A content pushing method and system based on a smart teaching terminal

Abstract

The application provides a content pushing method and system based on a smart teaching terminal, relates to the technical field of content pushing of the smart teaching terminal, and provides a content pushing method and system based on a smart teaching terminal.The application carries out multi-dimensional feature marking on teaching resources in a teaching resource library and constructs a resource knowledge graph, then extracts features of learning behavior data of a current class, converts the extracted learning behavior features into learning demand features required by students in the current class through a preset semester demand rule library, matches teaching resources from the teaching resource library based on the learning demand features, finally screens out a push content list with different resource types and matched with the learning demand of students in the current class through an expected utility evaluation and a resource diversity target optimization function, and pushes the push content list to the smart teaching terminal.The above method accurately captures the personalized learning demand of students and accordingly recommends the most suitable teaching resources.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of content pushing of smart teaching terminals, and particularly relates to a content pushing method and system based on a smart teaching terminal. BACKGROUND

[0002] With the development of information technology and the popularity of smart teaching terminals, how to efficiently and accurately push teaching resources suitable for the learning needs of students has become a problem to be solved. The traditional teaching resource pushing method on the smart teaching terminal often lacks in-depth understanding of the personalized needs of students. Different class students have different levels of participation in class, different levels of mastery of knowledge points, and different levels of difficulty of teaching resources adapted to the current class as a whole. Therefore, different classes need personalized, diverse and resource pushing methods. In addition, traditional resource pushing algorithms are mostly based on keyword matching or simple user behavior analysis. The recommendation mechanism based on keyword matching mainly relies on the matching degree between keywords in the resource title, description or tag and user queries to recommend content. Although this method is simple and direct, it ignores the actual content quality and applicability of the resources, which can easily cause the recommended results to be too concentrated on a certain type of resource, ignoring resources that are more beneficial to a specific student group. Not only does this cause the type of pushed resources to lack diversity, but it also leads to low matching degrees of teaching resources and student needs, affecting the teaching effect.

[0003] Therefore, it is necessary to provide a content pushing method and system based on a smart teaching terminal to solve the above technical problems. SUMMARY

[0004] To solve the above technical problems, the present application provides a content pushing method and system based on a smart teaching terminal, which achieves the beneficial effect of personalized and diversified pushing of teaching resources to different student groups in different classes.

[0005] The present application provides a content pushing method based on a smart teaching terminal, which comprises the following steps:

[0006] S1: constructing a resource knowledge graph based on resources in the teaching resource library that have been annotated with multi-dimensional features, and matching the teaching content on the current terminal with the resource knowledge graph to obtain a first candidate set;

[0007] S2: obtaining learning behavior data of the current class through the terminal, and extracting learning behavior features from the learning behavior data;

[0008] S3: converting the learning behavior features into learning demand features with the same dimension as the multi-dimensional annotated resource features using a pre-set learning demand rule library;

[0009] S4: Based on the learning demand characteristics, the resources in the first candidate set are scored using a preset expected utility evaluation formula, the expected utility scores obtained by scoring are arranged in descending order, and at least two resources with expected utility scores meeting a preset score threshold are selected to form a second candidate set;

[0010] S5: At least one resource is selected from the second candidate set using a preset optimization algorithm, and combined into a final push list pushed to the terminal.

[0011] Preferably, step S1 further comprises presetting an upper limit threshold of the set capacity when obtaining the first candidate set.

[0012] Preferably, the step of obtaining the first candidate set comprises:

[0013] extracting a knowledge point set contained in the current teaching content from the current terminal;

[0014] and finding resources associated with the knowledge point set contained in the current teaching content in the knowledge graph, and combining them into candidate resources;

[0015] The candidate resources are classified into multiple resource groups according to resource types, and resources are selected from multiple resource groups according to a preset selection rule to combine them into the first candidate set.

[0016] Preferably, the preset expected utility evaluation formula is:

[0017]

[0018] wherein, is the expected utility score, is the demand relevance score, is the knowledge mastery potential score, is the participation potential score, , , respectively, are the fusion weights of the demand relevance score, the knowledge mastery potential score and the participation potential score.

[0019] Preferably, the demand relevance score is obtained by calculating the distance between the learning demand characteristics and the resource characteristics extracted from the resources with multi-dimensional feature annotations.

[0020] Preferably, the calculation formula of the knowledge mastery potential score is:

[0021]

[0022]

[0023] wherein, is the knowledge mastery potential score, a set of knowledge points contained by the current resource, a set of knowledge points mastered by the class in the current resource, a difficulty level of the current resource, a comprehensive mastery score of the set of knowledge points contained by the current resource by the class.

[0024] Preferably, the participation potential score is obtained by acquiring interaction type features of the current resource and performing weighted fusion calculation based on a preset interaction type feature score standard.

[0025] Preferably, the step S5 further comprises using a preset resource diversity target optimization function to select the resources, and a formula of the preset resource diversity target optimization function is:

[0026]

[0027] wherein, is the resource diversity target optimization function, is a demand correlation score of the i-th resource, is a fusion weight of the demand correlation score, is a resource diversity score, is an adjustment parameter.

[0028] Preferably, the resource diversity score is evaluated by calculating a proportion of different types of resources in the final push list.

[0029] The present application provides a content push system based on a smart teaching terminal, which is applied to a content push method based on a smart teaching terminal, and the content push system comprises:

[0030] a first candidate set generation module configured to construct a resource knowledge graph based on resources in a teaching resource library that have been annotated with multi-dimensional features, and match teaching content on a current terminal with the resource knowledge graph to obtain a first candidate set;

[0031] a learning behavior feature extraction module configured to acquire learning behavior data of a current class through the terminal, and extract learning behavior features from the learning behavior data;

[0032] a learning behavior feature conversion module configured to convert the learning behavior features into learning demand features with the same feature dimension as the multi-dimensionally annotated resource features by using a preset learning demand rule library;

[0033] ​The second candidate set generation module is configured to score the resources in the first candidate set based on the learning demand features by using a preset expected utility evaluation formula, arrange the expected utility scores obtained by the scoring in descending order, and select at least two resources with expected utility scores meeting a preset score threshold to form a second candidate set;

[0034] The push list optimization module is configured to filter at least one resource from the second candidate set by using a preset optimization algorithm, and combine the at least one resource into a final push list to be pushed to the terminal.

[0035] Compared with the related art, the content push method based on the smart teaching terminal has the following beneficial effects:

[0036] The content push method based on the smart teaching terminal provided by the present application comprises the following steps: performing multi-dimensional feature labeling on teaching resources in a teaching resource library and constructing a resource knowledge graph, performing feature extraction on learning behavior data of a current class, converting the extracted learning behavior features into learning demand features required by students in the current class by using a preset semester demand rule library, matching the teaching resources from the teaching resource library based on the learning demand features, and finally filtering a push content list with different resource types matched with the learning demand of students in the current class by using an expected utility evaluation and a resource diversity target optimization function, and pushing the push content list to the smart teaching terminal. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 FIG. 1 is a flowchart of a content push method based on a smart teaching terminal according to the present application;

[0038] Figure 2 FIG. 2 is a schematic diagram of a module structure of a content push system based on a smart teaching terminal according to the present application. DETAILED DESCRIPTION

[0039] The present application will be further described below in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application. In addition, it should be noted that only the parts related to the present application are shown in the drawings for the convenience of description, and not all the structures. Furthermore, the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0040] It should also be noted that, for ease of description, only portions relevant to the present invention are shown in the accompanying drawings, rather than all of the contents. Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the various operations (or steps) as being processed sequentially, many of the operations can be performed in parallel, concurrently, or simultaneously. In addition, the order of the various operations can be rearranged. The process can be terminated when its operations are completed, but may also have additional steps not included in the accompanying drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0041] Example 1

[0042] A content push method based on smart teaching terminal, in the specific implementation process, such as Figure 1 As shown, it shows a flow chart of a content push method based on a smart teaching terminal, and the content push method includes:

[0043] Step S1: Construct a resource knowledge graph based on the resources in the teaching resource library that have been annotated with multi-dimensional features, and match the teaching content on the current terminal with the resource knowledge graph to obtain a first candidate set.

[0044] Specifically, step S1 further includes presetting a set capacity upper limit threshold when acquiring the first candidate set.

[0045] Specifically, the steps for obtaining the first candidate set are:

[0046] Extracting a set of knowledge points contained in the current teaching content from the current terminal;

[0047] And search the knowledge graph for resources related to the knowledge point set contained in the current teaching content, and combine them into candidate resources;

[0048] The candidate resources are classified into a plurality of resource groups according to resource types, and resources are selected from the plurality of resource groups according to preset selection rules to form a first candidate set.

[0049] In the implementation process, the teaching resources in the teaching resource library are marked with multi-dimensional features, including but not limited to knowledge points, difficulty levels, applicable grades, subject affiliations, resource types, and interaction types. A structured resource knowledge graph is constructed for the resources in the teaching resource library that have been marked with multi-dimensional features. The resource knowledge graph represents the relationships between various teaching resources and their relevance to different knowledge points, facilitating subsequent matching with teaching content. After constructing the resource knowledge graph, the set of knowledge points contained in the ongoing teaching content is extracted from the current smart teaching terminal. The set of knowledge points contained in the extracted teaching content is matched with the resources in the resource knowledge graph using the constructed resource knowledge graph. Resources associated with the set of knowledge points contained in the teaching content are searched for, and the selected resources are combined into a first candidate set according to a preset filtering rule. The first candidate set has a preset upper limit threshold for the set capacity, which is set to consider the efficiency of subsequent push content filtering and calculation, to avoid affecting the system response speed due to processing too much data. The acquisition of the first candidate set also needs to consider the diversity of resource types to ensure that the resources in the subsequent filtered push list are not single. First, the set of knowledge points contained in the ongoing teaching activity is extracted from the current smart teaching terminal. Then, all resources associated with the set of knowledge points contained in the ongoing teaching content are searched for from the resource knowledge graph, and the searched relevant resources are combined to form a preliminary candidate resource pool. The preliminary candidate resource pool is classified according to resource types to form multiple resource groups of different resource types. Finally, according to the preset selection rule, a certain number of resources are selected from the multiple resource groups and added to the first candidate set, ensuring that the resources pushed to students have diversity and comprehensiveness, rather than being limited to a certain form. The number of resources in the first candidate set is controlled based on the upper limit threshold of the set capacity to ensure the smooth progress of subsequent data processing.

[0050] Step S2: Obtain the learning behavior data of the current class through the terminal, and extract the learning behavior features from the learning behavior data.

[0051] In the implementation process, the learning behavior data of the current class is collected from the teaching smart terminal. The learning behavior data includes but is not limited to the stage test scores of knowledge points, homework completion, and class question and answer participation data on the current smart teaching terminal. After preprocessing the collected learning behavior data, the learning behavior features are extracted. For example, the participation of students in the class is calculated by summarizing the class interaction of all students, including but not limited to the frequency of answering questions, the proportion of participating in discussions, and the performance of completing group tasks. The overall mastery of knowledge points by the students in the current class and the difficulty of receiving the current knowledge points are evaluated by obtaining the stage test scores of the students in the class and the statistical values of homework completion from the teaching affairs system.

[0052] Step S3: using the preset learning demand rule base, converting the learning behavior features into learning demand features with the same resource feature dimensions as the multi-dimensional feature labels.

[0053] In the implementation process, first, a detailed learning demand rule base is constructed. The learning demand rule base is established based on various theoretical foundations of educational psychology, subject knowledge system and teaching experience, and there are various mapping relationships to convert learning behavior features into learning demand features. The content of the learning demand rule base can include but is not limited to defining different levels of knowledge point mastery and the current class students' overall acceptance of the difficulty of the knowledge points being learned according to the students' test scores, homework completion and other learning behavior data, thereby mapping the difficulty level of the corresponding push resources, mapping the students' interest in the knowledge points being learned according to the students' classroom interaction, the higher the students' interest degree helps the mastery and in-depth exploration of the knowledge points, and in the rule base, the knowledge points with high student participation are mapped to resources with high difficulty level and interactive type. Before conversion, in order to facilitate calculation, the learning behavior features are converted into learning behavior feature vectors, and when converting the learning demand features, considering that different features have different effects on learning effect, appropriate weight coefficients need to be assigned to each feature in the conversion process. Combining all the converted learning demand features, a comprehensive learning demand feature vector is formed, which reflects the characteristics of the push content needed by the current learning state of the class, facilitating subsequent feature matching in the teaching resource library. In addition, in order to better match, the learning demand features need to be the same as the feature dimensions of the resources in the teaching resource library which have been labeled with multi-dimensional features, and the mapping rules in the learning demand rule base are used to achieve this, ensuring that the learning demand features are directly mapped to specific resources in the resource library, thereby achieving accurate content recommendation.

[0054] Step S4: based on the learning demand features, using a preset expected utility evaluation formula to score the resources in the first candidate set, arranging the expected utility scores obtained by scoring in descending order, and selecting at least two resources with expected utility scores meeting a preset score threshold to form a second candidate set.

[0055] In the implementation process, according to the obtained learning demand characteristics, the expected utility score of the resources in the first candidate set is calculated through the preset expected utility evaluation formula. The higher the expected utility score, the higher the matching degree of the resource to the learning demand characteristics, the more suitable the resource is to the current class learning situation, and the more positive promotion effect the resource has on the students in the current class to the knowledge points being learned. All the resources in the first candidate set are sorted in descending order of the expected utility score. According to the preset score threshold, at least two resources with the expected utility score meeting the requirements are selected and added to the second candidate set, so as to ensure that there are enough choices in the final push list. The second candidate set also has a set capacity upper threshold to control the upper limit of the number of resources accommodated by the second candidate set, thereby reducing the data processing and calculation amount.

[0056] Specifically, the preset expected utility evaluation formula is:

[0057]

[0058] wherein, is the expected utility score, is the demand relevance score, is the knowledge mastery potential score, is the participation potential score, , , are the fusion weights of the demand relevance score, the knowledge mastery potential score and the participation potential score respectively.

[0059] In the implementation process, the expected utility evaluation formula includes the demand relevance, the knowledge mastery potential score and the participation potential score. The expected utility score is obtained by weighted fusion. The fusion weights are adjusted according to the teaching objectives and the characteristics of students. The expected utility score is used to screen resources to form the second candidate set.

[0060] Specifically, the demand relevance score is obtained by calculating the distance between the learning demand characteristics and the resource features extracted from the resources with multi-dimensional feature annotation.

[0061] In the implementation process, the demand relevance score represents the correlation degree of the resource to the current learning demand. The learning demand characteristics are converted into vector representation. Then, the feature extraction is performed on each resource in the first candidate set and the resource feature vector is obtained. For example, for each resource in the first candidate set, the Euclidean distance between the resource feature vector and the current class learning demand feature vector is calculated as the demand relevance score of the resource.

[0062] Specifically, the calculation formula of the knowledge mastery potential score is:

[0063]

[0064]

[0065] wherein, is a knowledge mastery potential score, is a knowledge point set contained in the current resource, is a knowledge point set mastered by the class in the current resource, is a difficulty level of the current resource, is a comprehensive mastery score of the class on the knowledge point set contained in the current resource.

[0066] In the specific implementation process, the knowledge mastery potential score represents the potential of the resource to improve the knowledge mastery of the students, and the knowledge mastery potential score formula considers the proportion of the number of knowledge points mastered by the class in the current resource in the number of knowledge points contained in the current resource, the difficulty level of the resource, and the comprehensive mastery score of the class on the knowledge point set contained in the current resource. By considering multiple factors, it is ensured that the pushed resource not only meets the overall learning needs of the students in the class, but also effectively improves the knowledge mastery.

[0067] Specifically, the participation potential score is obtained by acquiring the interaction type features of the current resource and performing weighted fusion calculation based on the preset interaction type feature scoring standard.

[0068] In the specific implementation process, by analyzing the interaction type features in the resource, a scoring standard is preset for each interaction type, and the scoring standard includes but is not limited to interaction frequency, i.e. how many times of interaction is expected to be triggered, interaction depth, i.e. whether the interaction can deeply explore the theme or only answer at the surface level, participation breadth, i.e. whether it can attract students of different levels to participate, the proportion of class student participation, and instant feedback, i.e. whether it can provide timely feedback to enhance learning effect. The participation potential score is obtained by weighted fusion of these scores. Through this method, the expected utility score can more accurately predict the influence of the resource on the participation of the students, thereby optimizing the recommendation strategy and improving the teaching effect.

[0069] Step S5: filtering at least one resource from the second candidate set using a preset optimization algorithm, and combining to form a final push list pushed to the terminal.

[0070] Specifically, in step S5, a preset resource diversity target optimization function is also used for resource selection, and the formula of the preset resource diversity target optimization function is:

[0071]

[0072] wherein, is a resource diversity target optimization function, a demand correlation score for the first resource, a fusion weight for the demand correlation score, a resource diversity score, a regulation parameter.

[0073] Specifically, the resource diversity score is evaluated by calculating the proportion of different types of resources in the final push list.

[0074] In the implementation process, the resources in the final push list are selected from the second candidate set. In order to select resources with resource type diversity and ensure that the pushed content does not fall into the limitation of single type content, an exemplary genetic algorithm is used to find a resource combination that maximizes the resource diversity target optimization function value. The selected resource combination is combined into the final push list for pushing to the terminal. In the generated final push list, the number of each type of resource is counted, the proportion of each type of resource in the total number of resources is calculated, and finally the concept of entropy is used to calculate the resource diversity. Through the preset optimization algorithm and in combination with the resource diversity target optimization function, it is ensured that the selected resources not only have a high expected utility score, but also have resource type diversity.

[0075] The working principle of the content push method based on the intelligent teaching terminal provided by the application is as follows:

[0076] First, the resources in the teaching resource library are annotated in detail through multi-dimensional feature annotation, and a structured resource knowledge graph is constructed based on this, and a preliminary screening is performed to obtain a first candidate set, which provides a basis for subsequent resource matching and recommendation. Then, the learning behavior data of the class is collected and analyzed, and the learning behavior features are extracted therefrom. Then, the preset learning demand rule library is used to convert the learning behavior features into learning demand features with the same feature dimensions as the resources in the teaching resource library that have been annotated with multi-dimensional features, so as to directly match the resources in the resource knowledge graph. This process ensures that the learning behavior of students can be accurately mapped to learning demand and matched to specific educational resources. Secondly, based on the learning demand features, the candidate resources are scored using the expected utility evaluation formula. The at least two resources with expected utility scores meeting the preset score threshold are selected from the second candidate set in descending order according to the expected utility scores. The expected utility evaluation formula comprehensively considers demand correlation, knowledge mastery potential, participation potential and other dimensions to comprehensively evaluate the value of each resource for the student group of the class. Finally,

[0077] ​The preset optimization algorithm is applied in combination with the resource diversity target optimization function to screen an optimal combination from the candidate resources, the optimal combination is a resource combination that maximizes the value of the resource diversity target optimization function, the resource combination is taken as a final resource list pushed to the terminal, and the purpose of personalized and diversified pushing of teaching resources for different classes of student groups is achieved.

[0078] Embodiment two

[0079] A content pushing system based on a smart teaching terminal is applied to a content pushing method based on a smart teaching terminal. Figure 2 As shown in the figure, a module structure schematic diagram of the content pushing system based on the smart teaching terminal is shown, and the content pushing system comprises:

[0080] The first candidate set generation module 100 is configured to construct a resource knowledge graph based on resources in the teaching resource library that have been labeled with multi-dimensional features, and match the teaching content on the current terminal with the resource knowledge graph to obtain a first candidate set.

[0081] The learning behavior feature extraction module 200 is configured to obtain learning behavior data of the current class through the terminal, and extract learning behavior features from the learning behavior data.

[0082] The learning behavior feature conversion module 300 is configured to convert the learning behavior features into learning demand features with the same feature dimension as the multi-dimensional labeled resource features by using a preset learning demand rule library.

[0083] The second candidate set generation module 400 is configured to score the resources in the first candidate set based on the learning demand features by using a preset expected utility evaluation formula, arrange the expected utility scores obtained by the scoring in descending order, and select at least two resources with expected utility scores meeting a preset score threshold to form a second candidate set.

[0084] The pushing list optimization module 500 is configured to screen at least one resource from the second candidate set by using a preset optimization algorithm, and combine the at least one resource into a final pushing list pushed to the terminal.

[0085] The working principle of the content pushing system based on the smart teaching terminal provided by the application is as follows:

[0086] Firstly, the first candidate set generation module 100 constructs a resource knowledge graph based on the teaching resource library that has been labeled with multi-dimensional features, and matches the teaching content on the current terminal with the resource knowledge graph to generate the first candidate set. Then, the learning behavior feature extraction module 200 collects learning behavior data of the current class through the terminal and extracts key learning behavior features from the learning behavior data. Then, the learning behavior feature conversion module 300 converts the learning behavior features into learning demand features with the same dimension as the resource features by using a preset learning demand rule library, so as to facilitate the subsequent matching process. The second candidate set generation module 400 scores the resources in the first candidate set according to the learning demand features by using an expected utility evaluation formula, and selects resources with an expected utility score higher than a preset threshold to form the second candidate set. Finally, the push list optimization module 500 applies an optimization algorithm to select at least one resource from the second candidate set, ensures that the selected resource not only has a high expected utility score, but also maintains a certain diversity, and combines the resource into a resource list that is finally pushed to the terminal.

[0087] The present application is described with reference to flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device realize functions specified in the flowcharts and / or block diagrams. Figure 1 The device that realizes the functions specified in one flow or multiple flows and / or blocks. Figure 1 The device that realizes the functions specified in one flow or multiple flows and / or blocks.

[0088] Those skilled in the art can understand that all or part of the steps of various methods in the above embodiments can be completed by instructing the relevant hardware by means of a program, and the program can be stored in a computer readable storage medium, including Read-Only Memory (ROM), Random Access Memory (RAM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), One-time Programmable Read-Only Memory (OTPROM), Electrically-Erasable Programmable Read-Only Memory (EEPROM), Compact Disc Read-Only Memory (CD-ROM) or other optical disk storage, magnetic disk storage, magnetic tape storage, or any other medium that can be used to carry or store data in a computer readable manner.

[0089] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements, but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the statement "comprising a" does not exclude the existence of additional identical elements in the process, method, article, or apparatus that includes the element.

Claims

1. A content pushing method based on a smart teaching terminal, characterized in that, The content pushing method comprises the following steps: S1: constructing a resource knowledge graph based on resources in a teaching resource library that have been subjected to multi-dimensional feature labeling, and matching teaching content on a current terminal with the resource knowledge graph to obtain a first candidate set; S2: obtaining learning behavior data of a current class through the terminal, and extracting learning behavior features from the learning behavior data; S3: converting the learning behavior features into learning demand features of the same dimension as the multi-dimensional feature-labeled resource features using a preset learning demand rule library; S4: scoring the resources in the first candidate set based on the learning demand features using a preset expected utility evaluation formula, arranging the expected utility scores obtained by the scoring in descending order, and selecting at least two resources with expected utility scores that meet a preset score threshold to form a second candidate set; S5: selecting at least one resource from the second candidate set using a preset optimization algorithm, and combining the at least one resource into a final push list to be pushed to the terminal; The preset expected utility evaluation formula is: wherein, is a prospective utility score, is a demand relevance score, is a knowledge mastery potential score, is an engagement potential score, , , are fusion weights for the demand relevance score, the knowledge mastery potential score, and the engagement potential score, respectively. In step S5, a preset resource diversity objective optimization function is also used for resource selection, and the formula of the preset resource diversity objective optimization function is: wherein, is a resource diversity objective optimization function, is a demand correlation score for the th resource, is a fusion weight for the demand correlation score, is a resource diversity score, is a tuning parameter. 2.The content pushing method based on the intelligent teaching terminal according to claim 1, characterized in that, Step S1 further comprises presetting a set capacity upper threshold when obtaining the first candidate set. 3.The content pushing method based on the intelligent teaching terminal according to claim 2, characterized in that, The obtaining step of the first candidate set is: extracting a knowledge point set contained in the current teaching content from the current terminal; finding resources associated with the knowledge point set contained in the current teaching content in the knowledge graph, and combining the resources into candidate resources; classifying the candidate resources into multiple resource groups according to resource types, and selecting resources from the multiple resource groups according to a preset selection rule to combine the resources into the first candidate set.

4. The content pushing method based on the intelligent teaching terminal according to claim 3, characterized in that, The demand relevance score is obtained by calculating the distance between the learning demand features and the resource features extracted from the resources that have been subjected to multi-dimensional feature labeling.

5. The content pushing method based on the intelligent teaching terminal according to claim 4, characterized in that, The formula for calculating the knowledge mastery potential score is: wherein, is a knowledge mastery potential score, is a set of knowledge points contained in the current resource, is a set of knowledge points that the class has mastered in the current resource, is a difficulty level of the current resource, is a comprehensive mastery score of the class on the set of knowledge points contained in the current resource.

6. The content pushing method based on the intelligent teaching terminal according to claim 5, characterized in that, The participation potential score is obtained by obtaining interaction type features of the current resources, and performing weighted fusion calculation based on a preset interaction type feature scoring standard.

7. The content pushing method based on the intelligent teaching terminal according to claim 6, characterized in that, The resource diversity score is obtained by calculating the proportion of different types of resources in the final push list to evaluate the diversity.

8. A content pushing system based on a smart teaching terminal, characterized in that, A content pushing method based on a smart teaching terminal, applied to one of claims 1 to 7, the content pushing system comprising: a first candidate set generation module configured to construct a resource knowledge graph based on resources in a teaching resource library that have been subjected to multi-dimensional feature labeling, and match teaching content on a current terminal with the resource knowledge graph to obtain a first candidate set; a learning behavior feature extraction module configured to obtain learning behavior data of a current class through the terminal, and extract learning behavior features from the learning behavior data; a learning behavior feature conversion module configured to convert the learning behavior features into learning demand features of the same dimension as the multi-dimensional feature-labeled resource features using a preset learning demand rule library; a learning behavior feature conversion module configured to convert the learning behavior features into learning demand features of the same dimension as the multi-dimensional feature-labeled resource features using a preset learning demand rule library; The second candidate set generating module is configured to score the resources in the first candidate set based on the learning requirement features by using a preset expected utility evaluation formula, arrange the expected utility scores obtained by the scoring in descending order, and select at least two resources with expected utility scores meeting a preset score threshold to form a second candidate set; The push list optimizing module is configured to filter at least one resource from the second candidate set by using a preset optimization algorithm, and combine the at least one resource into a final push list to be pushed to the terminal.

Images