Learning resource recommendation method and device, electronic equipment and readable storage medium

Abstract

The invention discloses a learning resource recommendation method and apparatus, an electronic device and a readable storage medium. The method comprises the steps of obtaining implicit learning feedback information of a target user; wherein the implicit learning feedback information comprises behavior information of the target user on various learning resources; mapping the implicit learning feedback information to a quantum state space, encoding the implicit learning feedback information into quantum state information, and constructing comprehensive preference state information of the target user for the various learning resources according to quantum characteristics of the quantum state information; wherein the comprehensive preference state information is used for representing the preference degree of the target user on the various learning resources; and according to the comprehensive preference state information of the target user on the various learning resources, performing learning resource recommendation on the target user.

Description

Technical Field

[0001] The present application belongs to the field of computer technology, and specifically relates to a method, device, electronic device, and readable storage medium for recommending learning resources. Background Art

[0002] Current online learning and management platforms have broken down information silos, enabling information sharing and integration. These platforms facilitate personnel management and the timely communication of work schedules and task requirements. However, existing learning and management systems often directly use implicit user feedback (such as browsing behavior and dwell time) to predict user preferences when recommending learning resources. However, certain user behaviors, such as clicks, do not necessarily indicate a preference for a particular content; they may simply be clicks motivated by curiosity or accidental clicks. This can lead to inaccurate predictions of user preferences based on implicit feedback, making it difficult to accurately recommend learning resources. Summary of the Invention

[0003] The embodiments of the present application provide a learning resource recommendation method, device, electronic device, and readable storage medium, which can solve the problem of incorrectly predicting user preferences based on user implicit feedback information, thereby failing to accurately recommend learning resources to users.

[0004] In a first aspect, an embodiment of the present application provides a method for recommending learning resources, the method comprising: Obtaining implicit learning feedback information of a target user; wherein the implicit learning feedback information includes behavior information of the target user on various learning resources; Mapping the implicit learning feedback information into a quantum state space and encoding it into quantum state information, and constructing the target user's comprehensive preference state information for the various types of learning resources based on the quantum properties of the quantum state information; wherein the comprehensive preference state information is used to represent the target user's preference for the various types of learning resources; Recommend learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources.

[0005] In a second aspect, an embodiment of the present application provides a device for recommending learning resources, the device comprising: An acquisition module, configured to acquire implicit learning feedback information of a target user; wherein the implicit learning feedback information includes behavior information of the target user on various learning resources; A construction module is used to map the implicit learning feedback information into a quantum state space and encode it into quantum state information, and to construct the target user's comprehensive preference state information for the various types of learning resources based on the quantum properties of the quantum state information; wherein the comprehensive preference state information is used to represent the target user's preference for the various types of learning resources; The recommendation module is used to recommend learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources.

[0006] In a third aspect, an embodiment of the present application provides an electronic device comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method described in the first aspect.

[0007] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0008] In a fifth aspect, an embodiment of the present application provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the steps of the method described in the first aspect.

[0009] In a sixth aspect, an embodiment of the present application provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes a program or instructions, which, when executed, implement the steps of the method described in the first aspect.

[0010] In an embodiment of the present application, by obtaining the implicit learning feedback information of the target user, the implicit learning feedback information is mapped to the quantum state space and encoded as quantum state information, and according to the quantum characteristics of the quantum state information, the comprehensive preference state information of the target user for the various types of learning resources is constructed, and the user's preference for various types of learning resources is deeply excavated. Based on the determined comprehensive preference state information of the target user for the various types of learning resources, the degree of preference of the target user for the various types of learning resources is clarified. In this way, the implicit learning feedback information is encoded into quantum state information and its quantum characteristics are used to deeply understand the target user's current demand for learning resources from multiple aspects, determine the target user's preference for learning resources, and accurately recommend learning resources to the target user. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a flowchart of a method for recommending learning resources provided in an embodiment of the present application; Figure 2 This is a structural diagram of a learning resource recommendation device provided in an embodiment of the present application; Figure 3 This is a structural diagram of a learning resource recommendation system provided in an embodiment of the present application; Figure 4 This is a flowchart of another method for recommending learning resources provided in an embodiment of the present application; Figure 5 This is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0013] The terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects, and are not used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of this application can be implemented in an order other than that illustrated or described herein, and that the objects distinguished by "first," "second," and the like are generally of the same type, and do not limit the number of objects; for example, the first object can be one or more. In addition, the term "and / or" in the specification and claims refers to at least one of the connected objects, and the character " / " generally indicates that the objects connected are in an "or" relationship.

[0014] Existing online learning and management platforms have broken down information silos and enabled information sharing and integration. However, the following issues remain: (1) Cold-start users usually lack sufficient historical learning behavior data, which makes it difficult for existing learning and management systems to accurately recommend learning resources to cold-start users. Moreover, due to the restrictions of various user privacy protection regulations, without the user's explicit authorization, less user information is available, which exacerbates the difficulty of recommending learning resources to cold-start users.

[0015] (2) When recommending learning resources to users, existing learning and management systems usually use users' implicit feedback (such as their browsing behavior, dwell time, etc.) to directly predict their preferences. However, in reality, some user behaviors, such as clicks, do not necessarily indicate a preference for a particular content. They may simply be clicks out of curiosity or accidental clicks, which leads to incorrect predictions of user preferences based on implicit feedback information, making it impossible to accurately recommend learning resources to users.

[0016] (3) In traditional learning and management systems, highly rated content is often recommended frequently, causing the system to favor popular content while ignoring the personalized needs of users. This bias is particularly detrimental in highly professional or specialized learning. For example, some highly professional learning content is not popular, which can easily lead to some important learning content not being fully displayed, affecting learning outcomes.

[0017] The following, in conjunction with the accompanying drawings, describes in detail the learning resource recommendation method, device, electronic device, and readable storage medium provided in the embodiments of the present application through specific embodiments and their application scenarios.

[0018] Figure 1 The flowchart of a method for recommending learning resources provided by an embodiment of the present application is shown, and the method can be executed by an electronic device. Figure 1 , the method may include the following steps.

[0019] Step 102: Obtain implicit learning feedback information of the target user; wherein the implicit learning feedback information includes the target user's behavior information on various learning resources.

[0020] Target users can include cold-start users or other users whose preferences require further exploration. Implicit learning feedback information may include, but is not limited to, the following: target users' interactive behavior with specific content (e.g., click-through rate, depth of browsing, bookmarking, favorites, etc.), dwell time, frequency of discussion participation, selected learning resources, duration of study using various learning resources, and frequency and duration of activity participation.

[0021] Step 104: Map the implicit learning feedback information to the quantum state space and encode it into quantum state information. Based on the quantum characteristics of the quantum state information, construct the comprehensive preference state information of the target user for the various types of learning resources; wherein the comprehensive preference state information is used to characterize the degree of preference of the target user for the various types of learning resources.

[0022] Implicit learning feedback information characterizes the target user's various learning behaviors, but these behaviors cannot directly reflect the target user's actual preferences. By mapping implicit learning feedback information into quantum state space and leveraging the properties of quantum superposition and quantum entanglement, this quantum state is used to deeply explore the explicit or implicit relationship between the target user's implicit learning feedback information and learning resources. This allows the target user's learning intention to be accurately identified, thereby accurately determining the target user's preference for learning resources.

[0023] Step 106: Recommend learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources.

[0024] In an embodiment of the present application, by obtaining the implicit learning feedback information of the target user, the implicit learning feedback information is mapped to the quantum state space and encoded as quantum state information, and according to the quantum characteristics of the quantum state information, the comprehensive preference state information of the target user for the various types of learning resources is constructed, such as quantum superposition and quantum entanglement, to deeply explore the user's preferences for various types of learning resources. Based on the determined comprehensive preference state information of the target user for the various types of learning resources, the degree of preference of the target user for the various types of learning resources is clarified. In this way, the implicit learning feedback information is encoded into quantum state information and its quantum characteristics are used to deeply understand the target user's current demand for learning resources from many aspects, determine the target user's preference for learning resources, and accurately recommend learning resources to the target user.

[0025] In one implementation, the above-mentioned step 104 of mapping the implicit learning feedback information into quantum state space encoding as quantum state information may include the following steps.

[0026] Step 1041: determining a basis state vector of the implicit learning feedback information in each learning dimension according to the implicit learning feedback information; Among them, the basis state vector is used to represent various learning dimensions, such as ideological, policy, historical and other learning dimensions. That is to say, implicit learning feedback information can reflect in which learning dimensions the target user has behavior, and understand the target user's behavior as a whole.

[0027] Step 1042: Determine the high-order mapping state corresponding to the implicit learning feedback information based on the basis state vector and the interaction weights of the implicit learning feedback information between the various learning dimensions; wherein the high-order mapping state is used to characterize the correlation relationship between the implicit learning feedback information in the various learning dimensions.

[0028] The target user's various learning dimensions include different dimensions reflected through implicit learning feedback. For example, if a target user frequently browses learning resources related to historical events, this reflects their learning progress in the historical dimension. Another example is if a target user repeatedly likes learning resources related to policies, this reflects their learning progress in the policy dimension. These learning dimensions also reflect the target user's learning intentions and help identify preferences reflected in implicit learning feedback.

[0029] In one implementation, the high-order configuration corresponding to the implicit learning feedback information is determined by the basis state vector and the interaction weights between the various learning dimensions, as shown below: ; in, Feedback for implicit learning The corresponding high-order mapping state is used to describe the implicit learning feedback information Feature associations across various learning dimensions.

[0030] For the The basis state vector corresponding to each learning dimension represents the specific learning dimension.

[0031] is the number of learning dimensions, each learning dimension Can represent different latent feature spaces, It represents the value range of the kth learning dimension. Different learning dimensions may have different numbers of values. For example, the thought dimension may have multiple different schools of thought.

[0032] It is an element of high-order mapping, which represents the interaction and association weight between implicit learning feedback information in various learning dimensions, such as the relationship between different historical events and current policies.

[0033] In one implementation, the above-mentioned step 104 constructs the target user's comprehensive preference state information for the various types of learning resources based on the quantum characteristics of the quantum state information, which may include the following steps.

[0034] Step 1043, based on the superposition characteristics of the quantum state information, according to the group state of each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information, determine the first comprehensive preference state information of the target user for the various types of learning resources; wherein, the learning resource group includes at least one type of learning resources; the group state corresponding to the learning resource group is used to characterize the target user's preference degree for the learning resource group.

[0035] Among them, by dividing various types of learning resources into different groups, each group includes at least one type of learning resources, one type of learning resources can be regarded as a source body, and each group is a multi-source group. The comprehensive state of the multi-source group is used to represent the target user's overall preference for various types of learning resources in the group, and the target user's preference for a single category of learning resources is expanded to a preference for learning resources of multiple categories, so as to capture the target user's complex or related preferences for learning resources.

[0036] The first comprehensive preference state information is used to represent the target user's overall preference for learning resources in multiple categories. Based on the classification of various learning resources into different groups, the first comprehensive preference state information of the target user for each type of learning resource is determined based on the group state of each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information as follows: ; in, Provide implicit learning feedback for target users The multi-source stacking configuration of the various types of learning resources represents first comprehensive preference state information, which is used to represent the overall preference of the target user for the various types of learning resources in the grouped learning resource group.

[0037] The group weight of each learning resource group.

[0038] Feedback for implicit learning The corresponding higher-order configurations.

[0039] is the target user’s quantum state information for the p-th type of learning resource in the k-th learning resource group, representing his or her interest level. It is the process of superimposing the quantum state information of the learning resources in the kth learning resource group by the target user.

[0040] Step 1044, based on the entanglement characteristics of the quantum state information, according to the state of the associated learning resources in each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information, determine the target user's second comprehensive preference state information for the various types of learning resources; wherein the associated learning resources include at least two types of interrelated learning resources.

[0041] Among them, step 1043 considers the target user's overall preference for each type of learning resource in each learning resource group based on the comprehensive state of the multi-source group. In step 1044, the target user's specific preference for the associated learning resources is further considered. In the embodiment of the present application, taking two types of interrelated learning resources as an example, the second comprehensive preference state information of the target user for the various types of learning resources is determined by the state of the associated learning resources in each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information as shown below: ; in, Provide implicit learning feedback for target users The second preference state information for the various types of learning resources is used to represent the target user's comprehensive preference for interrelated learning resources.

[0042] The group weight of each learning resource group.

[0043] Feedback for implicit learning The corresponding higher-order configurations.

[0044] Provide implicit learning feedback for target users The quantum state of learning resource A represents the degree of interest. Provide implicit learning feedback for target users The quantum state of learning resource B represents the degree of interest.

[0045] Step 1045 : linearly superimpose the first comprehensive preference information and the second comprehensive preference information to construct the comprehensive preference information of the target user for the various types of learning resources.

[0046] The first comprehensive preference state information and the second comprehensive preference state information are optimized and integrated through the super-superposition mechanism to generate global comprehensive preference state information, as shown below: ; in, It is a super-overlap state, that is, comprehensive preference state information, which is used to represent the target user's comprehensive preference for the various learning resources.

[0047] To obtain the first comprehensive preference status information and the second comprehensive preference status information.

[0048] represents a super-superposition operation, which linearly superimposes the first comprehensive preference state information and the second comprehensive preference state information.

[0049] In an embodiment of the present application, the target user's comprehensive preference state information for the various types of learning resources may include first comprehensive preference state information and second comprehensive preference state information. The first comprehensive preference state information is determined based on the superposition characteristics of the quantum state information, according to the group state corresponding to each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information. The second comprehensive preference state information is determined based on the entanglement characteristics of the quantum state information, according to the state of the associated learning resources in each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information. The first comprehensive preference state information is used to represent the target user's preference for a single category of learning resources, which is extended to a preference for learning resources of multiple categories, forming a global preference for learning resources. The second comprehensive preference state information is used to more finely represent the target user's preference for associated learning resources, that is, the second comprehensive preference state information can reflect the impact of the association of learning resources on the target user's preference, which is different from the first comprehensive preference state information. Therefore, based on the first comprehensive preference state information and the second comprehensive preference state information, the target user's comprehensive preference state information for various types of learning resources is determined, and the target user's complex or associated preferences for learning resources are more accurately captured for subsequent recommendation of learning resources.

[0050] In one implementation, because the target user's preferences may vary at different times, this embodiment of the present application also takes time into account, introducing the target user's dynamic preferences for various learning resources at a preset time. Thus, step 104 described above constructs the target user's comprehensive preference state information for the various learning resources based on the quantum properties of the quantum state information, and may include the following steps.

[0051] Step 1046: Based on the dynamic superposition characteristics of the quantum state information, the dynamic group states corresponding to the respective learning resource groups and the higher-order mapping states corresponding to the implicit learning feedback information, determine the target user's third comprehensive preference state information for the learning resources. The dynamic group states corresponding to the learning resource groups are used to represent the target user's preference for the learning resource groups at a preset time. The third comprehensive preference state information adds a preset time constraint to the first comprehensive preference state information, representing the target user's preference for learning resources in multiple categories at the preset time. The third comprehensive preference state information, determined by the dynamic group states corresponding to the learning resource groups and the high-order mapping state corresponding to the implicit learning feedback information, is as follows: ; in, Provide implicit learning feedback for target users The third comprehensive preference information for the various types of learning resources at a preset time is used to represent the target user's comprehensive preference for the various types of learning resources in the grouped learning resource group at a preset time.

[0052] The group weights of each learning resource group can be dynamically adjusted at a preset time. Feedback for implicit learning The corresponding higher-order configurations.

[0053] The dynamic quantum state information of the target user for the p-th type of learning resource in the k-th learning resource group at the preset time represents the degree of interest. It is the process of superimposing the quantum state information of the learning resources in the kth learning resource group by the target user at a preset time.

[0054] Step 1047 : Linearly superimpose the first comprehensive preference information, the second comprehensive preference information, and the third comprehensive preference information of the target user for the learning resources to construct the comprehensive preference information of the target user for the various types of learning resources.

[0055] Among them, through the super-superposition mechanism, the first comprehensive preference state information, the second comprehensive preference state information and the third comprehensive preference state information are optimized and integrated to generate global comprehensive preference state information, which is specifically shown as follows: ; in, It is a super-overlap state, that is, comprehensive preference state information, which is used to represent the target user's comprehensive preference for the various learning resources.

[0056] are the first comprehensive preference status information, the second comprehensive preference status information and the third comprehensive preference status information obtained above. represents a super-superposition operation, which linearly superimposes the first comprehensive preference state information, the second comprehensive preference state information, and the third comprehensive preference state information.

[0057] In an embodiment of the present application, based on the above-mentioned first comprehensive preference state information and second comprehensive preference state information, taking into account that the preferences of the target user may be different at different times, the dynamic preferences of the target user for various types of learning resources at a preset time are introduced, and the high-order mapping state corresponding to the implicit learning feedback information is used to jointly determine the third comprehensive preference state information of the target user for the various types of learning resources. The first comprehensive preference state information, the second comprehensive preference state information and the third comprehensive preference state information are then linearly superimposed to obtain comprehensive and accurate comprehensive preference state information of the target user for various types of learning resources, providing an accurate basis for recommending learning resources.

[0058] In one implementation, with respect to the above-obtained comprehensive preference information of the target user for the learning resource, the method may further include the following steps.

[0059] Step 112: Acquire the learning goal information of the target user and determine the matching degree between the various learning resources and the learning goal information.

[0060] Among them, the learning goal information of the target user provides a clear learning purpose. Based on the learning purpose, by matching it with various existing learning resources, the matching degree between various learning resources and the learning goal information can be obtained. This matching degree can measure the degree of matching between various learning resources and the learning goal information. Using it as the basis for recommending learning resources can provide more accurate learning resources for the target user.

[0061] Step 114 : determining the target user's matching preference value for the various types of learning resources based on the matching degree and the target user's comprehensive preference information for the various types of learning resources.

[0062] Among them, the matching preference value integrates the matching degree between the target user's learning goal and learning resources and the above-mentioned comprehensive preference state information, which can more comprehensively represent the target user's preference for various learning resources.

[0063] Step 116 : Decompose the target user's matching preference values ​​for the learning resources to determine the target user's target preference values ​​for the various types of learning resources.

[0064] Among them, through the deconfiguration mechanism, the matching preference values ​​of target users for various learning resources, which are integrated with multiple aspects and dimensions, are decomposed to obtain more specific target preference values ​​of target users for various learning resources, as shown below: ; in, It is a deconfiguration used to characterize the sub-quantum state information after decomposing the matching preference value, that is, the specific preference value of the target user for each type of learning resources.

[0065] Represents an unmarshal operation.

[0066] is the matching degree, which is used to represent the importance of various learning resources in the recommendation.

[0067] In an embodiment of the present application, after the comprehensive preference information is determined based on the target user's implicit learning feedback information and the target user's preference for learning resources, the target user's learning goal information is introduced to clarify the target user's current learning intention, and the matching preference value of the target user for the various learning resources is determined based on the matching degree between the learning goal information and the various learning resources and the comprehensive preference information of the target user for the various learning resources. The matching preference value integrates the multi-faceted and multi-dimensional preferences of the target user for the various learning resources. The matching preference value is decomposed to determine the target preference value of the target user for the various learning resources. The target preference value represents the target user's more specific preference for the various learning resources, providing a more accurate recommendation basis for the recommendation of learning resources, thereby improving the accuracy of the recommendation of learning resources.

[0068] In one implementation, after determining the target preference values ​​of the target user for the various types of learning resources in step 116, the method further includes the following steps.

[0069] Step 118: replacing the comprehensive preference information with the target user's target preference values ​​for the various types of learning resources; Step 120 : Recommending learning resources to the target user based on the target user's target preference values ​​for the various types of learning resources.

[0070] In an embodiment of the present application, based on obtaining the learning goal information of the target user and determining the target preference value obtained by decomposing the target user's preferences for various types of learning resources that integrate multiple aspects and dimensions, the target preference value can more accurately and more in line with the learning goal to characterize the target user's preferences for various types of learning resources, thereby using the target preference value to replace the above-mentioned comprehensive preference state information, and based on the target preference value of the target user for the various types of learning resources, the accuracy of learning resource recommendations to the target user can be further improved.

[0071] In one implementation, the above-mentioned learning resource recommendation method may further include the following steps.

[0072] Step 110: Acquire initial information of the target user; wherein the initial information includes identity information and learning tendency information.

[0073] Among them, the target user's identity information may include but is not limited to the following: user identity, such as students, employees, freelancers and unemployed persons; employee positions, such as grassroots employees, middle-level managers and senior managers; and age. Taking enterprises as an example, the target user's learning tendency information may include but is not limited to the following: professional learning content, culture and concepts, rules and regulations, development history, and feedback. The target user's identity information is used to characterize the target user's identity or position for learning. To a certain extent, the target user's learning needs are related to his or her identity. Therefore, the target user's identity information can be used as a basis for recommending learning resources. The target user's learning tendency information is used to characterize what the target user wants to learn, and further provides a basis for recommending learning resources based on his or her identity information.

[0074] Step 112: determining the target user's initial preference value for the initial information based on the initial information, the weight corresponding to the initial information, and the association relationship between the initial information.

[0075] In an embodiment of the present application, the initial preference value is used to represent the predicted preference value of the target user under the constraints of his or her identity information and learning tendency information. Through this initial preference value, the target user's preferences under the current identity and learning tendency can be known, thereby further supplementing and confirming the target user's preferences and improving the accuracy of learning resource recommendations.

[0076] In one implementation, the above-mentioned step 112 determines the target user's initial preference value for the initial information based on the initial information, the weight corresponding to the initial information, and the association relationship between the initial information, and may include the following steps.

[0077] Step 1121 : Determine an information matrix corresponding to the initial information based on a unit matrix corresponding to the initial information, a weight matrix corresponding to the weight, and a relationship matrix corresponding to the association relationship.

[0078] The unit matrix corresponding to the initial information is a unit matrix with the number of initial information as its dimension. For example, if the initial information includes five identity information and four learning tendency information, the unit matrix is ​​a 9×9 unit matrix. The elements of the weight matrix represent the importance of each piece of information.

[0079] In this embodiment of the present application, the association relationships between the initial information include: explicit relationships and implicit relationships between the identity information and the learning propensity information; and the relationship matrix corresponding to the association relationships includes: an explicit relationship matrix corresponding to the explicit relationships and an implicit relationship matrix corresponding to the implicit relationships. Explicit relationships are direct relationships between individual pieces of information, while implicit relationships are more complex relationships between individual pieces of information, such as where certain pieces of information may influence or interact with each other.

[0080] The information matrix corresponding to the initial information is determined by the identity matrix corresponding to the initial information, the weight matrix corresponding to the weight, and the relationship matrix corresponding to the association relationship as follows: ; in, is the information matrix corresponding to the initial information, is the weight matrix of the kth information layer, is the explicit relationship matrix of the kth information layer, is the implicit relationship matrix of the kth information layer. is the bias matrix, which is used to adjust the deviation between information levels.

[0081] Step 1122: construct a diagonal matrix corresponding to the initial information based on the identity information and the learning tendency information.

[0082] Among them, the diagonal matrix is ​​composed of identity information and learning tendency information. First, the eigenvector corresponding to the identity information is constructed And the feature vector corresponding to the learning tendency information , and then construct the diagonal matrix as follows: .

[0083] Step 1123: Determine the implicit eigenvalue of the initial information according to the information matrix, the diagonal matrix, and the identity matrix.

[0084] The implicit eigenvalues ​​of the initial information determined by the information matrix, the diagonal matrix, and the identity matrix are as follows: .

[0085] Among them, by adding the unit matrix into the calculation, the stability of the calculation of the implicit eigenvalue is ensured. That is, even after multiple interactive mappings, the original identity information and learning tendency information are still retained in the result, and will not be completely replaced by new interactive terms due to continuous mapping, thereby maintaining the structural integrity of the implicit features and preventing the overflow or disappearance of identity information and learning tendency information during the calculation of the implicit eigenvalue.

[0086] In step 1124 , the latent feature value is subjected to multi-level mapping through a nonlinear activation function to obtain the target user's initial preference value for the initial information.

[0087] The initial preference value of the target user for the initial information obtained by performing multi-level mapping of the latent feature value through a nonlinear activation function is as follows: ; in, is the initial preference value of the target user in the k+1th information layer, is the initial preference value of the target user in the kth information layer, and f is the nonlinear activation function; is the bias vector of the k+1th information layer.

[0088] In the embodiment of the present application, by fully understanding and mining the initial information of the target user, it is ensured that the generated initial preference value can fully represent the preferences shown by the target user based on the current initial information, providing a strong basis for the recommendation of learning resources.

[0089] In one implementation, the above-mentioned learning resource recommendation method may further include determining the recommendation value of each type of learning resource based on historical user ratings of the each type of learning resource, the popularity of the each type of learning resource, and the suppression weight corresponding to the popularity, which may specifically include the following steps.

[0090] Step 114 , stratify the various types of learning resources according to their importance or priority to obtain a plurality of learning resource layers; wherein the learning resource layers include at least one type of learning resource.

[0091] Among them, in the embodiment of the present application, the basis for stratifying various types of learning resources is the importance or priority of the learning resources. At this time, whether the learning resources can be recommended and the degree of recommendation are considered in terms of the situation of the various types of learning resources themselves and the historical learning situation of users on various types of learning resources.

[0092] Step 116 : Perform Bayesian smoothing on the scores of the various types of learning resources in each of the learning resource layers to obtain smoothed scores of the various types of learning resources in each of the learning resource layers.

[0093] In one embodiment, the scores of the various types of learning resources in each of the learning resource layers are subjected to Bayesian smoothing processing to obtain smoothed scores of the various types of learning resources in each of the learning resource layers, which may include step a, for each of the learning resource layers, determining the sum of the scores of the learning resources in each of the learning resource layers, the average score of the learning resources in each of the learning resource layers, and the number of learning resources in each of the learning resource layers based on the historical user scores of the learning resources; and step b, determining the smoothed score of each type of learning resource in each of the learning resource layers based on the sum of the scores of the learning resources in each of the learning resource layers, the average score of the learning resources in each of the learning resource layers, and the number of learning resources in each of the learning resource layers.

[0094] The smoothed scores of each type of learning resources in each learning resource layer are determined as follows: ; in, is the smoothed score of historical user u for learning resource i at the lth learning resource layer.

[0095] is the smoothing coefficient of the lth learning resource layer.

[0096] It is the average rating of all learning resources by the entire historical user group at the lth learning resource layer, usually expressed as the global average rating of a specific learning resource.

[0097] represents the original rating of historical user u on learning resource i at the lth learning resource layer.

[0098] M represents the total number of historical ratings given by user u to learning resource i at the lth learning resource layer.

[0099] Step 118 : Determine the target popularity of each type of learning resource according to the popularity of each type of learning resource and the suppression weight corresponding to the popularity.

[0100] Among them, the suppression weight corresponding to popularity is used to control the impact of popularity on learning resources. For example, a smaller suppression weight is assigned to the popularity of learning resources that reduce the impact of popularity. By adjusting the suppression weight, the proportion of popular content in the recommendation can be increased or decreased, thereby affecting the bias of the recommendation.

[0101] Step 120 : Determine the recommendation value of each type of learning resource according to the smoothed scores of each type of learning resource in each learning resource layer, the weights corresponding to each learning resource layer, and the target popularity of each type of learning resource.

[0102] The recommended values ​​of the various learning resources are determined based on the smoothed scores of the various learning resources in each learning resource layer, the weights corresponding to each learning resource layer, and the target popularity of the various learning resources as follows: ; in, is a dynamically adjusted weight function; is the suppression weight of popularity, which is used to adjust the influence of popular content in recommendation.

[0103] The weight corresponding to each learning resource layer is used to influence the bias of the recommendation. represents the popularity of learning resource i.

[0104] In an embodiment of the present application, the recommendation value of each type of learning resource is determined based on the smoothed score obtained by performing hierarchical Bayesian smoothing on the historical users' ratings of each type of learning resource, combined with the popularity of each type of learning resource and the suppression weight corresponding to the popularity. The evaluation indicators such as the rating and popularity of the learning resource are affected by multiple factors. By combining the historical users' ratings of each type of learning resource, the popularity of each type of learning resource, and the suppression weight used to suppress popularity for comprehensive evaluation, a fair recommendation value is given to each type of learning resource, thereby providing the target users with fair and diverse recommended learning resources, reducing the deviation caused by high ratings and popular content, and ensuring the fairness and diversity of recommended learning resources.

[0105] In one implementation, after the target user's initial preference value for the initial information, the target user's comprehensive preference information for the various types of learning resources, and the recommendation values ​​for the various types of learning resources are determined, the above-mentioned step 106 of recommending learning resources to the target user may include: recommending learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources, the target user's initial preference value for the initial information, and the recommendation values ​​for the various types of learning resources.

[0106] In an embodiment of the present application, by obtaining the initial information and implicit learning feedback information of the target user, the initial preference value of the target user for the initial information is determined based on the initial information, the weight corresponding to the initial information, and the correlation between the initial information. The implicit learning feedback information is then mapped to the quantum state space, and the quantum superposition characteristics and quantum entanglement characteristics are used to deeply explore the user's preferences for various types of learning resources, and determine the comprehensive preference state information of the target user for the various types of learning resources. Furthermore, the recommendation value of each type of learning resource is determined based on the historical user's ratings for the various types of learning resources, the popularity of the various types of learning resources, and the suppression weights corresponding to the popularity. The introduction of suppression weights can weaken the recommendation value of popular learning resources and ensure the fairness and diversity of recommended learning resources. After obtaining the above-mentioned initial preference value, comprehensive preference state information, and recommended value of learning resources, learning resources are recommended to the target user based on the initial preference value, the comprehensive preference state information, and the recommended value of the various types of learning resources. By deeply understanding the target user's current demand for learning resources from multiple aspects and combining the recommendation status of currently recommended learning resources, we can accurately determine the target user's preference information and the recommendation value of learning resources, and accurately recommend learning resources to the target user.

[0107] It should be noted that the learning resource recommendation method provided in the embodiments of the present application can be executed by a learning resource recommendation device, or a control module within the learning resource recommendation device that is used to execute the learning resource recommendation method. The embodiments of the present application use the learning resource recommendation device execution method as an example to illustrate the learning resource recommendation device provided in the embodiments of the present application.

[0108] Figure 2 FIG. 1 shows a schematic diagram of a learning resource recommendation device provided in an embodiment of the present application. Figure 2 As shown, the device 200 includes: an acquisition module 21, a construction module 22, and a recommendation module 23.

[0109] Among them, the acquisition module 21 is used to obtain the implicit learning feedback information of the target user; wherein, the implicit learning feedback information includes the behavioral information of the target user towards various types of learning resources; the first construction module 22 is used to map the implicit learning feedback information to the quantum state space and encode it into quantum state information, and construct the comprehensive preference state information of the target user for the various types of learning resources based on the quantum characteristics of the quantum state information; wherein, the comprehensive preference state information is used to characterize the degree of preference of the target user for the various types of learning resources; the recommendation module 25 is used to recommend learning resources to the target user based on the comprehensive preference state information of the target user for the various types of learning resources.

[0110] In one implementation, the above-mentioned construction module 22 can be used to determine the basis state vector of the implicit learning feedback information in each learning dimension based on the implicit learning feedback information; determine the high-order mapping state corresponding to the implicit learning feedback information based on the basis state vector and the interaction weight between the implicit learning feedback information in the various learning dimensions; wherein the high-order mapping state is used to characterize the correlation relationship between the implicit learning feedback information in the various learning dimensions.

[0111] In one implementation, the above-mentioned construction module 22 can be used to determine the first comprehensive preference state information of the target user for the various types of learning resources based on the superposition characteristics of the quantum state information, according to the group states corresponding to the various learning resource groups and the high-order mapping states corresponding to the implicit learning feedback information; wherein, the learning resource group includes at least one type of learning resources; the group state corresponding to the learning resource group is used to characterize the degree of preference of the target user for the learning resource group; based on the entanglement characteristics of the quantum state information, according to the states of the associated learning resources in the various learning resource groups and the high-order mapping states corresponding to the implicit learning feedback information, determine the second comprehensive preference state information of the target user for the various types of learning resources; wherein, the associated learning resources include at least two types of mutually related learning resources; the first comprehensive preference state information and the second comprehensive preference state information are linearly superimposed to construct the comprehensive preference state information of the target user for the various types of learning resources.

[0112] In one implementation, the above-mentioned construction module 22 can be used to determine the target user's third comprehensive preference state information for the various types of learning resources based on the dynamic superposition characteristics of the quantum state information, according to the dynamic group state corresponding to each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information; wherein, the dynamic group state corresponding to the learning resource group is used to characterize the target user's preference degree for the learning resource group at a preset time; the first comprehensive preference state information, the second comprehensive preference state information and the third comprehensive preference state information are linearly superimposed to construct the target user's comprehensive preference state information for the various types of learning resources.

[0113] In one implementation, the device 200 may also include a determination module that can be used to obtain the target user's learning goal information and determine the degree of match between the various types of learning resources and the learning goal information; determine the target user's matching preference value for the various types of learning resources based on the matching degree and the target user's comprehensive preference information for the various types of learning resources; decompose the target user's matching preference value for the various types of learning resources to determine the target user's target preference value for the various types of learning resources. Thus, the recommendation module 23 may also be used to: replace the comprehensive preference information with the target user's target preference value for the various types of learning resources; and recommend learning resources to the target user based on the target user's target preference value for the various types of learning resources.

[0114] In one implementation, the above-mentioned acquisition module 21 can also be used to: obtain the initial information of the target user; wherein the initial information includes identity information and learning tendency information; the above-mentioned determination module can also be used to: determine the initial preference value of the target user for the initial information based on the initial information, the weight corresponding to the initial information, and the association relationship between the initial information.

[0115] In one implementation, the above-mentioned determination module can be used to determine the information matrix corresponding to the initial information based on the unit matrix corresponding to the initial information, the weight matrix corresponding to the weight, and the relationship matrix corresponding to the association relationship; construct a diagonal matrix corresponding to the initial information based on the identity information and the learning tendency information; determine the implicit eigenvalues ​​of the initial information based on the information matrix, the diagonal matrix, and the unit matrix; perform multi-level mapping on the implicit eigenvalues ​​through a nonlinear activation function to obtain the initial preference value of the target user for the initial information.

[0116] In one implementation, the association relationship between the initial information includes: the explicit relationship and the implicit relationship between the identity information and the learning tendency information; the relationship matrix corresponding to the association relationship includes: the explicit relationship matrix corresponding to the explicit relationship and the implicit relationship matrix corresponding to the implicit relationship.

[0117] In one implementation, the above-mentioned determination module can also be used to: stratify the various types of learning resources according to their importance or priority to obtain multiple learning resource layers; wherein the learning resource layers include at least one type of learning resources; perform Bayesian smoothing on the scores of the various types of learning resources in each of the learning resource layers to obtain smoothed scores of the various types of learning resources in each of the learning resource layers; determine the target popularity of the various types of learning resources based on the popularity of the various types of learning resources and the suppression weights corresponding to the popularity; determine the recommendation value of the various types of learning resources based on the smoothed scores of the various types of learning resources in each of the learning resource layers, the weights corresponding to the various learning resource layers, and the target popularity of the various types of learning resources.

[0118] In one implementation, the above-mentioned Bayesian smoothing processing is performed on the scores of the various types of learning resources in each of the learning resource layers to obtain the smoothed scores of the various types of learning resources in each of the learning resource layers, including: for each of the learning resource layers, according to the scores of the learning resources, determining the sum of the scores of the learning resources in each of the learning resource layers, the average score of the learning resources in each of the learning resource layers, and the number of learning resources in each of the learning resource layers; according to the sum of the scores of the learning resources in each of the learning resource layers, the average score of the learning resources in each of the learning resource layers, and the number of learning resources in each of the learning resource layers, determining the smoothed score of each type of learning resource in each of the learning resource layers.

[0119] In one implementation, the above-mentioned recommendation module 23 can be used to recommend learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources, the target user's initial preference value for the initial information, and the recommendation value of the various types of learning resources.

[0120] The learning resource recommendation device in the embodiments of the present application can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device. For example, the mobile electronic device can be a mobile phone, tablet computer, laptop computer, PDA, in-vehicle electronic device, wearable device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA). The non-mobile electronic device can be a server, network attached storage (NAS), personal computer (PC), television (TV), ATM, or self-service machine, etc., which is not specifically limited in the embodiments of the present application.

[0121] The learning resource recommendation device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0122] The learning resource recommendation device provided in the embodiment of the present application can implement each process implemented in the method embodiment described in the figure, and will not be described again here to avoid repetition.

[0123] Figure 3 A structural diagram of a learning resource recommendation system provided by an embodiment of the present application is shown. Figure 3 The system 300 includes: a multi-level interactive mapping network 31, a quantum state preference generator 32, a learning resource specialization optimizer 33, and a learning resource recommender 34.

[0124] The multi-level interaction mapping network 31 is used to extract relevant features from the target user's initial information, including identity information and learning propensity information. The extracted relevant features are then subjected to multi-level interaction mapping processing to capture the implicit interaction dimension between the target user and the learning resources.

[0125] The quantum state preference generator 32 represents the target user's implicit learning feedback information as a superposition state, an entangled state, a dynamic superposition state, etc. through quantum state theory, and extracts the target user's implicit comprehensive preference for learning resources through the collapse process of the quantum state.

[0126] The learning resource specialized optimizer 33 uses a hierarchical Bayesian smoothing model to determine the recommendation value of learning resources based on historical user ratings of learning resources and the popularity of learning resources, combined with priority / importance and popularity suppression mechanisms, to reduce the bias caused by high ratings and popular content, and ensure the fairness and diversity of recommended learning resources.

[0127] The learning resource recommender 34 recommends learning resources to the target user by combining the implicit interaction dimension between the target user and the learning resource, the target user's explicit comprehensive preference for the learning resource, and the recommendation value of the learning resource.

[0128] It should be noted that the learning resource recommendation system provided in the embodiment of the present application can be applied to a variety of application software, such as government application software, learning application software, office application software, etc.

[0129] In one implementation, Figure 4 A flow chart showing another method for recommending learning resources provided by an embodiment of the present application is shown. This method can be applied to Figure 3 The recommendation system for learning resources shown in Figure 4 , the method includes the following steps.

[0130] Step 401: Acquire the target user's identity information and ideological tendency information.

[0131] Taking enterprises as an example, ideological orientation information is reflected through five characteristics: professional learning content, culture and philosophy, rules and regulations, development history, and feedback. Identity information includes four characteristics: user identity (for example, student, employee, freelancer, and unemployed), employee position (for example, grassroots employee, middle manager, and senior manager), functional position (for example, administration, finance, R&D, and production), and age.

[0132] Step 402: Determine the identity matrix based on the target user's identity information and ideological tendency information.

[0133] Among them, the unit matrix is ​​a 9×9 matrix, and its dimension is consistent with the number of information.

[0134] Step 403: construct an identity-ideology nesting matrix based on the target user's identity information and ideological tendency information.

[0135] Among them, the identity thought nesting matrix Determined by: .

[0136] in, is the weight matrix of the kth feature layer, indicating the importance of each feature. is the relationship matrix of the kth feature layer, which is used to capture the direct relationship between features. is the nested relationship matrix, which is a new interaction term introduced in the k-th feature layer to capture more complex relationships between features. is the bias matrix, which is used to adjust the deviation between levels.

[0137] Step 404 : determining the implicit interaction dimension of the target user based on the target user's identity information, ideological tendency information, and the identity-ideology nesting matrix.

[0138] Among them, the implicit interaction dimension Determined by: .

[0139] in, is the feature vector corresponding to the identity information, is the feature vector corresponding to the ideological tendency information. The transpose and Construct a diagonal matrix.

[0140] Step 405 : Perform multi-level mapping on the implicit interaction dimension to generate an initial preference value.

[0141] Among them, the implicit interaction dimension is mapped at multiple levels in the following way: .

[0142] in, is the initial preference value of the kth feature layer. F is the nonlinear activation function, is the bias vector of the kth feature layer.

[0143] Step 406: Obtain implicit learning feedback information of the target user.

[0144] Implicit learning feedback includes: Length of time spent studying rules and regulations: The time users spend studying rules and regulations; Length of time spent studying professional learning content: The time users spend studying professional learning content (such as relevant literature and courses); Frequency of participation in corporate events: The frequency and duration of users' participation in corporate events (such as training sessions and seminars); and Interaction: User interaction with professional learning content, such as click-through rate, depth of browsing, bookmarking, and favorites.

[0145] It's important to note that the specific behaviors in the target user's implicit learning feedback can also reflect the learning dimensions along which the target user is engaged. For example, if the target user studies historical research works, popular history readings, and historical literary works, this can reflect their learning in the historical dimension. Another example is if the target user browses official policies and their interpretations, this can reflect their learning in the policy interpretation dimension. Understanding the target user's behavior from different learning dimensions can further improve the accuracy of predicting and determining the target user's preferences.

[0146] Step 407: construct the target user's implicit learning feedback information into a high-order mapping state.

[0147] Among them, the higher-order configuration As shown below: .

[0148] in, It is an element of high-order mapping, which represents the interaction and correlation weight of implicit learning feedback information between different learning dimensions, such as the relationship between different historical events and current policies. Represents the basis state vector of the nth learning dimension, representing a specific learning dimension, such as ideological education, policy interpretation, etc. n represents the number of learning dimensions, and each learning dimension It can represent different latent feature spaces. It represents the value range of the kth learning dimension. Different learning dimensions may have different numbers of values. For example, the thought dimension may have multiple different schools of thought as the base state.

[0149] Step 408: For various learning resources, construct the target user's first comprehensive preference for the learning resources through multi-source stacking configuration.

[0150] The first comprehensive preference can be determined as follows: .

[0151] in, Indicates the target user's implicit learning feedback information The multi-source stacking configuration is used to comprehensively reflect users' preferences for different learning resources. is the weight of the stack configuration, which is used to reflect the target user's preference for the kth stack configuration. Indicates the The first Source state. represents the combination process of multiple sources, is the number of sources, i.e. the number of learning resources considered in the multi-source stack.

[0152] Step 409: For mutually related learning resources, construct the target user's second comprehensive preference for the learning resources through the multi-source entangled state.

[0153] The second comprehensive preference can be determined as follows: .

[0154] in, Indicates the target user's implicit learning feedback information The multi-source entangled state is used to comprehensively reflect users' preferences for interrelated learning resources. For example, after learning "rules and regulations", the preference for "feedback" is enhanced, or the synergistic learning effect of "culture and philosophy" and "development history" is achieved. Provide implicit learning feedback for target users The quantum state of learning resource A represents the degree of interest. Provide implicit learning feedback for target users The quantum state of learning resource B represents the degree of interest.

[0155] Step 410 : constructing the target user's third comprehensive preference for learning resources through dynamic multi-source stacking configuration at different times.

[0156] The second comprehensive preference can be determined as follows: .

[0157] in, Provide implicit learning feedback for target users The dynamic multi-source stacking configuration of the various types of learning resources at a preset time is used to represent the comprehensive preference of the target user for the various types of learning resources in the grouped learning resource group at the preset time. The weight of the multi-source stack configuration at a preset time can be dynamically adjusted. For the target user at the preset time The state of the individual source represents its level of interest.

[0158] In step 411, the first comprehensive preference, the second comprehensive preference, and the third comprehensive preference are optimized and integrated through a hyper-connected mechanism, and then decomposed through a deconfiguration mechanism under the constraints of a specific learning goal to determine the target preference value of the target user for the learning resource under the specific learning goal.

[0159] Among them, the super-superposition mechanism is: .in, Represents a super-cascade operation, are the first comprehensive preference, second comprehensive preference and third comprehensive preference obtained above. The deconfiguration mechanism is: .in, Used to represent the importance of various learning resources in recommendation.

[0160] Step 412 : Layer the learning resources according to importance or priority, and perform multi-level Bayesian smoothing on each layer of learning resources based on historical user ratings of the learning resources to obtain smoothed ratings of the learning resources.

[0161] Among them, the multi-level Bayesian smoothing process is: .

[0162] in, is the smoothed score of historical user u for learning resource i at the lth learning resource layer. is the smoothing coefficient of the lth learning resource layer. It is the average rating of all learning resources by the entire historical user group at the lth learning resource layer, usually expressed as the global average rating of a specific learning resource. represents the original rating of historical user u on learning resource i at the lth learning resource layer. M represents the total number of historical ratings of historical user u on learning resource i at the lth learning resource layer.

[0163] Step 413 : Based on the popularity of the learning resources, the suppression weight corresponding to the popularity, and the smoothed score of the learning resources, popularity suppression and integration processing is performed on the learning resources to obtain a recommendation value of the learning resources.

[0164] The recommended value of learning resources can be determined by the following formula: .

[0165] in, is a dynamically adjusted weight function; is the suppression weight of popularity, which is used to adjust the influence of popular content in recommendation. The weight corresponding to each learning resource layer is used to influence the bias of the recommendation. represents the popularity of learning resource i.

[0166] Step 414 : Recommend learning resources to the target user based on the initial preference value, the target user's target preference value for the learning resource, and the recommendation value of the learning resource.

[0167] The obtained initial preference value, the target preference value of the target user for the learning resource, and the recommended value of the learning resource are all input into the learning resource recommender for processing, so as to recommend accurate learning resources that meet the learning goals of the target user.

[0168] In an embodiment of the present application, by extracting implicit interaction dimensions from the target user's identity information and ideological tendency information and performing multi-level mapping, an initial preference value that meets the needs of the target user is generated. By mapping implicit learning feedback information into a preset quantum state space, the implicit learning feedback information and learning resources are constructed into high-order mapping states, multi-source stacking configurations, multi-source entangled states, and dynamic multi-source stacking configurations, and the target preference value that represents the target user's true and implicit preference is extracted, solving the problem of incorrectly predicting the user's preference based on the user's implicit feedback information. The historical user's rating of the learning resource is processed by multi-level Bayesian smoothing, and combined with the priority and popularity suppression mechanism of the learning resource, the deviation caused by high ratings and popular content is reduced to ensure the fairness and diversity of the recommended learning resources. By comprehensively analyzing the initial preference value, target preference value, and the recommended value of the learning resource, the target user's preference and the recommendability of the learning resource are accurately judged, thereby accurately recommending learning resources to the target user.

[0169] Based on the same technical concept, an embodiment of the present application further provides an electronic device, which is used to execute the above-mentioned learning resource recommendation method. Figure 5 A schematic diagram of the structure of an electronic device for implementing various embodiments of the present application. Electronic devices may have relatively large differences due to different configurations or performances, and may include a processor (processor) 501, a communication interface (Communications Interface) 502, a memory (memory) 503 and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 communicate with each other via the communication bus 504. The processor 501 can call a computer program stored in the memory 503 and can be run on the processor 501 to execute the various steps of the above-mentioned learning resource recommendation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0170] It should be noted that the electronic devices in the embodiments of the present application include: servers, terminals, or other devices other than terminals, such as cars, robots, and handheld devices.

[0171] The above electronic device structure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, the input unit may include a graphics processing unit (GPU) and a microphone, and the display unit may be configured as a display panel in the form of a liquid crystal display, an organic light-emitting diode, etc. The user input unit includes at least one of a touch panel and other input devices. A touch panel is also called a touch screen. Other input devices may include, but are not limited to, a physical keyboard, function keys (such as volume control buttons, power buttons, etc.), a trackball, a mouse, and a joystick, which will not be detailed here.

[0172] The memory can be used to store software programs and various data. The memory may mainly include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store an operating system, applications or instructions required for at least one function (such as a sound playback function, an image playback function, etc.). In addition, the memory may include volatile memory or non-volatile memory, or the memory may include both volatile and non-volatile memory. The non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct rambus random access memory (DRRAM).

[0173] The processor may include one or more processing units; optionally, the processor may integrate an application processor and a modem processor, wherein the application processor primarily handles operations related to the operating system, user interface, and application programs, and the modem processor primarily processes wireless communication signals, such as a baseband processor. It is understood that the modem processor may not be integrated into the processor.

[0174] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the various processes of the above-mentioned learning resource recommendation method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0175] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer-readable storage medium, such as a computer read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk.

[0176] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned learning resource recommendation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0177] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0178] An embodiment of the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes a program or instruction. When the program or instruction is executed, the various processes of the above-mentioned learning resource recommendation method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0179] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be noted that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0180] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a more preferred embodiment. Based on this understanding, the technical solution of this application, or the part that contributes to the existing technology, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of this application.

[0181] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

Claims

1. A method for recommending learning resources, characterized in that: include: Obtaining implicit learning feedback information of a target user; wherein the implicit learning feedback information includes behavior information of the target user on various learning resources; Mapping the implicit learning feedback information into a quantum state space and encoding it into quantum state information, and constructing the target user's comprehensive preference state information for the various types of learning resources based on the quantum properties of the quantum state information; wherein the comprehensive preference state information is used to represent the target user's preference for the various types of learning resources; Recommend learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources.

2. The method according to claim 1, characterized in that Mapping the implicit learning feedback information to a quantum state space and encoding it into quantum state information includes: Determining a basis state vector of the implicit learning feedback information in each learning dimension according to the implicit learning feedback information; According to the basis state vector and the interaction weights of the implicit learning feedback information between the various learning dimensions, the high-order mapping state corresponding to the implicit learning feedback information is determined; wherein the high-order mapping state is used to characterize the association relationship between the implicit learning feedback information in the various learning dimensions.

3. The method according to claim 2, characterized in that The step of constructing the target user's comprehensive preference state information for the various types of learning resources based on the quantum characteristics of the quantum state information includes: Based on the superposition characteristics of the quantum state information, and according to the group states corresponding to the respective learning resource groups and the higher-order mapping states corresponding to the implicit learning feedback information, determining the first comprehensive preference state information of the target user for the various types of learning resources; wherein the learning resource group includes at least one type of learning resource; and the group state corresponding to the learning resource group is used to represent the target user's preference for the learning resource group; Based on the entanglement characteristics of the quantum state information, and according to the states of the associated learning resources in each learning resource group and the high-order mapping state corresponding to the implicit learning feedback information, determining the target user's second comprehensive preference state information for the various types of learning resources; wherein the associated learning resources include at least two types of mutually associated learning resources; The first comprehensive preference information and the second comprehensive preference information are linearly superimposed to construct the comprehensive preference information of the target user for the various types of learning resources.

4. The method according to claim 3, characterized in that The step of constructing the target user's comprehensive preference state information for the various types of learning resources based on the quantum characteristics of the quantum state information includes: Based on the dynamic superposition characteristics of the quantum state information, and according to the dynamic group states corresponding to the respective learning resource groups and the high-order mapping states corresponding to the implicit learning feedback information, the third comprehensive preference state information of the target user for the various types of learning resources is determined; wherein the dynamic group states corresponding to the learning resource groups are used to represent the target user's preference for the learning resource groups at a preset time; The first comprehensive preference information, the second comprehensive preference information, and the third comprehensive preference information are linearly superimposed to construct the comprehensive preference information of the target user for the various types of learning resources.

5. The method according to any one of claims 1 to 4, characterized in that After constructing the target user's comprehensive preference information for the various types of learning resources, the method further includes: Obtaining the learning goal information of the target user, and determining the matching degree between the various learning resources and the learning goal information; Determining the target user's matching preference value for the various types of learning resources based on the matching degree and the target user's comprehensive preference information for the various types of learning resources; The matching preference values ​​of the target user for the various types of learning resources are decomposed to determine the target preference values ​​of the target user for the various types of learning resources.

6. The method according to claim 5, characterized in that After determining the target user's target preference values ​​for the various types of learning resources, the method further includes: Replacing the comprehensive preference state information with the target user's target preference value for the various types of learning resources; Recommend learning resources to the target user based on the target user's target preference values ​​for the various types of learning resources.

7. The method according to claim 1, characterized in that The method further comprises: Acquire initial information of the target user; wherein the initial information includes identity information and learning tendency information; An initial preference value of the target user for the initial information is determined according to the initial information, the weight corresponding to the initial information, and the association relationship between the initial information.

8. The method according to claim 7, characterized in that The determining, based on the initial information, the weights corresponding to the initial information, and the associations between the initial information, the initial preference value of the target user for the initial information includes: Determining an information matrix corresponding to the initial information according to a unit matrix corresponding to the initial information, a weight matrix corresponding to the weight, and a relationship matrix corresponding to the association relationship; Constructing a diagonal matrix corresponding to the initial information according to the identity information and the learning tendency information; Determining the implicit eigenvalues ​​of the initial information according to the information matrix, the diagonal matrix, and the identity matrix; The latent feature value is subjected to multi-level mapping through a nonlinear activation function to obtain the initial preference value of the target user for the initial information.

9. The method according to claim 7, characterized in that The method further comprises: The various types of learning resources are layered according to their importance or priority to obtain a plurality of learning resource layers; wherein the learning resource layers include at least one type of learning resource; Performing Bayesian smoothing on the scores of the various types of learning resources in each of the learning resource layers to obtain smoothed scores of the various types of learning resources in each of the learning resource layers; Determining target popularity of each type of learning resource according to the popularity of each type of learning resource and the suppression weight corresponding to the popularity; The recommendation value of each type of learning resource is determined according to the smoothed scores of each type of learning resource in each of the learning resource layers, the weights corresponding to each of the learning resource layers, and the target popularity of each type of learning resource.

10. The method according to claim 9, characterized in that The Bayesian smoothing process is performed on the scores of the various types of learning resources in each of the learning resource layers to obtain smoothed scores of the various types of learning resources in each of the learning resource layers, including: For each learning resource layer, determining, based on the scores of the learning resources, the sum of the scores of the learning resources in each learning resource layer, the average score of the learning resources in each learning resource layer, and the number of learning resources in each learning resource layer; The smoothed score of each type of learning resource in each learning resource layer is determined according to the sum of the scores of the learning resources in each learning resource layer, the average score of the learning resources in each learning resource layer, and the number of learning resources in each learning resource layer.

11. The method according to claim 9, characterized in that The recommending learning resources to the target user includes: Recommend learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources, the target user's initial preference value for the initial information, and the recommendation value of the various types of learning resources.

12. A learning resource recommendation device, characterized in that: include: An acquisition module, configured to acquire implicit learning feedback information of a target user; wherein the implicit learning feedback information includes behavior information of the target user on various learning resources; A construction module is used to map the implicit learning feedback information into a quantum state space and encode it into quantum state information, and to construct the target user's comprehensive preference state information for the various types of learning resources based on the quantum properties of the quantum state information; wherein the comprehensive preference state information is used to represent the target user's preference for the various types of learning resources; The recommendation module is used to recommend learning resources to the target user based on the target user's comprehensive preference information for the various types of learning resources.

13. An electronic device, characterized in that: The method comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method for recommending learning resources as claimed in any one of claims 1 to 11.

14. A readable storage medium, characterized in that The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the method for recommending learning resources according to any one of claims 1 to 11 are implemented.

15. A computer program product, characterized in that The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program includes a program or instructions, and when the program or instructions are executed, the steps of the learning resource recommendation method according to any one of claims 1 to 11 are implemented.

Images