.

[0004] To achieve the above-mentioned purpose, the present application provides a smart library reading promotion management method based on a knowledge graph, comprising the following steps:

[0005] S1, obtain library resources, user behavior, and scene data to construct a knowledge graph, provide semantic association basis for intelligent recommendation, and intelligent matching of library resources and user needs;

[0006] S2, dynamically adjust the knowledge association weight, improve the accuracy, self-adaptation ability and intelligent decision-making of the promotion strategy of the system;

[0007] S3, analyze user multi-dimensional behavior data, mine user implicit interests, and dynamically update the user interest portrait in real time;

[0008] S4, encode the promotion strategy, generate the optimal promotion combination, and intelligently optimize the allocation of resources;

[0009] S5, set up a feedback mechanism, dynamically adjust the strategy in combination with the feedback information, intelligently evaluate and iteratively optimize the promotion effect;

[0010] S6. Construct an evaluation system and update the knowledge graph and algorithm parameters in real time to enhance the intelligent self-learning ability of the promotion strategy.

[0011] S7 provides users with an intelligent and innovative promotional approach by generating AR scenes and dynamic knowledge maps based on knowledge graphs.

[0012] Preferably, in step S2, dynamically adjusting the knowledge association weights includes the following steps:

[0013] S21. Calculate the entropy of book knowledge and the entropy of user needs, construct a multi-objective fitness function, and decompose and quantify the global entropy change of knowledge association. The formula is:

[0014]

[0015] In the formula, The result of global entropy change calculation for knowledge association. For three-dimensional tensors Tensor information entropy, To measure the true correlation of three-dimensional tensors Errors associated with model predictions, To influence the weight, As a penalty factor, The sparsity value that affects the weights;

[0016] S22. Utilizing the quantum tunneling effect to correlate interdisciplinary correlation strengths and combining it with chaotic sequences, highly diverse initial weight solutions are generated, breaking through local optima and deeply exploring the correlation between reading resources and users. This supports precise and diversified reading promotion strategies. The formula is:

[0017]

[0018] In the formula, For users For resources The initial degree of correlation, For users For resources The basic correlation strength, To measure the quantum tunneling temperature threshold that determines the ability to break through local optima, It is a chaotic sequence;

[0019] S23. Construct a non-uniform temperature field to drive the mutation probability. Based on the weight configuration and the disorder factor of the solution, dynamically adjust the mutation probability to optimize the resource recommendation and reading promotion effect. The formula is:

[0020]

[0021] In the formula, The probability of mutation during the optimization of reading promotion strategies. This represents the gradient difference between the current recommendation weight and the ideal optimal recommendation weight. For location-dependent temperature fields, The coefficient of variation is the sensitivity coefficient. It is a tiny path element vector;

[0022] S24. Using fiber bundle theory, through the mapping between Lie algebras and solution manifolds and local optimization operators, the derivation process of the collaborative optimization reading promotion strategy is described by the following formula:

[0023]

[0024] In the formula, Genetic Algorithm exist The weight matrix at time step, This represents the mapping from Lie algebras to solution manifolds. for The reverse, For local optimization operators, For particle swarm optimization algorithm exist The weight matrix at each time step;

[0025] S25. Combining drift, diffusion, and jump terms following a Poisson distribution, this characterizes the continuous changes in the relationships between knowledge entities over time, while also incorporating the impact of sudden trending events. This allows for precise capture and strategy adaptation of dynamic relationships in reading promotion. The formula is:

[0026]

[0027] In the formula, For users For resources The correlation of small changes, It is a drift term, and , for Time users For resources The strength of the association, For time intervals, For a tiny time interval, The diffusion coefficient is used to measure the degree of random fluctuation in the relationship between knowledge entities. To drive the small increments of random perturbation factors that cause correlation changes, To measure the coefficient of the jump term in an emergency, This represents the slight increase in the average number of sudden hotspot events occurring per unit of time. For users For resources the associated state quantity.

[0028] Preferably, in the step S3, the evaluation and optimization include the following steps:

[0029] S31, integrate the user's spatial behavior and timing behavior across platforms, capture the spatio-temporal coupling effect by using tensor product, construct high-dimensional interest representation, and accurately depict user interest;

[0030] S32, combine quantum entangled state, fuse basic interest state and co-reading state, combine interest energy level difference and quantum tunneling mechanism, mine user's hidden, cross-domain nonlinear interest, and promote personalized reading for users, formula:

[0031]

[0032] In the formula, is the hidden interest entangled state, is the total number of basic interest states or co-reading state related states participating in the calculation, is the quantum state representation describing the user's basic interest, is the tensor product, is the co-reading state, is the quantumization deformation of the Boltzmann distribution;

[0033] S33, map high-dimensional user behavior data to an interpretable interest manifold, dynamically regulate the compactness of different users and interest clusters, formula:

[0034]

[0035] In the formula, is the high-dimensional behavior data tensor of the object to be processed, and is the product operation of the high-dimensional behavior data tensor and the interest matrix in a specific dimension, is a regularization coefficient for controlling the influence degree on the optimization target, is the manifold curvature for controlling the compactness of the interest cluster regularization term;

[0036] S34, fuse the spatio-temporal attention mechanism, dynamically adapt the interest portrait dynamic evolution of user short-term hot response and long-term reading preference, formula:

[0037]

[0038] In the formula, is the fractional derivative reflecting the historical cumulative influence of interest evolution, is the user interest vector, is the interest decay coefficient, The coefficient representing the influence of attention mechanisms. and These are the query and key-cast matrices, respectively. Associate tensors with users;

[0039] S35. Verify the model error rate based on historical data, dynamically adjust the fuzzy rules and membership function parameters, and complete the intelligent update of the requirement model.

[0040] Preferably, in step S31, accurately depicting the reader's interests includes the following steps:

[0041] S311. Collect spatial and temporal behaviors from multiple systems in the library, map all behaviors onto a behavior manifold, and describe continuous behavior trajectories using the following formula:

[0042]

[0043] In the formula, For users To behavior The shortest behavioral trajectory, The curvature used to characterize the difficulty of an action. The intensity of the occurrence of the behavior, These are constraints;

[0044] S312. Introduce supersymmetric variables to spatial behavior, distinguish between real behavior and potential behavior, and use non-standard analysis to describe temporal behavior. Limit the discrete time window of the graph to capture the behavioral correlation of infinitesimal time intervals.

[0045] S313. Integrate all behavioral fragments to construct a high-dimensional interest representation, characterizing the user's interest direction. The formula is:

[0046]

[0047] In the formula, To represent users' high-dimensional interests, The total number of behavioral dimensions. For the first Credibility weight of the behavior For the first Type of spatial behavior matrix, For the first Type of temporal behavior matrix, It is the tensor product.

[0048] Preferably, in step S4, intelligent optimization uses a quantum neural network to encode resource allocation, combined with a knowledge graph-driven target tensor, to adapt to the high-dimensional associations of the knowledge graph under the constraints of balancing error and allocation entropy, thereby achieving efficient reading promotion. The formula is as follows:

[0049]

[0050] wherein, is a resource allocation matrix, is a tensor and matrix related to promotion strategy, is a quantum neural network output quantum-optimized resource allocation mapping, is a target tensor, is an entropy weight coefficient, is an entropy of resource allocation.

[0051] Preferably, in the step S5, the intelligent evaluation and iterative optimization comprises the following steps:

[0052] S51, encode the user feedback as the measurement result of the quantum state, and evaluate the promotion effect by using quantum state tomography, and the formula is:

[0053]

[0054] wherein, is a feedback density matrix of the promotion effect quantum state mixture, is the number of feedback types, is a feedback ground state, is the conjugate transpose of, is a quantum measurement operator for projecting and measuring the user behavior tensor ;

[0055] S52, combine the feedback information, and use a quantum stochastic microequation to describe the dynamic adjustment of the strategy tensor, and the formula is:

[0056]

[0057] wherein, is a strategy tensor for adjusting the promotion strategy over time differential change, is a coefficient of natural decay of the strategy without feedback, is a knowledge graph guided item, is a random fluctuation in the simulation of promotion, is a control of the influence of random fluctuation on the strategy, is a feedback learning rate, is a feedback difference operator;

[0058] S53, compare the matching degree of the current strategy and the knowledge graph, and evaluate the optimality of the promotion effect by using quantum variational inequality, when the inequality is established, the current strategy is locally optimal, and the formula is:

[0059]

[0060] wherein, is the ideal strategy driven by the knowledge graph, is the set of strategies constrained by resources and channels, is the composite operator containing the knowledge graph association, feedback error, and resource cost;

[0061] S54, taking the knowledge graph association as the reward function, using quantum reinforcement learning to complete the iterative optimization of the strategy, and through the superposition of quantum states, exploring popular promotion and hidden association promotion at the same time, the formula is:

[0062]

[0063] wherein, and are the updated and pre-updated quantum policy parameters, is the learning rate, is the quantum reward operator of the driven promotion strategy of the knowledge graph is the strategy probability distribution.

[0064] Preferably, in the step S6, updating the knowledge graph and algorithm parameters includes the following steps:

[0065] S61, synchronously measuring the promotion effect and the quality of the knowledge graph, using quantum information entropy to construct a multi-dimensional evaluation index, evaluating the uncertainty of the promotion effect and measuring the complexity of the knowledge association, the formula is:

[0066]

[0067] wherein, is the multi-dimensional evaluation index parameter, , and are the corresponding weight coefficients, the von Neumann entropy of the feedback density matrix, is the quantum entanglement entropy of the knowledge graph, is the quantum Fisher information;

[0068] S62, fusing user feedback and promotion effect by means of quantum state evolution mechanism, using quantum master equation to fuse user feedback and promotion effect, dynamically updating the knowledge graph in real time, stably and adaptively to feedback, the formula is:

[0069]

[0070] wherein, is the evolution rate of the knowledge graph with respect to time , is the Lyapunov operator of the knowledge graph acting on the time​ the state of the knowledge graph at the time, for updating the intensity coefficient, adjusting the quantum update unitary operator of the internal association of the knowledge graph according to user behavior, for the partial trace operation;

[0071] S63, dynamically optimize algorithm parameters according to evaluation indicators, adopt quantum meta-learning to quickly adapt to new tasks of wisdom library reading promotion, prevent overfitting by quantum regularization, ensure the adaptability of promotion effect and knowledge graph quality, and the formula is:

[0072]

[0073] wherein, the optimal algorithm parameter obtained after optimization, the original optimal algorithm parameter, the quantum regularization term for preventing overfitting of the algorithm in the reading promotion parameter optimization; according to the distribution take the expectation, the meta-learning update parameter, the loss function of the algorithm in the reading promotion task, the regularization strength, the quantum regularization term for preventing overfitting of the algorithm in the reading promotion parameter optimization;

[0074] S64, process the knowledge graph from the aspects of mining knowledge causal association, aligning multi-modal information, and verifying strategy stability, strengthen the dynamic updating ability, multi-modal fusion effect and strategy robustness of the knowledge graph, and continuously deepen the deep integration of quantum technology and library scene.

[0075] Preferably, in step S63, the meta-learning update parameter is used to quickly adapt to new tasks of reading promotion, and the formula is:

[0076]

[0077] wherein, the learning rate of the meta-learning update parameter for controlling the update step size, the loss function the gradient of the original optimal algorithm parameter .

[0078] Preferably, in step S64, processing the knowledge graph includes the following steps:

[0079] S641, distinguish the causal and false association between the promotion strategy and user behavior, use quantum causal inference to identify the causal association between the promotion strategy and user behavior, and avoid false association interference update, and the formula is:

[0080]

[0081] In the formula, For quantum causality coefficient, For quantum causal unitary operators that propagate causality from encoding strategy to behavior, Quantum mutual information is used to measure the correlation between strategy and behavior. For smart library reading promotion strategies, For user behavior, To extract the real part of a complex number, i.e., the extraction strategy After quantum causal unitary operator Obtain user behavior The real part of the inner product of this quantum state;

[0082] S642. By superimposing quantum states, align user textual, spatial, and temporal multimodal feedback data to resolve modal conflicts and accurately update the knowledge graph. The formula is:

[0083]

[0084] In the formula, To unify the updating of the multimodal superposition state of the knowledge graph, Modal weights, which reflect the degree of influence of different modes when superimposed, To verify the multimodal feedback state and knowledge graph subspace Quantum exchange test of alignment This represents the total number of modes in the multimodal data. For modal index variables;

[0085] S643. Extreme scenarios are simulated using quantum random walks to verify the robustness of the reading promotion strategy, avoiding the strategy's applicability only to conventional scenarios and ensuring its effective operation in unpopular and unexpected situations. The formula is:

[0086]

[0087] In the formula, To determine the robustness score for policy stability, Strategies for Promoting Reading in Smart Libraries The adjoint operator, Quantum robustness operators that apply extreme scenario interference to generalization strategies This is the worst-case density matrix. For trace-finding operations.

[0088] The beneficial effects of the knowledge graph-based smart library reading promotion and management method provided by this invention are as follows:

[0089] 1. By integrating user cross-platform spatial and temporal behavior, using tensor product to capture spatio-temporal coupling effect, combining quantum entangled state, interest manifold dimension reduction and spatio-temporal attention mechanism, the deep mining and dynamic characterization of user implicit interest are realized, the cross-domain interest preference that is not explicitly expressed by the user can be accurately identified, and the real-time updated interest portrait can adapt to the dynamic changes of the reading habit of the reader, so that the reading promotion is upgraded from the extensive recommendation based on explicit behavior to the accurate matching based on implicit interest, the relevance of the recommendation and the user satisfaction are significantly improved, and a more scientific decision basis is provided for personalized reading promotion.

[0090] 2. By means of quantum neural network coding resource allocation, combining knowledge graph target tensor to optimize resource allocation, and through quantum state chromatography, stochastic differential equation, variational inequality and reinforcement learning to construct feedback iteration mechanism, the intelligent optimization of resource allocation is realized, the optimal promotion combination can be quickly found in the complex knowledge graph association, and the strategy is dynamically adjusted according to the user feedback, which can effectively balance the popular resource promotion and the cold knowledge mining, maximize the promotion effect under the condition of limited resources, and through the parallelism and randomness of quantum technology, the response ability of the strategy to sudden hot spots and changes in user demand is improved, so that the reading promotion strategy can continuously maintain high efficiency in the dynamic environment.

[0091] 3. By constructing a multi-dimensional evaluation system based on quantum information entropy, combining quantum master equation, meta-learning, causal mining, cross-modal alignment, robustness verification and other technologies, the real-time updating of knowledge graph and algorithm parameters is realized, the intelligent self-learning ability of the promotion strategy is improved, the real causal relationship between the promotion strategy and user behavior can be automatically identified, multi-modal feedback data is fused, the dynamic nature and accuracy of the knowledge graph are enhanced, the stability of the strategy is verified through simulation of extreme scenarios, and the promotion strategy is ensured to run effectively in various complex situations, so that the intelligent library system can adapt to the evolution of the knowledge graph and the changes of user demand, continuously optimize the promotion strategy, continuously improve the quality and effect of reading promotion, and promote the intelligent library to develop in the direction of more intelligent and more adaptive. BRIEF DESCRIPTION OF DRAWINGS

[0092] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, a brief introduction will be given below to the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creating any inventive labor.

[0093] Figure 1 A process framework schematic diagram of a knowledge graph-based intelligent library reading promotion management method provided by the present application. DETAILED DESCRIPTION

[0094] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0095] like Figure 1 As shown, this embodiment proposes a smart library reading promotion and management method based on knowledge graphs, including the following steps:

[0096] S1. Acquire library resources, user behavior, and scenario data to construct a knowledge graph, providing a semantic association foundation for intelligent recommendation and intelligent matching of book resources with user needs;

[0097] S2. Dynamically adjust the knowledge association weights to improve the accuracy, adaptability, and intelligent decision-making of promotion strategies for the system.

[0098] S3. Analyze multi-dimensional user behavior data, uncover users' latent interests, create interest profiles for users, and update them dynamically in real time.

[0099] S4. Encode promotion strategy to generate the optimal promotion combination and intelligently optimize resource allocation;

[0100] S5. Establish a feedback mechanism to dynamically adjust strategies based on feedback information, and conduct intelligent evaluation and iterative optimization of promotion effects;

[0101] S6. Construct an evaluation system and update the knowledge graph and algorithm parameters in real time to enhance the intelligent self-learning ability of the promotion strategy.

[0102] S7 provides users with an intelligent and innovative promotional approach by generating AR scenes and dynamic knowledge maps based on knowledge graphs.

[0103] In this embodiment, step S2, dynamically adjusting the knowledge association weights, includes the following steps:

[0104] S21. Calculate the entropy of book knowledge and the entropy of user needs, construct a multi-objective fitness function, and decompose and quantify the global entropy change of knowledge association. The formula is:

[0105]

[0106] In the formula, The result of global entropy change calculation for knowledge association. For three-dimensional tensors Tensor information entropy, To measure the true correlation of three-dimensional tensors Errors associated with model predictions, To influence the weight, As a penalty factor, The sparse value of the influence weight;

[0107] S22, using quantum tunneling effect correlation cross-disciplinary correlation strength, combined with chaotic sequence, generate high diversity initial weight solution, break through local optimization, deep mining reading resources and user association, support accurate and multiple reading promotion strategy, formula is:

[0108]

[0109] In the formula, The initial association degree of the user For resources , The basic association strength of the user For resources , The quantum tunneling temperature threshold for measuring the ability to break through local optimization, Chaotic sequence;

[0110] S23, construct non-uniform temperature field driven variation probability, according to the weight configuration, solution chaos degree factor, dynamic adjustment of mutation probability, optimization of resource recommendation and reading promotion effect, formula is:

[0111]

[0112] In the formula, The probability of mutation in the reading promotion strategy optimization process, The gradient difference between the current recommended weight and the ideal optimal recommended weight, The position related temperature field, The mutation sensitive coefficient, The micro path vector;

[0113] S24, with the help of fiber bundle theory, through the mapping of Lie algebra and solution manifold and local optimization operator, collaborative optimization of reading promotion strategy calculation process, formula is:

[0114]

[0115] In the formula, The weight matrix of genetic algorithm At Time, The mapping relationship from Lie algebra to solution manifold, The inverse of , Local optimization operator, The weight matrix of particle swarm algorithm At Time;

[0116] S25, combined with drift, diffusion and jump term obeying Poisson distribution, describes the continuous change of knowledge entity association over time, while taking into account the influence of sudden hot events, and accurately captures and adapts to the strategy of reading promotion dynamic association, the formula is:

[0117]

[0118] In the formula, is the association of user to resource , is the drift term, and , is the association strength of user to resource at time , is the time interval, is the diffusion coefficient measuring the degree of random fluctuation of the change of knowledge entity association, is the small increment of the random disturbance factor driving the change of association, is the jump term coefficient measuring the sudden event, is the small increment of the average number of sudden hot events occurring per unit time, is the association state of user to resource .

[0119] In this embodiment, in step S3, evaluation and optimization include the following steps:

[0120] S31, integrate user spatial behavior and time sequence behavior across platforms, capture spatio-temporal coupling effect by using tensor product, construct high-dimensional interest representation, and accurately describe user interest;

[0121] S32, combined with quantum entangled state, fusion basic interest state and co-reading state, combined with interest energy level difference and quantum tunneling mechanism, to explore user hidden, cross-domain nonlinear interest, for user personalized reading promotion, the formula is:

[0122]

[0123] In the formula, is the hidden interest entangled state, is the total number of basic interest states or co-reading state related states involved in the calculation, is the quantum state representation describing the user's basic interest, is the tensor product, is the co-reading state, is the quantumization deformation of the Boltzmann distribution;

[0124] S33, dimensionality reduction of high-dimensional user behavior data is mapped to an interpretable interest manifold, and the tightness of different users and interest clusters is dynamically regulated, and the formula is:

[0125]

[0126] In the formula, is the high-dimensional behavior data tensor of the object to be dimensionally reduced, and is the product operation of the high-dimensional behavior data tensor and the interest matrix of a specific dimension, is a regularization coefficient for controlling the influence degree on the optimization target, is a manifold curvature for controlling the tightness of the interest cluster regularization term;

[0127] S34, fusion of space-time attention mechanism, dynamic adaptation of user short-term hot spot response and long-term reading preference interest portrait dynamic evolution, formula:

[0128]

[0129] In the formula, is the fractional derivative reflecting the historical cumulative influence of interest evolution, is the user interest vector, is the interest decay coefficient, is the attention mechanism influence coefficient, and are the query and key projection matrices respectively, is the user associated tensor;

[0130] S35, based on historical data to verify the model error rate, dynamically adjust the fuzzy rules and membership function parameters, and complete the intelligent update of the demand model.

[0131] In this embodiment, in step S31, the reader interest is accurately described, including the following steps:

[0132] S311, collect spatial behavior and time series behavior from multiple library systems, map all behaviors to a behavior manifold, and describe continuous behavior trajectories, formula:

[0133]

[0134] In the formula, is the shortest behavior trajectory of the user to the behavior , is the curvature describing the difficulty of behavior occurrence, is the occurrence intensity of the behavior, is the constraint condition;

[0135] S312. Introduce supersymmetric variables to spatial behavior, distinguish between real behavior and potential behavior, and use non-standard analysis to describe temporal behavior. Limit the discrete time window of the graph to capture the behavioral correlation of infinitesimal time intervals.

[0136] S313. Integrate all behavioral fragments to construct a high-dimensional interest representation, characterizing the user's interest direction. The formula is:

[0137]

[0138] In the formula, To represent users' high-dimensional interests, The total number of behavioral dimensions. For the first Credibility weight of the behavior For the first Type of spatial behavior matrix, For the first Type of temporal behavior matrix, It is the tensor product.

[0139] Specifically, this invention integrates users' cross-platform spatial and temporal behaviors, utilizes tensor products to capture spatiotemporal coupling effects, and combines quantum entangled states, interest manifold dimensionality reduction, and spatiotemporal attention mechanisms to achieve in-depth mining and dynamic characterization of users' implicit interests. It can accurately identify users' unexpressed cross-domain interest preferences, and the real-time updated interest profiles can adapt to the dynamic changes in readers' reading habits. This upgrades reading promotion from extensive recommendations based on explicit behavior to precise matching based on implicit interests, significantly improving the relevance of recommendations and user satisfaction, and providing a more scientific decision-making basis for personalized reading promotion.

[0140] In this embodiment, in step S4, intelligent optimization uses a quantum neural network to encode resource allocation, combined with a knowledge graph-driven target tensor. Under the constraints of balancing error and allocation entropy, it adapts to the high-dimensional associations of the knowledge graph to achieve efficient reading promotion. The formula is:

[0141]

[0142] In the formula, For resource allocation matrix, For the tensors and matrices related to the promotion strategy, To output a quantum-optimized resource allocation map for the quantum neural network. For the target tensor, For entropy weighting coefficients, Entropy for resource allocation.

[0143] In this embodiment, step S5, intelligent evaluation and iterative optimization includes the following steps:

[0144] S51, encode the user feedback as the measurement result of the quantum state, and evaluate the promotion effect using quantum state tomography, the formula is:

[0145]

[0146] wherein, is the feedback density matrix of the promotion effect quantum state mixture, is the number of feedback types, is the feedback ground state, is the conjugate transpose of, is the quantum measurement operator for projecting the user behavior tensor measurement;

[0147] S52, combined with the feedback information, use quantum stochastic differential equation to describe the dynamic adjustment of the strategy tensor, the formula is:

[0148]

[0149] wherein, is the strategy tensor for adjusting the promotion strategy over time differential change, is the natural decay coefficient of the strategy without feedback, is the knowledge graph guide item, is the random fluctuation in the simulation of promotion, is the control of the influence of random fluctuations on the strategy, is the feedback learning rate, is the feedback difference operator;

[0150] S53, compare the matching degree of the current strategy and the knowledge graph, and evaluate the optimality of the promotion effect using quantum variational inequality, when the inequality is established, the current strategy is locally optimal, the formula is:

[0151]

[0152] wherein, is the ideal strategy driven by the knowledge graph, is the strategy set constrained by resources and channels, is a composite operator containing knowledge graph association, feedback error, and resource cost;

[0153] S54, take the knowledge graph association as the reward function, and complete the iterative optimization of the strategy using quantum reinforcement learning, and explore popular promotion and hidden association promotion simultaneously through the superposition of quantum states, the formula is:

[0154]

[0155] wherein, and These are the quantum policy parameters before and after the update, respectively. For learning rate, For knowledge graphs Quantum reward operators driving promotion strategies, This represents the policy probability distribution.

[0156] Specifically, this invention utilizes quantum neural networks to encode resource allocation, combines knowledge graph target tensors to optimize resource configuration, and constructs a feedback iteration mechanism through quantum state tomography, stochastic differential equations, variational inequalities, and reinforcement learning. This achieves intelligent optimization of resource allocation, enabling the rapid identification of optimal promotion combinations within complex knowledge graph relationships. Furthermore, it dynamically adjusts strategies based on user feedback, effectively balancing the promotion of popular resources with the mining of niche knowledge, maximizing promotional effectiveness with limited resources. The parallelism and randomness of quantum technology enhance the strategy's responsiveness to sudden hot topics and changes in user needs, ensuring the reading promotion strategy remains highly efficient in dynamic environments.

[0157] In this embodiment, step S6, updating the knowledge graph and algorithm parameters includes the following steps:

[0158] S61. Simultaneously measure the promotion effect and the quality of the knowledge graph. Use quantum information entropy to construct a multi-dimensional evaluation index to assess the uncertainty of the promotion effect and measure the complexity of knowledge associations. The formula is:

[0159]

[0160] In the formula, For multidimensional evaluation index parameters, , and For the corresponding weighting coefficients, The von Neumann entropy of the feedback density matrix, The quantum entanglement entropy of the knowledge graph. For quantum Fisher information;

[0161] S62. Integrating User Feedback and Promotion Effects: Leveraging the quantum state evolution mechanism and utilizing the quantum master equation, user feedback and promotion effects are integrated to achieve real-time, stable, and dynamic updates of the knowledge graph that adapt to feedback. The formula is as follows:

[0162]

[0163] In the formula, For knowledge graphs at any time The rate of evolution, For the Lyapunov operator of the knowledge graph, the action at time... The state of the knowledge graph at that time. to update the intensity coefficient, adjusting the quantum update unitary operator of the internal association of the knowledge graph according to user behavior, for partial trace operation;

[0164] S63, dynamically optimize algorithm parameters according to evaluation indicators, adopt quantum meta-learning to quickly adapt to new tasks of wisdom library reading promotion, prevent overfitting by quantum regularization, ensure the adaptability of promotion effect and knowledge graph quality, and the formula is:

[0165]

[0166] In the formula, is the optimal algorithm parameter obtained after optimization, is the original optimal algorithm parameter, is the optimal algorithm parameter obtained after optimization, according to the distribution take the expectation, is the meta-learning update parameter, the loss function of the algorithm in the reading promotion task, is the regularization strength, is the quantum regularization term to prevent overfitting of the algorithm in the reading promotion parameter optimization;

[0167] S64, respectively from mining knowledge causal association, aligning multi-modal information, verifying strategy stability, processing knowledge graph, strengthening knowledge graph dynamic updating ability, multi-modal fusion effect and strategy robustness, and continuously deepening the deep integration of quantum technology and library scene.

[0168] In this embodiment, in step S63, the meta-learning update parameter is used to quickly adapt to new tasks of reading promotion, and the formula is:

[0169]

[0170] In the formula, is the learning rate of the meta-learning update parameter, is the loss function the gradient of the original optimal algorithm parameter .

[0171] In this embodiment, in step S64, processing the knowledge graph includes the following steps:

[0172] S641, distinguish the causal and false association of promotion strategy and user behavior, use quantum causal inference to identify the causal association of promotion strategy and user behavior, and avoid false association interference update, and the formula is:

[0173]

[0174] In the formula, For quantum causality coefficient, For quantum causal unitary operators that propagate causality from encoding strategy to behavior, Quantum mutual information is used to measure the correlation between strategy and behavior. For smart library reading promotion strategies, For user behavior, To extract the real part of a complex number, i.e., the extraction strategy After quantum causal unitary operator Obtain user behavior The real part of the inner product of this quantum state;

[0175] S642. By superimposing quantum states, align user textual, spatial, and temporal multimodal feedback data to resolve modal conflicts and accurately update the knowledge graph. The formula is:

[0176]

[0177] In the formula, To unify the updating of the multimodal superposition state of the knowledge graph, Modal weights, which reflect the degree of influence of different modes when superimposed, To verify the multimodal feedback state and knowledge graph subspace Quantum exchange test of alignment This represents the total number of modes in the multimodal data. For modal index variables;

[0178] S643. Extreme scenarios are simulated using quantum random walks to verify the robustness of the reading promotion strategy, avoiding the strategy's applicability only to conventional scenarios and ensuring its effective operation in unpopular and unexpected situations. The formula is:

[0179]

[0180] In the formula, To determine the robustness score for policy stability, Strategies for Promoting Reading in Smart Libraries The adjoint operator, Quantum robustness operators that apply extreme scenario interference to generalization strategies This is the worst-case density matrix. For trace-finding operations.

[0181] Specifically, the application realizes real-time updating of the knowledge graph and algorithm parameters by constructing a multi-dimensional evaluation system based on quantum information entropy, combining quantum master equation, meta-learning, causal mining, cross-modal alignment, robustness verification and other technologies, improves the intelligent self-learning ability of the promotion strategy, can automatically identify the real causal relationship between the promotion strategy and the user behavior, fuses multi-modal feedback data, enhances the dynamics and accuracy of the knowledge graph, verifies the stability of the strategy through simulation of extreme scenarios, ensures the effective operation of the promotion strategy under various complex conditions, enables the intelligent library system to adapt to the evolution of the knowledge graph and the changes of the user demand, continuously optimizes the promotion strategy, continuously improves the quality and effect of the reading promotion, and promotes the intelligent library to develop in the direction of being more intelligent and more adaptive.

[0182] The above embodiments are only used to illustrate the present application, but not to limit the present application. Although the present application is explained in detail with reference to the embodiments, those skilled in the art should understand that various combinations, modifications or equivalent replacements of the technical solutions of the present application do not deviate from the spirit and scope of the present application, and should be covered in the scope of the claims of the present application.

Claims

1.A knowledge graph-based intelligent library reading promotion management method, characterized in that, Comprise the following steps: S1, obtain the collection resources, user behavior, scene data, build knowledge graph, provide semantic association basis and intelligent matching of book resources and user demand for intelligent recommendation; S2, dynamically adjust the knowledge correlation weight, improve the accuracy, self-adaptability and intelligent decision of the promotion strategy of the system; S3, analyze multi-dimensional behavior data of users, mine implicit interests of users, and dynamically update interest portraits of users; S4, encode the promotion strategy, generate the optimal promotion combination, and intelligently optimize the allocation of resources; S5, set up a feedback mechanism, dynamically adjust the strategy combined with feedback information, and intelligently evaluate and iteratively optimize the promotion effect; S6, build an evaluation system, update the knowledge graph and algorithm parameters in real time, and improve the intelligent self-learning ability of the promotion strategy; S7, based on the knowledge graph, generate AR scene and dynamic knowledge map intelligent interaction scheme, and provide intelligent and innovative promotion methods for users; In step S2, dynamically adjusting the knowledge correlation weight comprises the following steps: S21, calculate the book knowledge entropy and user demand entropy, build a multi-objective fitness function, decompose and quantify the global entropy change of knowledge correlation, and the formula is: ； wherein, is a global entropy change accounting result of knowledge association, is a three-dimensional tensor of tensor information entropy, is a real association three-dimensional tensor error associated with model prediction association, is an influence weight, is a penalty factor, is a sparsity value of the influence weight; S22, use quantum tunneling effect to associate the cross-disciplinary correlation strength, combine chaotic sequence to generate high diversity initial weight solution, break through local optimum, deeply mine the correlation between reading resources and users, support accurate and diversified reading promotion strategy, and the formula is: ； In the formula, for a user for a resource an initial correlation degree, for a user for a resource a basic correlation strength, a quantum tunneling temperature threshold for measuring the ability to break through local optimality, a chaotic sequence; S23, build a non-uniform temperature field to drive variation probability, dynamically adjust the variation probability according to weight configuration and solution chaos degree factors, optimize resource recommendation and reading promotion effect, and the formula is: ； In the formula, is the probability of mutation in the reading promotion strategy optimization process, is the gradient gap between the current recommended weight and the ideal optimal recommended weight, is the position-related temperature field, is the mutation sensitivity coefficient, is the micro-path element vector; S24, use fiber bundle theory, map and local optimization operator of Lie algebra and solution manifold, and cooperatively optimize the calculation process of reading promotion strategy, and the formula is: ； wherein is a genetic algorithm at the weight matrix at time instant is a local optimization operator is a particle swarm algorithm at the weight matrix at time instant S25, combine drift, diffusion and jump items subject to Poisson distribution to depict the continuous change of knowledge entity correlation over time, and incorporate the influence of sudden hot events to accurately capture the dynamic correlation of reading promotion and adapt the strategy, and the formula is: ； In the formula, For users For resources The correlation of small changes, It is a drift term, and , for Time users For resources The strength of the association, For time intervals, For a tiny time interval, The diffusion coefficient is used to measure the degree of random fluctuation in the relationship between knowledge entities. To drive the small increments of random perturbation factors that cause correlation changes, To measure the coefficient of the jump term in an emergency, This represents the slight increase in the average number of sudden hotspot events occurring per unit of time. For users For resources The associated state variables. 2.The knowledge graph-based smart library reading promotion management method of claim 1, wherein, In step S3, evaluation and optimization comprises the following steps: S31, integrate user spatial behavior and time sequence behavior across platforms, use tensor product to capture spatio-temporal coupling effect, build high-dimensional interest representation, and accurately depict user interest; S32, combine quantum entangled state, fuse basic interest state and co-reading state, combine interest energy level difference and quantum tunneling mechanism, mine hidden, cross-domain nonlinear interest of users, and formula for personalized reading promotion of users: ； wherein, is a hidden interest entangled state, is the total number of underlying interest states or co-reading state correlated states participating in the computation, is a quantum state representation describing the underlying interests of the user, is a tensor product, is a co-reading state, is a quantumization of the Boltzmann distribution; S33, reduce high-dimensional user behavior data to an interpretable interest manifold, dynamically regulate the compactness of different users and interest clusters, and the formula is: ； In the formula, is a high-dimensional behavior data tensor of an object to be processed by dimension reduction, and is a high-dimensional behavior data tensor and an interest matrix are subjected to product operation of a specific dimension, is a regularization coefficient for controlling the degree of influence on the optimization target, is a manifold curvature for controlling the compactness of the interest cluster regularization term S34, fuse spatio-temporal attention mechanism, dynamically adapt the dynamic evolution of interest portrait of user short-term hot response and long-term reading preference, and the formula is: ； wherein, is the fractional derivative embodying the historical cumulative influence of interest evolution, is the user interest vector, is the interest decay coefficient, is the attention mechanism influence coefficient, and are the query and key projection matrices, respectively, is the user associated tensor; S35, verify the model error rate based on historical data, dynamically adjust fuzzy rules and membership function parameters, and complete intelligent update of demand model. 3.The knowledge graph-based wisdom library reading promotion management method of claim 2, wherein, In step S31, accurately depicting reader interest comprises the following steps: S311. Collect spatial and temporal behaviors from multiple systems in the library, map all behaviors onto a behavior manifold, and describe continuous behavior trajectories using the following formula: ； wherein, is a user to the behavior the shortest behavior trajectory, is a curvature that characterizes the difficulty of the behavior to occur, is the occurrence intensity of the behavior, is a constraint condition; S312. Introduce supersymmetric variables to spatial behavior, distinguish between real behavior and potential behavior, and use non-standard analysis to describe temporal behavior. Limit the discrete time window of the graph to capture the behavioral correlation of infinitesimal time intervals. S313. Integrate all behavioral fragments to construct a high-dimensional interest representation, characterizing the user's interest direction. The formula is: ； wherein, is a high-dimensional representation of the user's interest, is the total number of behavior dimensions, is a confidence weight of the th behavior, is a spatial behavior matrix of the th behavior, is a temporal behavior matrix of the th behavior, is a tensor product. 4.The knowledge graph-based wisdom library reading promotion management method of claim 1, wherein, In step S4, intelligent optimization uses a quantum neural network to encode resource allocation, combined with a knowledge graph-driven target tensor, and adapts to the high-dimensional associations of the knowledge graph under the constraints of balancing error and allocation entropy, to achieve efficient reading promotion. The formula is as follows: ； wherein, is a resource allocation matrix, are tensors and matrices related to promotion strategies, is a quantum neural network outputting a quantum-optimized resource allocation mapping, is a target tensor, is an entropy weight coefficient, is an entropy of resource allocation. 5.The knowledge graph-based wisdom library reading promotion management method of claim 3, wherein, In step S5, intelligent evaluation and iterative optimization include the following steps: S51. Encode user feedback into measurement results of quantum states, and use quantum state tomography to evaluate the promotion effect. The formula is: ； wherein is the feedback density matrix for promoting the effect of quantum state mixing, is the feedback type number, is the feedback ground state, is the is the conjugate transpose of is the quantum measurement operator for the projection measurement on the user behavior tensor is the quantum measurement operator for the projection measurement on the user behavior tensor S52. Combining feedback information, the dynamic adjustment of the strategy tensor is described using quantum stochastic microequations, with the following formula: ； wherein, a policy tensor to adjust promotion strategy over time a differential change, a coefficient for natural decay of policy without feedback, a knowledge graph guided item, a random fluctuation in simulated promotion, a control of the impact of random fluctuations on the policy, a feedback learning rate, a feedback difference operator; S53. Compare the matching degree between the current strategy and the knowledge graph, and use the quantum variational inequality to evaluate the optimality of the promotion effect. When the inequality holds, the current strategy is locally optimal. The formula is: ； wherein, is a knowledge graph driven ideal strategy, is a set of resource and channel constrained strategies, is a composite operator including knowledge graph correlation, feedback error, and resource cost. S54. Using knowledge graph associations as the reward function, quantum reinforcement learning is employed to iteratively optimize the strategy. Leveraging the superposition property of quantum states, both popular referencing and hidden association referencing are explored simultaneously. The formula is: ； In the formula, and respectively, the quantum policy parameters after and before updating, is a learning rate, is a knowledge graph a quantum reward operator of the driving promotion strategy, is a policy probability distribution. 6.The knowledge graph-based wisdom library reading promotion management method of claim 5, wherein, In step S6, updating the knowledge graph and algorithm parameters includes the following steps: S61. Simultaneously measure the promotion effect and the quality of the knowledge graph. Use quantum information entropy to construct a multi-dimensional evaluation index to assess the uncertainty of the promotion effect and measure the complexity of knowledge associations. The formula is: ； wherein, is a multi-dimensional evaluation index parameter, , and are corresponding weight coefficients, von Loewis entropy of the feedback density matrix, is the quantum entanglement entropy of the knowledge graph, is the quantum Fisher information; S62. Integrating User Feedback and Promotion Effects: Leveraging the quantum state evolution mechanism and utilizing the quantum master equation, user feedback and promotion effects are integrated to achieve real-time, stable, and dynamic updates of the knowledge graph that adapt to feedback. The formula is as follows: ； wherein is the evolution rate of the knowledge graph at time is the state of the knowledge graph at time is the update strength coefficient, is the quantum update unitary operator that adjusts the internal relations of the knowledge graph according to user behavior, is the partial trace operation;​​ S63. Based on the evaluation indicators, dynamically optimize the algorithm parameters, adopt quantum meta-learning to quickly adapt to the new tasks of smart library reading promotion, and use quantum regularization to prevent overfitting, ensuring the adaptability of promotion effect to knowledge graph quality. The formula is: ； In the formula, is the optimal algorithm parameter obtained after optimization, is the original optimal algorithm parameter, is the optimal algorithm parameter obtained after optimization, According to the distribution Take the expectation, is the parameter of the meta-learning update, The loss function of the algorithm in the reading promotion task, is the regularization strength, is a quantum regularization term to prevent overfitting of the algorithm in the reading promotion parameter optimization; S64. By focusing on mining causal relationships in knowledge, aligning multimodal information, and verifying the stability of strategies, knowledge graphs are processed to enhance their dynamic update capabilities, multimodal fusion effects, and strategy robustness, thereby continuously deepening the integration of quantum technology with library scenarios. 7.The knowledge graph-based wisdom library reading promotion management method of claim 6, wherein, In step S63, the meta-learning update parameters are used to quickly adapt to the new task of reading promotion, and the formula is: ； wherein is the learning rate at the time of the meta-learning update for controlling the parameter update step size, is the loss function is the gradient of the original optimal algorithm parameter with respect to the parameter. 8.The knowledge graph-based wisdom library reading promotion management method of claim 6, wherein, In step S64, the processing of the knowledge graph includes the following steps: S641. Distinguish between causal and spurious associations between promotional strategies and user behavior. Employ quantum causal inference to identify the causal relationship between promotional strategies and user behavior, avoiding spurious associations from interfering with updates. The formula is: ； wherein, is the quantum causal coefficient, is the quantum causal unitary operator encoding the strategy-to-behavior causal propagation, is the quantum mutual information measuring the strategy-behavior association, is the wisdom library reading promotion strategy, is the user behavior, is the real part of the complex number, i.e., the extraction strategy by the quantum causal unitary operator gives the user behavior the real part of this quantum state inner product; S642. By superimposing quantum states, align user textual, spatial, and temporal multimodal feedback data to resolve modal conflicts and accurately update the knowledge graph. The formula is: ； In the formula, To unify the update of the multi-modal superposition state of the knowledge graph, To embody the influence degree of different modalities when superimposed, To verify the multi-modal feedback state And the knowledge graph subspace The quantum exchange test of the alignment degree, The total number of modalities of the multi-modal data, The index variable of the modality; S643. Extreme scenarios are simulated using quantum random walks to verify the robustness of the reading promotion strategy, avoiding the strategy's applicability only to conventional scenarios and ensuring its effective operation in unpopular and unexpected situations. The formula is: ； wherein is the robustness score for judging the stability of the strategy, is the quantum robustness operator for the smart library reading promotion strategy is the companion operator, is the quantum robustness operator for applying extreme scenario interference to the promotion strategy, is the worst-case density matrix, is the trace operation.

Images