A teaching resource recommendation method, system, device and medium

Abstract

The present invention provides a teaching resource recommendation method, system, device, and medium, relating to the field of digital teaching technology. The method involves segmenting learning behavior data into multiple behavioral event segments; determining explicit trigger labels for training paths between each behavioral event segment based on the association graph structure of cross-segment behaviors and the association trigger relationship between each behavioral event segment; further determining implicit dependency labels for training paths between each behavioral event segment based on the implicit semantic association of training intentions between each behavioral event segment and the training logic between each behavioral event segment; and jointly modeling the teaching resource configuration of the training path using the explicit trigger labels and implicit dependency labels of the training paths between each behavioral event segment to obtain joint constraints for the teaching resource configuration, thereby performing personalized recommendation of teaching resources based on the joint constraints. The solution of the present invention can achieve personalized recommendation of teaching resources driven by dependency semantic preferences.

Description

Technical Field

[0001] The present invention relates to the field of digital teaching technology, and more specifically, to a teaching resource recommendation method, system, device and medium. Background Art

[0002] Against the backdrop of the continuous development of digital education, the way of acquiring teaching resources is shifting from traditional offline unified distribution to platform-based online dynamic push. With the popularization of teaching terminal equipment and learning management systems, students' behavioral data in the learning process are widely collected and analyzed, providing the basic conditions for realizing more personalized teaching resource recommendations. Individualized teaching has gradually become an important path to improve teaching efficiency and student learning outcomes. Especially in large-scale teaching environments, the push of resources that accurately match students' needs has become the key to improving the intelligence level of teaching systems.

[0003] Existing technologies generally rely on single behavioral features or static behavioral sequences for matching in teaching resource recommendations based on behavioral preferences, ignoring the complex temporal logical relationships and semantic dependencies between learning behaviors. This approach fails to deeply explore the cross-segment semantic chains between learning events and the potential evolution of training intentions, resulting in low correlation in resource matching results. It is difficult to effectively respond to students' individualized, multi-stage training needs, especially in continuous learning paths. The lack of joint modeling of explicit triggers and implicit dependency structures in learning behaviors limits the recommendation mechanism's ability to understand the contextual semantics of teaching content. Therefore, how to achieve personalized recommendations of teaching resources driven by semantic preferences has become a difficult problem facing the industry. Summary of the Invention

[0004] The present invention provides a teaching resource recommendation method, system, device and medium, which can realize personalized recommendation of teaching resources driven by semantic preferences.

[0005] In a first aspect, the present invention provides a method for recommending teaching resources, comprising the following steps:

[0006] Obtaining learning behavior data of target students in a historical period from teaching terminal devices;

[0007] Segmenting the learning behavior data into learning events based on the behavioral characteristics of the target students to obtain multiple teaching and training behavioral event segments;

[0008] Performing a global behavior path graph analysis on all behavioral event segments to obtain a cross-segment behavior association graph structure, and then determining explicit trigger labels for the training paths between each behavioral event segment based on the association graph structure and the association trigger relationship between each behavioral event segment;

[0009] The implicit relationship modeling of the training intentions between each behavioral event segment is performed to obtain the implicit semantic correlation of the training intentions between each behavioral event segment. Then, based on all the implicit semantic correlations and the training logic between each behavioral event segment, the implicit dependency labels of the training paths between each behavioral event segment are determined.

[0010] The teaching resource configuration of the training path is jointly modeled by the explicit trigger labels and implicit dependency labels of the training path between each behavioral event segment, and the joint constraints of the teaching resource configuration are obtained, and then personalized recommendations of teaching resources are performed based on the joint constraints.

[0011] Preferably, the learning behavior data is segmented into learning events based on the behavioral characteristics of the target students to obtain multiple teaching and training behavior event segments, specifically including:

[0012] extracting behavioral characteristics of the target student in time series from the learning behavior data;

[0013] Determine the amount of difference in fluctuations of behavioral characteristics within adjacent time windows;

[0014] When the fluctuation difference exceeds a preset behavioral characteristic fluctuation threshold, it is determined to be an event segmentation boundary;

[0015] The learning behavior data is divided into multiple continuous behavior event segments according to all event segmentation boundaries.

[0016] Preferably, performing a global behavior path diagram analysis on all behavior event segments to obtain a cross-segment behavior association graph structure specifically includes:

[0017] Each behavioral event segment is used as a node in the path diagram;

[0018] Determine the similarity of behavioral patterns between any two behavioral event segments;

[0019] Constructing initial connection edges between nodes based on the similarity of the behavior patterns;

[0020] The initial connection edges are topologically optimized based on the time series relationship between the behavior event segments to generate an association graph structure of cross-segment behaviors.

[0021] Preferably, determining the explicit trigger label of the training path between each behavior event segment according to the association graph structure and the association trigger relationship between each behavior event segment specifically includes:

[0022] Extracting the connection edge weight between each two behavioral event segments from the association graph structure as the association strength value between each two behavioral event segments;

[0023] Analyze the correlation and triggering relationship between each behavioral event fragment in the time dimension;

[0024] Determining explicit dependency values ​​of training paths between each behavioral event segment based on all association strength values ​​and the association trigger relationship;

[0025] All explicit dependency values ​​are converted into explicit trigger labels for the training paths between each behavioral event segment through preset mapping rules.

[0026] Preferably, implicit relationship modeling is performed on the training intentions between the various behavioral event segments to obtain the implicit semantic correlation of the training intentions between the various behavioral event segments, specifically including:

[0027] Label each behavioral event segment with a learning objective, and then determine the degree of fit between the learning objectives of each behavioral event segment based on the labeling results;

[0028] Construct the intention association matrix of the behavioral event segments based on all the fit scores;

[0029] Based on the intention association matrix, the implicit association semantics of the training intentions between the various behavior event segments are analyzed, and then the implicit semantic association degree of the training intentions between the various behavior event segments is obtained.

[0030] Preferably, determining the implicit dependency labels of the training paths between the behavioral event segments based on all implicit semantic associations and the training logic between the behavioral event segments specifically includes:

[0031] All implicit semantic correlations are associated and fused with the training logic between each behavioral event segment to obtain the fusion result of the implicit relationship between each behavioral event segment;

[0032] Learning the dependency pattern of the fusion results through a pre-trained neural network model;

[0033] An implicit dependency matrix representing the potential logical relationship is output according to the learned dependency pattern, and then the implicit dependency labels of the training paths between the various behavioral event segments are determined through the implicit dependency matrix.

[0034] Preferably, the teaching terminal device refers to an intelligent hardware device used to collect, display, interact and record students' learning behavior information during the teaching process.

[0035] In a second aspect, the present invention provides a teaching resource recommendation system, comprising:

[0036] An acquisition module is used to obtain the learning behavior data of target students in a historical period from the teaching terminal device;

[0037] a processing module for performing learning event segmentation on the learning behavior data based on the behavioral characteristics of the target students to obtain a plurality of teaching and training behavioral event segments;

[0038] The processing module is further configured to perform a global behavior path graph analysis on all behavior event segments to obtain a cross-segment behavior association graph structure, and then determine explicit trigger labels for the training paths between each behavior event segment based on the association graph structure and the association trigger relationship between each behavior event segment;

[0039] The processing module is further used to perform implicit relationship modeling on the training intentions between the various behavioral event segments, obtain the implicit semantic association of the training intentions between the various behavioral event segments, and then determine the implicit dependency labels of the training paths between the various behavioral event segments based on all the implicit semantic associations and the training logic between the various behavioral event segments;

[0040] The execution module is used to jointly model the teaching resource configuration of the training path through the explicit trigger labels and implicit dependency labels of the training path between each behavioral event segment, obtain the joint constraints of the teaching resource configuration, and then perform personalized recommendations of teaching resources based on the joint constraints.

[0041] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a code, and the processor is configured to obtain the code and execute the above-mentioned teaching resource recommendation method.

[0042] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which implements the above-mentioned teaching resource recommendation method when executed by a processor.

[0043] The technical solutions provided by the embodiments disclosed in the present invention have the following beneficial effects:

[0044] In an embodiment of the present invention, learning behavior data of a target student in a historical time period is obtained from a teaching terminal device; learning event segmentation is performed on the learning behavior data based on the behavioral characteristics of the target student to obtain multiple behavioral event segments of teaching training; a global behavior path graph analysis is performed on all behavioral event segments to obtain an association graph structure of cross-segment behavior, and then the explicit trigger label of the training path between each behavioral event segment is determined based on the association graph structure and the association trigger relationship between each behavioral event segment; implicit relationship modeling is performed on the training intention between each behavioral event segment to obtain the implicit semantic correlation of the training intention between each behavioral event segment, and then the implicit dependency label of the training path between each behavioral event segment is determined based on all the implicit semantic correlations and the training logic between each behavioral event segment; the teaching resource configuration of the training path is jointly modeled through the explicit trigger label and implicit dependency label of the training path between each behavioral event segment to obtain the joint constraint conditions of the teaching resource configuration, and then personalized recommendation of teaching resources is performed based on the joint constraint conditions.

[0045] It can be seen that the present invention jointly models the teaching resource configuration of the training path through the explicit trigger labels and implicit dependency labels of the training path between the behavioral event fragments, and then performs personalized recommendation of teaching resources based on the joint constraints obtained by modeling; first, the explicit trigger labels of the training path between each behavioral event fragment are determined according to the association graph structure and the association trigger relationship between each behavioral event fragment, and the cross-fragment behavioral association graph structure is constructed through global behavioral path graph analysis to identify the explicit trigger labels between each event fragment, thereby revealing the clear sequential logic and excitation relationship between different learning behaviors, and effectively reflecting the behavioral advancement trajectory followed by students in the actual training process; secondly, the implicit dependency labels of the training path between each behavioral event fragment are determined according to all the implicit semantic associations and the training logic between each behavioral event fragment, and the implicit semantic modeling method is combined to conduct in-depth analysis of the training intention to extract the implicit dependency labels between each event fragment. , so that the deep intention connection that was originally difficult to identify from the surface behavior is clearly portrayed, and the potential semantic path is restored; then, the teaching resource configuration of the training path is jointly modeled through the explicit trigger labels and implicit dependency labels of the training path between the behavioral event fragments, and then personalized recommendation of teaching resources is performed based on the joint constraints obtained by modeling. On the basis of explicit trigger labels and implicit dependency labels, the joint constraints of resources in the teaching path are jointly modeled, so that personalized recommendation can not only fit the surface preferences of learning behavior, but also fully consider the deep logical needs in its semantic evolution process. Finally, through the combined explicit and implicit modeling mechanism, the teaching resource recommendation is significantly enhanced in terms of the appropriateness of content matching, the dynamics of path planning and the context adaptability of recommendation results, thereby realizing personalized teaching resource push driven by semantic preferences; in summary, the scheme of the present invention can realize personalized recommendation of teaching resources driven by semantic preferences. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is an exemplary flow chart of a teaching resource recommendation method according to some embodiments of the present invention;

[0047] Figure 2 is a schematic diagram of an application scenario of a teaching resource recommendation system according to some embodiments of the present invention;

[0048] Figure 3 is a schematic diagram of a process for determining an association graph structure according to some embodiments of the present invention;

[0049] Figure 4 is a structural diagram of a teaching resource recommendation system according to some embodiments of the present invention;

[0050] Figure 51 is a structural diagram of a computer device for implementing a teaching resource recommendation method according to some embodiments of the present invention. DETAILED DESCRIPTION

[0051] In order to better understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0052] refer to Figure 1 , which is an exemplary flow chart of a teaching resource recommendation method according to some embodiments of the present invention. The teaching resource recommendation method mainly includes the following steps:

[0053] In step 101, the learning behavior data of the target student in a historical time period is obtained from the teaching terminal device.

[0054] It should be noted that the teaching terminal device in the present invention refers to an intelligent hardware device used to collect, display, interact and record students' learning behavior information during the teaching process, and its role is to serve as an interactive medium between students and the teaching system; it should also be noted that the present invention assumes that the target students have used the teaching terminal device to participate in teaching and training activities within a historical time period; it should also be noted that the learning behavior data in the present invention refers to the operational data recorded by the teaching terminal device during the teaching and training process that reflects the students' learning status and behavioral characteristics.

[0055] In some embodiments, reference Figure 2 As shown in the figure, this figure is a schematic diagram of the application scenario of the teaching resource recommendation system shown in some embodiments of the present invention, which includes three main components: a terminal device, a server and a data storage device. The terminal device is used to record the students' learning behavior data, send the learning behavior data to the server through the communication network, run the execution code of the teaching resource recommendation system in the server, and finally store the processed execution recommendation data in the data storage device through the server.

[0056] In step 102, the learning behavior data is segmented into learning events based on the behavioral characteristics of the target students to obtain a plurality of teaching and training behavior event segments.

[0057] In some embodiments, the learning behavior data is segmented into learning events based on the behavior characteristics of the target student to obtain multiple teaching and training behavior event segments, which can be achieved by the following steps:

[0058] extracting behavioral characteristics of the target student in time series from the learning behavior data;

[0059] Determine the amount of difference in fluctuations of behavioral characteristics within adjacent time windows;

[0060] When the fluctuation difference exceeds a preset behavioral characteristic fluctuation threshold, it is determined to be an event segmentation boundary;

[0061] The learning behavior data is divided into multiple continuous behavior event segments according to all event segmentation boundaries.

[0062] It should be noted that the behavioral characteristics in the present invention are characteristics that describe students' learning operation methods, interaction frequency and behavioral change status; the fluctuation difference in the present invention is an indicator to measure the degree of change of students' behavioral characteristics in adjacent time windows; the event segmentation boundary in the present invention refers to the position where the behavioral pattern changes significantly in the learning behavior sequence; the behavioral event fragment in the present invention is used to represent a phased learning unit that is continuous and has consistent behavioral characteristics in the learning process.

[0063] In specific implementation, first, extracting the target student's behavioral characteristics in the time series from the learning behavior data can be achieved in the following manner, namely: extracting the behavioral characteristics of each time point from the learning behavior data, wherein the behavioral characteristics specifically include click frequency, dwell time, operation type code and sliding speed; secondly, determining the fluctuation difference of the behavioral characteristics in adjacent time windows can be achieved in the following manner, namely: presetting a time window of fixed length, and calculating the statistics of all behavioral characteristics in two adjacent time windows in a sliding manner of the time window, and then using Euclidean distance or Manhattan distance to calculate the difference between the statistics of the two adjacent time windows, and using the difference as the fluctuation difference; then, when the When the fluctuation difference exceeds the preset behavioral feature fluctuation threshold, it is determined to be an event segmentation boundary in the following manner, namely: comparing the fluctuation difference with the preset behavioral feature fluctuation threshold. When the fluctuation difference exceeds the behavioral feature fluctuation threshold, it is considered that there is a behavioral pattern mutation at that location, and it is determined to be an event segmentation boundary. It should be noted that the behavioral feature fluctuation threshold in the present invention can be determined based on historical behavioral feature data, which will not be elaborated here. Finally, according to all event segmentation boundaries, the learning behavior data is divided into multiple continuous behavioral event segments, which can be achieved in the following manner, namely: using the mutation point as the event boundary, and dividing the learning behavior data corresponding to the entire time series into multiple continuous behavioral event segments.

[0064] In step 103, a global behavior path graph analysis is performed on all behavior event segments to obtain an association graph structure of cross-segment behaviors, and then the explicit trigger labels of the training paths between each behavior event segment are determined based on the association graph structure and the association trigger relationship between each behavior event segment.

[0065] In some embodiments, reference Figure 3As shown in FIG. 1 , this figure is a schematic diagram of a process for determining an association graph structure in some embodiments of the present invention. In this embodiment, a global behavior path graph analysis is performed on all behavior event segments to obtain an association graph structure for cross-segment behaviors. The following steps can be used to achieve this:

[0066] In step 1031, each behavioral event segment is used as a node of a path graph;

[0067] In step 1032, the behavioral pattern similarity between any two behavioral event segments is determined;

[0068] In step 1033, initial connection edges between nodes are constructed based on the behavior pattern similarity;

[0069] In step 1034, the initial connection edges are topologically optimized based on the time series relationship between the behavior event segments to generate an association graph structure of the cross-segment behavior.

[0070] It should be noted that the behavioral pattern similarity in the present invention is an indicator to measure the degree of similarity of behavioral characteristics between different learning behavioral event fragments; the initial connecting edges in the present invention are used to represent the potential association relationship between behavioral event fragments based on behavioral similarity; the association graph structure in the present invention is a directed weighted graph constructed with behavioral event fragments as nodes and connecting edges reflecting their behavioral similarity and timing constraints as topological units, which is used to characterize the global association relationship between each behavioral event fragment in the training path.

[0071] In the specific implementation, first, each behavioral event segment can be used as a node of the path graph in the following way, that is, the behavioral event segment obtained by the above division can be used as a single node in the path graph to obtain a node set; then, the behavioral pattern similarity between any two behavioral event segments can be determined in the following way, that is, for any two nodes, the multi-dimensional behavioral feature sequence of the corresponding event segment is extracted, and the similarity of the two time series in the behavioral pattern is calculated using the dynamic time warping algorithm, and the similarity is used as the behavioral pattern similarity between the two behavioral event segments; then, a node is constructed based on the behavioral pattern similarity. The initial connection edges between nodes can be realized in the following way, namely: the normalized value of the behavior pattern similarity can be used as the initial connection edges between corresponding nodes; finally, the initial connection edges are topologically optimized based on the time series relationship between the behavior event fragments, and the association graph structure of cross-fragment behavior is generated in the following way, namely: according to the timestamp information of the behavior event fragments, time series constraint rules are constructed, and edges that are logically impossible to have direct associations are eliminated, such as edges from future nodes to past nodes, and edge weights are adjusted to reflect time proximity, so as to complete the topological structure optimization of the path graph, and the path graph obtained by the optimization process is used as the association graph structure of cross-fragment behavior.

[0072] In some embodiments, determining the explicit trigger labels of the training paths between the behavioral event segments based on the association graph structure and the association trigger relationship between the behavioral event segments can be achieved by using the following steps:

[0073] Extracting the connection edge weight between each two behavioral event segments from the association graph structure as the association strength value between each two behavioral event segments;

[0074] Analyze the correlation and triggering relationship between each behavioral event fragment in the time dimension;

[0075] Determining explicit dependency values ​​of training paths between each behavioral event segment based on all association strength values ​​and the association trigger relationship;

[0076] All explicit dependency values ​​are converted into explicit trigger labels for the training paths between each behavioral event segment through preset mapping rules.

[0077] It should be noted that the association trigger relationship in the present invention refers to the temporal logical association between two behavioral event fragments that may constitute a causal relationship in time sequence; the explicit dependency value in the present invention is an indicator to measure the direct dependency strength of two behavioral event fragments in the training path; the explicit trigger label in the present invention is an identifier used to represent the clear training dependency relationship between behavioral event fragments.

[0078] In specific implementation, first, the connection edge weight between each two behavioral event fragments is extracted from the association graph structure as the association strength value between each two behavioral event fragments, which can be implemented in the following way, namely: extracting the edge weight between each pair of connected nodes from the association graph structure as their association strength value; secondly, analyzing the association trigger relationship between each behavioral event fragment in the time dimension can be implemented in the following way, namely: combining the start and end time information of each event fragment, analyzing its sequential relationship on the time axis, and determining whether there is a successive trigger relationship, a judgment strategy based on a time difference threshold can be adopted, such as the occurrence of successive behaviors within a reasonable time range is regarded as having an association trigger relationship; then, determining the explicit dependency value of the training path between each behavioral event fragment based on all the association strength values ​​and the association trigger relationship can be implemented in the following way, namely: for two behavioral event fragments, obtaining the association strength value between the two behavioral event fragments and the association trigger relationship between the two behavioral event fragments. The associated trigger relationship in the time dimension, and then jointly calculate the association strength value and the trigger relationship coefficient that characterizes the associated trigger relationship. A common method includes setting a weight coefficient to linearly superimpose the two. The joint calculation result can be used as the explicit dependency value of the training path between the two behavioral event fragments to quantify the degree of direct correlation in the training path. Through the above method, the explicit dependency value of the training path between any two behavioral event fragments can be obtained; finally, all explicit dependency values ​​are converted into explicit trigger labels of the training path between each behavioral event fragment through the preset mapping rules. It can be achieved in the following way, namely: all explicit dependency values ​​are converted into discrete explicit trigger labels according to the preset mapping rules. It should be further explained that the preset mapping rules in the present invention can be implemented in the following way, namely: the explicit dependency values ​​can be mapped into weak trigger, medium trigger and strong trigger labels according to the numerical interval. The specific numerical interval mapping relationship is not specifically limited here.

[0079] It should be noted that the solution of the present invention solves the problems of inaccurate training path dependency identification and logical ambiguity caused by traditional reliance on a single feature by combining the behavioral pattern similarity and time series triggering relationship between behavioral event fragments. The multi-dimensional fusion method accurately quantifies the explicit dependency strength between fragments, improves the temporal consistency and correlation rationality of the training path, and the structured expression of explicit trigger labels achieved thereby enhances the refined scheduling capability of teaching resource allocation, effectively supports the dynamic optimization of personalized teaching training paths, and improves the system's adaptability to changes in student behavior and the targeted teaching effect.

[0080] In step 104, implicit relationship modeling is performed on the training intentions between each behavioral event segment to obtain the implicit semantic correlation of the training intentions between each behavioral event segment, and then the implicit dependency labels of the training paths between each behavioral event segment are determined based on all the implicit semantic correlations and the training logic between each behavioral event segment.

[0081] In some embodiments, implicit relationship modeling of the training intentions between the behavioral event segments and obtaining the implicit semantic correlation of the training intentions between the behavioral event segments can be achieved by the following steps:

[0082] Label each behavioral event segment with a learning objective, and then determine the degree of fit between the learning objectives of each behavioral event segment based on the labeling results;

[0083] Construct the intention association matrix of the behavioral event segments based on all the fit scores;

[0084] Based on the intention association matrix, the implicit association semantics of the training intentions between the various behavior event segments are analyzed, and then the implicit semantic association degree of the training intentions between the various behavior event segments is obtained.

[0085] It should be noted that the fit score in the present invention is an indicator for measuring the degree of matching between the learning objectives of two behavioral event segments; the intention association matrix in the present invention is a matrix for quantifying the correlation between the learning objectives of the behavioral event segments; and the latent semantic association in the present invention is an indicator for measuring the correlation between the deep semantic structures of the behavioral event segments.

[0086] In the specific implementation, first, each behavioral event segment is labeled with a learning target, and then the fit score of the learning target between each behavioral event segment is determined based on the labeling results. This can be achieved in the following way, namely: each behavioral event segment can be assigned a corresponding learning target label based on expert pre-labeling, and then the Euclidean distance of the learning target labels between each two behavioral event segments is calculated, and the value obtained by normalizing the Euclidean distance is used as the fit score of the learning target between each two behavioral event segments; secondly, the intention association matrix of the behavioral event segment is constructed based on all the fit scores. This can be achieved in the following way, namely: all the fit scores are combined into a matrix, and the matrix is ​​used as the intention Figure association matrix; then, based on the intention association matrix, the implicit association semantics of the training intentions between each behavior event segment is analyzed, and then the implicit semantic association degree of the training intentions between each behavior event segment is obtained. This can be achieved in the following way, namely: the intention association matrix can be used as input, and the matrix elements represent the fit of the learning objectives between the behavior event segments. Based on the intention association matrix, the spectral clustering algorithm model is used to cluster the behavior event segments, identify potential semantic groups, clarify the implicit intention association structure between the behavior event segments, and generate the implicit semantic association degree between each segment through similarity calculation to quantify its deep connection at the training intention level, so as to support the refined construction and resource allocation of subsequent training paths.

[0087] In some embodiments, determining the implicit dependency labels of the training paths between the behavioral event segments based on all implicit semantic associations and the training logic between the behavioral event segments can be achieved by using the following steps:

[0088] All implicit semantic correlations are associated and fused with the training logic between each behavioral event segment to obtain the fusion result of the implicit relationship between each behavioral event segment;

[0089] Learning the dependency pattern of the fusion results through a pre-trained neural network model;

[0090] An implicit dependency matrix representing the potential logical relationship is output according to the learned dependency pattern, and then the implicit dependency labels of the training paths between the various behavioral event segments are determined through the implicit dependency matrix.

[0091] It should be noted that the dependency pattern in the present invention is a pattern indicator that characterizes the structured association rules within the training path between behavioral event segments; the implicit dependency label in the present invention is an identifier used to express the potential dependency relationship between behavioral event segments.

[0092] In the specific implementation, first, all the implicit semantic associations and the training logic between each behavior event segment are associated and fused to obtain the fusion result of the implicit relationship between each behavior event segment. This can be achieved in the following way, namely: for every two behavior event segments, the implicit semantic association of the training intention between the two behavior event segments and the training logic value between the two behavior event segments can be spliced ​​to form a fusion vector, and the fusion vector is used as the fusion result of the implicit relationship between the behavior event segments. The fusion result of the implicit relationship between every two behavior event segments can be obtained in the above way. It should be further explained that the training logic value in the present invention is an indicator for measuring the predefined knowledge sequence relationship and teaching process dependency between behavior event segments; secondly, learning the dependency pattern of the fusion result through a pre-trained neural network model can be achieved in the following way, namely: all the fusion vectors are used as input features of the pre-trained neural network model, and the pre-trained neural network model is input. The model usually adopts a graph neural network structure, which gradually aggregates information of neighboring nodes through multi-layer graph convolution or attention mechanism to mine the complex dependency relationships and potential patterns between behavioral event fragments; then, based on the learned dependency pattern, an implicit dependency matrix representing the potential logical relationship is output, and then the implicit dependency label of the training path between each behavioral event fragment is determined by the implicit dependency matrix. It can be implemented in the following way, namely: in the training stage, the neural network model adjusts the network weights through supervised learning based on the labeled training path samples, so that it can accurately capture the implicit dependency structure in the fusion feature; in the inference stage, the neural network model can output a matrix reflecting the dependency strength between behavioral event fragments, and use the matrix as the implicit dependency matrix in the present invention; the elements in the implicit dependency matrix can measure the potential training path dependency strength between behavioral event fragments; the implicit dependency matrix is ​​further normalized, and the matrix obtained by normalization is used as the implicit dependency label of the training path between each behavioral event fragment.

[0093] It should be noted that the solution of the present invention can effectively solve the problems of lack of deep semantic expression and insufficient logical constraints in the existing technology of training path dependency identification by integrating implicit semantic correlation and predefined training logic. It uses deep models to mine the potential intention relationship between behavioral fragments, realizes the precise quantification and structured expression of implicit dependencies, and improves the semantic integrity and logical rigor of the training path. The technology of the present invention significantly optimizes the path planning of personalized resource allocation, improves the teaching system's ability to respond to the dynamic changes of complex learning behaviors and the targeted training effect.

[0094] In step 105, the teaching resource configuration of the training path is jointly modeled through the explicit trigger labels and implicit dependency labels of the training path between each behavioral event segment to obtain the joint constraints of the teaching resource configuration, and then personalized recommendation of teaching resources is performed based on the joint constraints.

[0095] In some embodiments, the teaching resource configuration of the training path is jointly modeled by the explicit trigger labels and implicit dependency labels of the training path between each behavioral event segment, and the joint constraint conditions of the teaching resource configuration are obtained by the following steps:

[0096] Mapping the explicit trigger tag to a timing constraint graph of teaching resource configuration;

[0097] Mapping the implicit dependency labels into a logical dependency graph of teaching resource configuration;

[0098] The timing constraint graph and the logic dependency graph are optimized at a minimum cost, and then a joint constraint condition including resource allocation priority and conditional rules is generated according to the optimization result.

[0099] It should be noted that the timing constraint graph in the present invention is a graph structure used to characterize the temporal sequence of resource allocation between behavioral event fragments; the logical dependency graph in the present invention is a graph structure used to express the logical dependency relationship of resource allocation between behavioral event fragments; the joint constraint conditions in the present invention are a set of constraint conditions used to standardize resource allocation priorities and dependencies.

[0100] In the specific implementation, first, mapping the explicit trigger label to the timing constraint graph of the teaching resource configuration can be achieved in the following way, namely: for each pair of behavioral event fragments with explicit trigger labels, parse the label content to determine the trigger strength and directionality, and use the corresponding behavioral event fragments as nodes in the graph, establish directed edges according to the label instructions, and the edge weights can be given specific weights according to the trigger strength. All explicit trigger labels are traversed, and a complete directed weighted graph is gradually constructed, and the obtained directed weighted graph is used as the timing constraint graph of the teaching resource configuration; then, the implicit dependency label is mapped to the logical dependency graph of the teaching resource configuration. The spectrum can be implemented in the following way, namely: each behavior event fragment can be regarded as a node in the logical dependency graph, the implicit dependency label can be parsed, the implicit dependency relationship and its strength between the behavior event fragments can be identified, and directed edges can be established between the nodes accordingly. The edge weight reflects the strictness of the implicit dependency, and all implicit dependency labels can be traversed to gradually build a directed weighted graph containing complete logical constraints and dependency strength. The directed weighted graph can be used as the logical dependency graph of teaching resource configuration to guide the conditional allocation and priority control of resources, so as to achieve accurate expression and management of implicit dependency relationships in the training path; finally, the timing constraint graph and the logical The logical dependency graph is used to perform minimum cost optimization, and then a joint constraint condition containing resource allocation priority and conditional rules is generated based on the optimization results. This can be achieved in the following way: the timing constraint graph and the logical dependency graph can be fused to obtain a unified resource allocation constraint set, wherein the timing constraint graph provides direct sequential restrictions on resource allocation between behavioral event segments, and the logical dependency graph provides implicit conditional dependencies and sharing restrictions on resource allocation between behavioral event segments. Furthermore, an integer linear programming model is initialized, and the resource allocation constraint set is used as the initialization parameter of the integer linear programming model, wherein the integer linear programming model The decision variables include the resource allocation time point and allocation status of each behavioral event segment. The objective function design can comprehensively consider the total cost of resource use, task priority weight and dependency conflict penalty term to balance resource efficiency and task execution priority. In the integer linear programming model, the branch and bound algorithm can be used for systematic search to improve the solution efficiency and avoid the local optimal trap. After the solution is completed, the integer linear programming model can output a specific resource allocation plan, including the resource allocation order, start time and priority scheduling rules of the behavioral event segment, and all rule sets in the resource allocation plan are used as the joint constraint conditions.

[0101] It should be noted that the scheme of the present invention solves the scheduling conflicts and unclear priorities caused by the separation of timing and logical constraints in existing teaching resource configuration schemes through the joint modeling of explicit trigger tags and implicit dependency tags. It adopts the minimum cost optimization of the fusion of timing constraint graph and logical dependency graph to realize resource allocation optimization under multi-dimensional constraints, improve resource utilization efficiency and scheduling accuracy. The method of the present invention can significantly enhance the constraint strictness and execution stability of training path teaching resource configuration, thereby improving the personalized adaptability and teaching effect of the teaching resource recommendation system in complex dynamic environments.

[0102] In some embodiments, performing personalized recommendation of teaching resources based on the joint constraint conditions refers to personalized push of teaching resources that meet the training path requirements based on resource allocation rules; in specific implementation, the joint constraint conditions can be used as constraints of the recommendation strategy in the teaching resource recommendation system, and personalized teaching resource recommendations can be made to the target students under the execution of the constraint conditions.

[0103] On the other hand, in some embodiments, the present invention provides a teaching resource recommendation system, referring to Figure 4 , which is a schematic diagram of the structure of a teaching resource recommendation system according to some embodiments of the present invention. The teaching resource recommendation system 400 includes: an acquisition module 401, a processing module 402, and an execution module 403, which are described as follows:

[0104] Acquisition module 401, in the present invention, the acquisition module 401 is mainly used to acquire the learning behavior data of the target student in the historical time period from the teaching terminal device;

[0105] Processing module 402, in the present invention, is used to segment the learning behavior data into learning events based on the behavior characteristics of the target student to obtain multiple teaching and training behavior event segments;

[0106] The processing module 402 of the present invention is further configured to perform a global behavior path graph analysis on all behavior event segments to obtain a cross-segment behavior association graph structure, and then determine explicit trigger labels for the training paths between each behavior event segment based on the association graph structure and the association trigger relationship between each behavior event segment;

[0107] The processing module 402 of the present invention is further used to perform implicit relationship modeling on the training intentions between the various behavioral event segments, obtain the implicit semantic association of the training intentions between the various behavioral event segments, and then determine the implicit dependency labels of the training paths between the various behavioral event segments based on all the implicit semantic associations and the training logic between the various behavioral event segments;

[0108] Execution module 403, the execution module 403 in the present invention is mainly used to jointly model the teaching resource configuration of the training path through the explicit trigger labels and implicit dependency labels of the training path between each behavioral event segment, obtain the joint constraints of the teaching resource configuration, and then perform personalized recommendations of teaching resources based on the joint constraints.

[0109] In addition, the present invention also provides a computer device, which includes a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the above-mentioned teaching resource recommendation method.

[0110] In some embodiments, reference Figure 5 , which is a schematic diagram of the structure of a computer device for implementing a teaching resource recommendation method according to some embodiments of the present invention. The teaching resource recommendation method in the above embodiment can be Figure 5 The computer device 500 shown in FIG. 5 is implemented as shown in FIG. 5 . The computer device 500 includes at least one processor 501 , a communication bus 502 , a memory 503 , and at least one communication interface 504 .

[0111] The processor 501 may be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0112] The communication bus 502 may be used to transmit information between the aforementioned components.

[0113] The memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), a magnetic disk or other magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto. The memory 503 may be independent and connected to the processor 501 via the communication bus 502. The memory 503 may also be integrated with the processor 501.

[0114] Memory 503 is used to store program code for implementing the present invention, and is controlled by processor 501 for execution. Processor 501 is used to execute the program code stored in memory 503. The program code may include one or more software modules. The teaching resource recommendation method in the above embodiment can be implemented by processor 501 and one or more software modules in the program code in memory 503.

[0115] The communication interface 504 uses any device such as a transceiver to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area network (WLAN), etc.

[0116] In a specific implementation, as an example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. A processor herein may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0117] The aforementioned computer device can be a general-purpose computer device or a dedicated computer device. In a specific implementation, the computer device can be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. The embodiments of the present invention do not limit the type of computer device.

[0118] In addition, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned teaching resource recommendation method is implemented.

[0119] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0120] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A teaching resource recommendation method, characterized in that: The steps include: Obtaining learning behavior data of target students in a historical period from teaching terminal devices; Segmenting the learning behavior data into learning events based on the behavioral characteristics of the target students to obtain multiple teaching and training behavioral event segments; Performing a global behavior path graph analysis on all behavioral event segments to obtain a cross-segment behavior association graph structure, and then determining explicit trigger labels for the training paths between each behavioral event segment based on the association graph structure and the association trigger relationship between each behavioral event segment; The implicit relationship modeling of the training intentions between each behavioral event segment is performed to obtain the implicit semantic correlation of the training intentions between each behavioral event segment. Then, based on all the implicit semantic correlations and the training logic between each behavioral event segment, the implicit dependency labels of the training paths between each behavioral event segment are determined. The teaching resource configuration of the training path is jointly modeled by the explicit trigger labels and implicit dependency labels of the training path between each behavioral event segment, and the joint constraints of the teaching resource configuration are obtained, and then personalized recommendations of teaching resources are performed based on the joint constraints.

2. The method according to claim 1, wherein The learning behavior data is segmented into learning events based on the behavioral characteristics of the target students to obtain multiple teaching training behavior event segments, specifically including: extracting behavioral characteristics of the target student in time series from the learning behavior data; Determine the amount of difference in fluctuations of behavioral characteristics within adjacent time windows; When the fluctuation difference exceeds a preset behavioral characteristic fluctuation threshold, it is determined to be an event segmentation boundary; The learning behavior data is divided into multiple continuous behavior event segments according to all event segmentation boundaries.

3. The method according to claim 1, wherein Performing a global behavior path graph analysis on all behavioral event segments, the association graph structure of cross-segment behaviors is obtained, including: Each behavioral event segment is used as a node in the path diagram; Determine the similarity of behavioral patterns between any two behavioral event segments; Constructing initial connection edges between nodes based on the similarity of the behavior patterns; The initial connection edges are topologically optimized based on the time series relationship between the behavior event segments to generate an association graph structure of cross-segment behaviors.

4. The method according to claim 1, wherein Determining the explicit trigger labels of the training paths between the behavioral event segments based on the association graph structure and the association trigger relationship between the behavioral event segments specifically includes: Extracting the connection edge weight between each two behavioral event segments from the association graph structure as the association strength value between each two behavioral event segments; Analyze the correlation and triggering relationship between each behavioral event fragment in the time dimension; Determining explicit dependency values ​​of training paths between each behavioral event segment based on all association strength values ​​and the association trigger relationship; All explicit dependency values ​​are converted into explicit trigger labels for the training paths between each behavioral event segment through preset mapping rules.

5. The method according to claim 1, wherein The implicit relationship modeling of the training intentions between each behavioral event segment is performed to obtain the implicit semantic correlation of the training intentions between each behavioral event segment, which specifically includes: Label each behavioral event segment with a learning objective, and then determine the degree of fit between the learning objectives of each behavioral event segment based on the labeling results; Construct the intention association matrix of the behavioral event segments based on all the fit scores; Based on the intention association matrix, the implicit association semantics of the training intentions between the various behavior event segments are analyzed, and then the implicit semantic association degree of the training intentions between the various behavior event segments is obtained.

6. The method according to claim 1, wherein The implicit dependency labels of the training paths between the behavioral event segments are determined based on all implicit semantic associations and the training logic between the behavioral event segments. Specifically, they include: All implicit semantic correlations are associated and fused with the training logic between each behavioral event segment to obtain the fusion result of the implicit relationship between each behavioral event segment; Learning the dependency pattern of the fusion results through a pre-trained neural network model; An implicit dependency matrix representing the potential logical relationship is output according to the learned dependency pattern, and then the implicit dependency labels of the training paths between the various behavioral event segments are determined through the implicit dependency matrix.

7. The method according to claim 1, wherein The teaching terminal device refers to an intelligent hardware device used to collect, display, interact and record students' learning behavior information during the teaching process.

8. A teaching resource recommendation system, characterized in that: include: An acquisition module is used to obtain the learning behavior data of target students in a historical period from the teaching terminal device; a processing module for performing learning event segmentation on the learning behavior data based on the behavioral characteristics of the target students to obtain a plurality of teaching and training behavioral event segments; The processing module is further configured to perform a global behavior path graph analysis on all behavior event segments to obtain a cross-segment behavior association graph structure, and then determine explicit trigger labels for the training paths between each behavior event segment based on the association graph structure and the association trigger relationship between each behavior event segment; The processing module is further used to perform implicit relationship modeling on the training intentions between the various behavioral event segments, obtain the implicit semantic association of the training intentions between the various behavioral event segments, and then determine the implicit dependency labels of the training paths between the various behavioral event segments based on all the implicit semantic associations and the training logic between the various behavioral event segments; The execution module is used to jointly model the teaching resource configuration of the training path through the explicit trigger labels and implicit dependency labels of the training path between each behavioral event segment, obtain the joint constraints of the teaching resource configuration, and then perform personalized recommendations of teaching resources based on the joint constraints.

9. A computer device comprising a memory and a processor, wherein the memory stores a code, wherein: The processor is configured to obtain the code and execute the teaching resource recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the teaching resource recommendation method according to any one of claims 1 to 7 is implemented.

Images