Self-adaptive learning content generation method of smart learning cloud platform

Abstract

The invention is applicable to the technical field of online learning platforms, and particularly relates to a self-adaptive learning content generation method of a smart learning cloud platform, which comprises the following steps: acquiring course learning data of a learner; wherein the course learning data of the learner is used for representing learning data of the learner in a certain course; the course learning data of the learner comprises learning interest data, course progress data, learning behavior data and test evaluation data; determining course learning features of the learner according to the course learning data of the learner; performing feature extraction on the course learning content in the smart learning cloud platform to obtain learning content features; and determining the adaptive learning content of the learner from the course learning content of the smart learning cloud platform according to the course learning features of the learner and the learning content features. According to the method, the adaptive learning content can be dynamically generated, so that the learning efficiency of the learner is improved.

Description

Technical Field

[0001] The present application belongs to the technical field of online learning platforms, and in particular to a method for generating adaptive learning content for a smart learning cloud platform. Background Art

[0002] The Smart Learning Cloud Platform is an online education platform based on cloud computing and intelligent technology, covering a variety of functions, such as online learning resource management, personalized learning path planning, learning progress tracking, teacher-student interaction and evaluation system, etc. The Smart Learning Cloud Platform aims to provide users with a personalized, efficient and interactive learning experience, and uses cutting-edge technologies such as big data, artificial intelligence, and cloud computing to provide teachers, students and managers with a complete set of digital education solutions.

[0003] In the existing technology, the learning status of learners changes dynamically. In traditional learning platforms, course recommendations and learning paths are often static and cannot be dynamically adjusted according to the learners' real-time performance, resulting in some learners being unable to obtain learning resource recommendations or adaptive learning guidance in a timely manner when they encounter bottlenecks in the learning process.

[0004] In summary, traditional learning platforms have the problem of poor dynamic adaptability in the process of generating learning content. Summary of the invention

[0005] The embodiment of the present application provides an adaptive learning content generation method for a smart learning cloud platform, which can solve the problem of poor dynamic adaptability in the process of generating learning content on a traditional learning platform in related technologies.

[0006] In a first aspect, an embodiment of the present application provides a method for generating adaptive learning content for a smart learning cloud platform, comprising: Acquire the course learning data of the learner; wherein the course learning data of the learner is used to represent the learning data of the learner in a certain course; the course learning data of the learner includes learning interest data, course progress data, learning behavior data and test evaluation data; Determining the course learning characteristics of the learner according to the course learning data of the learner; Extract features of course learning content in the smart learning cloud platform to obtain learning content features; According to the course learning characteristics of the learner and the learning content characteristics, the learner's adaptive learning content is determined from the course learning content of the smart learning cloud platform.

[0007] The above technical solutions in the embodiments of the present application have at least the following technical effects: The adaptive learning content generation method of the smart learning cloud platform provided in this application first obtains the learner's course learning data (learner's learning data in a certain course) to help learn the learner's learning interest data, course progress data, learning behavior data and test evaluation data, and then determines the learner's course learning characteristics based on the learner's course learning data, and then extracts features of the course learning content in the smart learning cloud platform to obtain learning content features, and finally determines the learner's adaptive learning content from the course learning content of the smart learning cloud platform based on the learner's course learning features and learning content features. Since the learner's learning characteristics are dynamically changing, as the course progresses and the evaluation data accumulates, the smart learning cloud platform can adjust the learner's learning content in real time through this method, realize adaptive learning, and help the learning process to be more in line with the learner's growth trajectory. By analyzing the learner's course learning data, the smart learning cloud platform can generate personalized learning paths for different learners, enabling learners to learn efficiently according to their own learning needs and rhythm, thereby improving learning efficiency.

[0008] In a second aspect, an embodiment of the present application provides an adaptive learning content generation device for a smart learning cloud platform, comprising: An acquisition unit, used for acquiring the course learning data of the learner; wherein the course learning data of the learner is used to represent the learning data of the learner in a certain course; the course learning data of the learner includes learning interest data, course progress data, learning behavior data and test evaluation data; A course learning feature obtaining unit, used to determine the course learning feature of the learner according to the course learning data of the learner; A learning content feature obtaining unit is used to extract features of course learning content in the smart learning cloud platform to obtain learning content features; The adaptive learning content determination unit is used to determine the adaptive learning content of the learner from the course learning content of the smart learning cloud platform according to the course learning characteristics of the learner and the learning content characteristics.

[0009] In the third aspect, an embodiment of the present application provides an adaptive learning content generation device for a smart learning cloud platform, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in any one of the embodiments of the first aspect when executing the computer program.

[0010] It can be understood that the beneficial effects of the second to third aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0012] Figure 1 It is a flowchart of a method for generating adaptive learning content on a smart learning cloud platform provided in one embodiment of the present application; Figure 2 It is a schematic diagram of the implementation process of step S200 in the adaptive learning content generation method of the smart learning cloud platform provided in the embodiment of the present application; Figure 3 It is a schematic diagram of the implementation process of step S300 in the adaptive learning content generation method of the smart learning cloud platform provided in the embodiment of the present application; Figure 4 It is a schematic diagram of the implementation process of step S400 in the method for generating adaptive learning content of the smart learning cloud platform provided in the embodiment of the present application; Figure 5 It is a structural diagram of an adaptive learning content generation device of a smart learning cloud platform provided in an embodiment of the present application; Figure 6 It is a structural diagram of the adaptive learning content generation device of the smart learning cloud platform provided in the embodiment of the present application. DETAILED DESCRIPTION

[0013] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0014] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0015] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0016] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0017] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0018] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0019] In related technologies, learners' learning status changes dynamically. In traditional learning platforms, course recommendations and learning paths are often static and cannot be dynamically adjusted according to learners' real-time performance, resulting in some learners being unable to obtain learning resource recommendations or adaptive learning guidance in a timely manner when they encounter bottlenecks in the learning process.

[0020] To solve the above problems, the embodiment of the present application provides a method for generating adaptive learning content for a smart learning cloud platform. In this method, firstly, by acquiring the learner's course learning data (learner's learning data in a certain course), it is beneficial to know the learner's learning interest data, course progress data, learning behavior data and test evaluation data, and then determine the learner's course learning characteristics according to the learner's course learning data, and then extract the characteristics of the course learning content in the smart learning cloud platform to obtain the learning content characteristics, and finally determine the learner's adaptive learning content from the course learning content of the smart learning cloud platform according to the learner's course learning characteristics and learning content characteristics. Since the learner's learning characteristics are dynamically changing, with the progress of the course and the accumulation of evaluation data, the smart learning cloud platform can adjust the learner's learning content in real time through this method, realize adaptive learning, and help the learning process to be more in line with the learner's growth trajectory. By analyzing the learner's course learning data, the smart learning cloud platform can generate personalized learning paths for different learners, so that learners can learn efficiently according to their own learning needs and rhythm, thereby improving learning efficiency.

[0021] The adaptive learning content generation method for the smart learning cloud platform provided in the embodiment of the present application can be applied to the adaptive learning content generation device of the smart learning cloud platform. At this time, the adaptive learning content generation device of the smart learning cloud platform is the executor of the adaptive learning content generation method for the smart learning cloud platform provided in the embodiment of the present application. The embodiment of the present application does not impose any restrictions on the specific type of the adaptive learning content generation device of the smart learning cloud platform.

[0022] For example, the adaptive learning content generation device of the smart learning cloud platform can be a cellular phone, a mobile phone, a tablet computer, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), a desktop computer, a smart large screen, a smart TV, a handheld device with wireless communication function, a computing device or other processing device connected to a wireless modem, a computer, a laptop computer, customer premise equipment (CPE) and / or other devices for communicating on a wireless system and a next-generation communication system, for example, a mobile terminal in a 5G network or a mobile terminal in a future evolved public land mobile network (PLMN), etc.

[0023] In order to better understand the adaptive learning content generation method of the smart learning cloud platform provided in the embodiment of the present application, the specific implementation process of the adaptive learning content generation method of the smart learning cloud platform provided in the embodiment of the present application is exemplarily introduced below.

[0024] Figure 1 A schematic flow chart of a method for generating adaptive learning content of a smart learning cloud platform provided in an embodiment of the present application is shown. The method for generating adaptive learning content of a smart learning cloud platform includes: S100, obtaining the course learning data of the learner. The course learning data of the learner is used to represent the learning data of the learner in a certain course; the course learning data of the learner includes learning interest data, course progress data, learning behavior data and test evaluation data.

[0025] It can be understood that learning interest data is used to understand learners' interest tendencies in course content. Learning interest data may include: learning time preference, when learners tend to study; course selection preference, the types of courses or topics that learners often choose; content access frequency, what types of content (videos, documents, exercises, etc.) learners access more frequently; interactive behavior, whether learners participate in discussions in the course discussion area and question-and-answer area.

[0026] Course progress data is used to track learners' learning progress in the course. Course progress data may include: learning progress, the learner's overall progress in the course; course unit progress, the completion status of specific chapters; learning time, the time spent in each unit or chapter.

[0027] Learning behavior data is used to analyze learners' specific learning activities on the smart learning cloud platform. Learning behavior data may include: learning frequency, the frequency of learning per day or per week; learning path, whether learning is done according to the recommended path or there is skipped learning; repeated learning, whether there is repeated learning of certain content.

[0028] Test evaluation data is used to evaluate learners’ mastery of course knowledge. Test evaluation data may include: test scores, test scores for each unit in the course; error analysis, knowledge points or question types that are prone to errors.

[0029] For example, the smart learning cloud platform provides a data interface or log system, through which learners' learning interest data, course progress data, learning behavior data and test evaluation data can be obtained. The interface can be an API interface or a database access. The log system can record learners' operation behaviors, such as login time, accessed courses, learning time, tests, etc. The acquired course learning data can be cleaned, redundant information can be removed, integrated into a unified learning data set, and data of different dimensions can be classified.

[0030] Through this step, the learner's course learning data can be fully obtained, thereby providing basic data support for the subsequent generation of learner's adaptive learning content.

[0031] S200, determining the course learning characteristics of the learner according to the course learning data of the learner.

[0032] For example, frequency analysis or cluster analysis (such as K-Means) can be used to calculate the content types and subject categories preferred by learners, which are the features corresponding to the learning interest data. For example, the selected algorithm is used to count the frequency and total time that learners visit various course contents, extract learner preferences (such as preferring videos or text materials); analyze the types of resources that learners like (such as video explanations, document reading, or test participation); and identify learners' areas of interest (such as programming, data science, etc.) by analyzing the categories and subjects of the courses that learners participate in.

[0033] Time series analysis can be used to identify the changing trends of learners' learning time. For example, the completion ratio of learners in the course, that is, the learning progress percentage, can be calculated; the relationship between the learner's completion progress and time can be calculated to determine the learner's learning speed (such as the amount of content learned per day / week). The learning progress percentage and the learner's learning speed can be determined as the corresponding features of the course progress data.

[0034] The learning frequency of learners can be calculated, such as the number of times they study per day or per week; the active time periods of learners (such as daytime and nighttime) can be determined through the timestamp data in the learning behavior data; path analysis algorithms can be used, such as Markov chain analysis of the pattern of learners' learning paths, that is, tracking learners' behavior paths to determine whether learners study in the recommended course sequence or whether there is skipping learning. The learning frequency of learners, the learning time period of learners, and the learning path of learners can be used as the corresponding features of learning behavior data.

[0035] The test scores can be extracted from the test evaluation data, and the average score of the test scores can be calculated. If the average score is greater than 80, the learner's performance characteristics can be high-level; if the average score is less than 50, it is low-level. Clustering algorithms can be used to find out the knowledge points where learners often make mistakes, determine which knowledge points the learners have weak links, and generate knowledge point mastery portraits. The performance level and knowledge point mastery portrait can be used as the corresponding features of the test evaluation data.

[0036] Based on the analysis results of the above dimensions, the features corresponding to the learning interest data, the features corresponding to the course progress data, the features corresponding to the learning behavior data and the features corresponding to the test evaluation data can be integrated to generate the learner's course learning features, which are used to characterize the learner's overall learning situation.

[0037] Through this step, the learner's course learning characteristics can be extracted based on the learner's course learning data. Course learning characteristics can not only be used to generate personalized learning content, but also help teachers understand the learner's learning status and optimize teaching strategies.

[0038] In one possible implementation, see Figure 2 , S200, determining the learner's course learning characteristics according to the learner's course learning data, including: S210, extracting features from the learning interest data to obtain a first feature.

[0039] For example, the interactive data (such as questions and comments) in the learner's learning interest data can be segmented, and the TF-IDF algorithm can be used to calculate the weight of each word, and the keywords with high frequency and high weight can be extracted as the key representation of learning interest. TF-IDF (Term Frequency-Inverse Document Frequency) can measure the importance of keywords. For example, the machine learning and regression model keywords in a certain course have high weights, indicating that the learner is interested in machine learning and regression models. The extracted keywords can be clustered using topic modeling algorithms such as LDA (Latent Dirichlet Allocation) to summarize the learner's main interest topics in the course, such as data analysis, algorithm implementation, mathematical derivation, etc.

[0040] The browsing time, browsing frequency and number of interactions in the learning interest data can be calculated to quantify the learner's interest intensity and assign interest weights to different content or topics.

[0041] After extracting the learner's interest topics and interest strength, an interest distribution vector can be constructed to represent the learner's interest in the course. Each dimension of the vector can represent a learning topic or knowledge point, for example, [machine learning, algorithm design, data cleaning, regression model]. The value corresponding to each dimension of the vector represents the learner's interest strength in the learning topic or knowledge point. For example, if the learner has a high interest in machine learning and a medium interest in data cleaning, an interest distribution vector such as [0.8, 0.5, 0.3, 0.7] can be obtained, where the higher the value, the greater the interest strength. The interest distribution vector represents the learner's interest weight in different topics or knowledge points.

[0042] The interest distribution vector is further processed to extract the first feature. The weights in the interest distribution vector can be normalized so that the interest intensity value is between a uniform scale (such as 0-1) to facilitate subsequent analysis and recommendation; if the interest vector dimension is too high, it can be reduced by PCA (principal component analysis) or clustering methods to extract several core interest topics to form a more concise first feature. The first feature can be represented as a vector, such as [0.8, 0.5, 0.3, 0.7], or a combination of refined topics, such as machine learning: high, data analysis: medium, data cleaning: low.

[0043] S220, numerically process the course progress data to obtain a second feature.

[0044] For example, the course completion rate can be calculated based on the learning progress in the course progress data, and the calculation formula is: ,in, Indicates the progress completion rate, Indicates that the chapter has been completed. Indicates the total number of chapters in the course. Between 0 and 1. For example, if a course contains 10 chapters and a learner has completed 6 chapters, the progress completion rate is 0.6.

[0045] Courses can be divided into different stages according to learning difficulty, such as basic, advanced, and advanced, and the current learning stage of the learner can also be quantified. For example, the entire course can be divided into 3 stages, and each stage of the course is mapped to a value, such as 1 for basic, 2 for advanced, and 3 for advanced. If the learner is currently in the advanced stage, then the learning stage eigenvalue can be 2 / 3=0.67.

[0046] After obtaining the numerical indicators of progress completion and learning stage, the progress completion and learning stage can be combined into a comprehensive feature vector (second feature). The second feature can be expressed as a vector, such as [0.6, 0.67].

[0047] S230, extracting features from the learning behavior data and the test evaluation data to obtain a third feature.

[0048] For example, the total learning time can be extracted from the learning behavior data to calculate the average learning time of each learner logging in; the total number of visits of the learners in the learning behavior data can be counted to calculate the average number of visits per day.

[0049] All test scores can be extracted from the test evaluation data to calculate the learner's average score in all tests; the learner's average error rate in all tests can be calculated; the average time it takes for the learner to answer each question can be calculated to reflect the learner's proficiency in the knowledge points.

[0050] The features calculated above are integrated into a feature vector, i.e., the third feature. For example, if the learner's total learning time is 100 hours, the average learning time is 2 hours, the average number of visits per day is 3 times, the average score is 85 points, the average error rate is 0.15, and the average answering time is 1 minute, the feature vector can be expressed as [100, 2, 3, 85, 0.15, 1].

[0051] Optionally, S230, feature extraction is performed on the learning behavior data and the test evaluation data to obtain a third feature, including: S231, constructing a learning time sequence diagram based on the learning behavior data and the test evaluation data.

[0052] It can be understood that the learning time series diagram is used to represent the relationship between the learner's learning behavior over time and the test evaluation data.

[0053] Exemplarily, a learning behavior can be used as a learning behavior node in a learning sequence diagram, and the learning behavior can include the learning activity information of the learner at a certain point in time, such as learning time, number of reviews, access frequency, etc. The evaluation data such as the test score, error rate, answering speed, etc. of the learner at a certain point in time can be used as a test evaluation node in the learning sequence diagram. Each node has attributes, and the node attributes can be numerical features extracted from the learning behavior data and the test evaluation data. For example, the learning behavior node attributes can include: learning time, learning frequency, number of reviews, etc.; the test evaluation node attributes can include: average score, error rate, answering time, etc.

[0054] The learning behavior nodes and test evaluation nodes can be connected in time order, and the direction of the edge indicates the flow of time. Learning behavior nodes and time points The test evaluation nodes of are connected by an edge. According to the attribute similarity between different nodes, different time nodes of the same learner (such as between multiple similar learning behaviors) or nodes of similar learners are connected to capture similar learning patterns.

[0055] In the learning time series graph, edges can be assigned weights, which represent the correlation between nodes. For example, weights can be calculated based on the time interval or attribute similarity between two nodes, so that close or similar nodes are more closely connected.

[0056] S232, extracting features from the learning time series graph using the first graph neural network to obtain a third feature, wherein the first graph neural network is trained based on learning behavior data and test evaluation data of different learners.

[0057] Exemplarily, the learning time series graph is input into the first graph neural network. Through the graph convolution operation, the features of each node in the learning time series graph are aggregated with the features of the adjacent nodes, and the feature representation of the node is updated. Graph convolution can help learn the association between nodes, so that the model can capture the interaction between learning behavior and test data. Based on the graph convolution, the attention mechanism is introduced to dynamically adjust the weights of different neighbor nodes to further improve the aggregation effect of the model. The graph attention layer can capture nodes with greater influence in the learning behavior, thereby extracting key node features in the learning process. Through multi-layer graph convolution operations, the feature vector of each node is updated to an aggregated feature vector containing the neighbor node information corresponding to each node.

[0058] After multiple graph convolutions, the node features of the last layer represent the global features of the learner, which can be regarded as the third feature. The first graph neural network outputs the third feature.

[0059] Exemplarily, S231, constructing a learning time sequence diagram according to the learning behavior data and the test evaluation data, including: S2311, extracting time information of learning behavior data and time information of test evaluation data.

[0060] For example, the timestamp (such as the specific date and time) of each learning behavior of the learner can be extracted from the learning behavior data, and the time information of the learning behavior data can be used to determine the sequence between learning behaviors. The timestamp of each test or evaluation of the learner can be extracted from the test evaluation data to indicate the time point of each evaluation.

[0061] By obtaining timestamp information, the sequence between learning behavior and test evaluation can be clarified on the timeline.

[0062] S2312, determining a learning behavior node of the learning sequence diagram based on the learning behavior data.

[0063] It can be understood that the purpose of constructing learning behavior nodes and test evaluation nodes is to be able to represent learners' learning activities and evaluation results in the learning sequence diagram, so as to facilitate the subsequent analysis of learners' learning patterns and knowledge mastery.

[0064] Exemplarily, each learning behavior in the learning behavior data may be regarded as a learning behavior node, and each learning behavior node may have a set of attributes, which may include the time when the learning behavior occurs, the learning duration, and the access frequency.

[0065] S2313, based on the test evaluation data, determine the test evaluation node of the learning timing diagram.

[0066] Exemplarily, each test evaluation data in the test evaluation data may be regarded as a test evaluation node, and each test evaluation node may have a set of attribute nodes, which may include the test occurrence time, test score, error rate, and average answering time.

[0067] S2314, based on the time information of the learning behavior data, the learning behavior nodes of the learning sequence diagram, the time information of the test evaluation data, and the test evaluation nodes of the learning sequence diagram, determine the time sequence edges of the learning sequence diagram. The time sequence edges of the learning sequence diagram are used to represent the sequence between different learning behaviors of the learner and the test evaluation results.

[0068] It can be understood that the time sequence edge is used to connect the learning behavior nodes and the test evaluation nodes at different time points to represent the sequence of the learning behavior nodes and the test evaluation nodes.

[0069] For example, the nodes can be connected according to the time order in the time information of the learning behavior data and the time information of the test evaluation data, that is, the node that occurs first points to the node that occurs later. A test evaluation node will appear after the learning behavior node to indicate the effect of the learning behavior, so a time order edge can be added between adjacent learning behavior nodes and test evaluation nodes. For example, at a time point The learning behavior node at the time point Before the test evaluation node, create a slave node Point to Node Similarly, if two learning behavior nodes are at adjacent time points, a time sequence edge can also be added to indicate the order in which the behaviors occur.

[0070] S2315, determining edge attributes of the learning timing graph based on the time information of the learning behavior data and the time information of the test evaluation data.

[0071] It can be understood that edge attributes are used to represent information such as the strength of the relationship between nodes and the time interval, and the determination of edge attributes can provide additional information for subsequent feature extraction.

[0072] For example, the time difference between adjacent nodes can be calculated and used as the attribute of the edge. With Node The time difference between them is 3 hours, so the node With Node The attribute of the edge between is 3.

[0073] S2316, constructing a learning timing graph according to the learning behavior nodes of the learning timing graph, the test evaluation nodes of the learning timing graph, the time sequence edges of the learning timing graph, and the edge attributes of the learning timing graph.

[0074] For example, assuming that the learning behavior data has learning behavior 1: time is , the learning time is 1 hour; learning behavior 2: the time is , the learning time is 2 hours. Test evaluation data 1: time is , the test score is 80, and the error rate is 10%. Construct the nodes of the learning sequence diagram: Learning behavior 1 is the learning behavior node (time ), learning behavior 2 is the learning behavior node (time ), test evaluation data 1 is the test evaluation node (time ); Construct the time sequence edge and edge attributes of the learning timing graph: the time sequence edge starts from the learning behavior node Point to the learning behavior node , the edge attribute is the time interval ; Time order edge from learning behavior node Point to the test evaluation node , the edge attribute is the time interval .

[0075] The learning time series diagram constructed through the above steps provides rich data support for subsequent feature extraction and analysis, and helps to capture learners' behavior patterns and knowledge mastery trends in the course.

[0076] S240, performing feature fusion on the first feature, the second feature, and the third feature to obtain the course learning feature of the learner.

[0077] Exemplarily, feature fusion of the first feature, the second feature, and the third feature can be performed by feature vector concatenation, that is, the dimensions of the first feature, the second feature, and the third feature are directly concatenated into a comprehensive feature vector as the learner's course learning feature. For example, the first feature is [0.8, 0.5, 0.3, 0.7], the second feature is [0.6, 0.67], and the third feature is [100, 2, 3, 85, 0.15, 1]. The first feature, the second feature, and the third feature are concatenated into a comprehensive feature vector: the learner's course learning feature = [0.8, 0.5, 0.3, 0.7, 0.6, 0.67, 100, 2, 3, 85, 0.15, 1].

[0078] The first feature, the second feature, and the third feature can be fused by weighted fusion, that is, weighted summation is performed according to the weight value corresponding to each feature to obtain a comprehensive feature vector (learner's course learning feature). For example, the weight value of the first feature is 0.3, the weight value of the second feature is 0.4, and the weight value of the third feature is 0.3. The weighted summation of each feature vector according to the weight value is the comprehensive feature vector = 0.3 × first feature + 0.4 × second feature + 0.3 × third feature. The weight value of each different feature can be set manually.

[0079] Through the above steps, the features of learning interest data, course progress data, learning behavior data and test evaluation data can be extracted and integrated to construct a comprehensive and multi-dimensional course learning feature of learners, thereby achieving personalized and dynamic learning support and content recommendation.

[0080] Optionally, S240, performing feature fusion on the first feature, the second feature, and the third feature to obtain the course learning feature of the learner includes: S241, obtaining a first weight, a second weight, and a third weight, wherein the first weight is used to characterize the weight value corresponding to the first feature, the second weight is used to characterize the weight value corresponding to the second feature, and the third weight is used to characterize the weight value corresponding to the third feature.

[0081] For example, a large number of learners' historical data can be analyzed by statistical analysis methods to determine the importance of different features in predicting learning effects or generating learning content accuracy, so that the weight of each feature can be set manually. Statistical analysis methods can include: feature correlation analysis, analyzing the correlation between the first, second and third features of a large number of learners and learning outcomes, and the features with higher correlation can be assigned higher weights; machine learning feature importance analysis, using machine learning models (such as decision trees, random forests, etc.) to calculate the importance score of each feature in the model, and using the score results to determine the weight.

[0082] You can optimize the value of each feature weight through multiple groups of A / B tests. For example, set the initial weights (such as 0.3, 0.3, 0.4), and test them over a period of time to observe the impact of each group of weight combinations on the learning effect. Then, adjust the weights according to the impact of each group of weight combinations on the learning effect until the optimal weight combination is found.

[0083] S242, calculating a weighted sum of the first feature, the second feature, and the third feature according to the first weight, the second weight, and the third weight to obtain a fused feature vector.

[0084] Exemplarily, the three corresponding feature vectors can be weighted summed using the three obtained weight values ​​to obtain a fused feature vector. Assume that the first feature vector is The second eigenvector is and the third eigenvector is , the first weight value is The second weight value is and the third weight value is , the calculation formula of the fused feature vector is , Represents the fused feature vector.

[0085] For example, the first weight , the second weight , the third weight , the first eigenvector , the second eigenvector , the third eigenvector , then the fusion feature vector , that is, the fused feature vector is .

[0086] S243, determining the fused feature vector as the learner's course learning feature.

[0087] For example, the fused feature vector is used as the learner's course learning feature. For example, the fused feature vector is , then the learner's course learning characteristics are .

[0088] By integrating the first, second and third features according to the weights, the learner's course learning features can fully reflect the learner's interests, learning progress and knowledge mastery. This comprehensive feature can be used for personalized recommendations, learning achievement tracking and adaptive learning path design in intelligent learning systems.

[0089] S300, extracting features of course learning content in the smart learning cloud platform to obtain learning content features.

[0090] For example, the course learning content of the smart learning cloud platform is mostly presented in text form (such as course introduction, chapter content, articles, PDF documents, etc.), and the text content features of the course learning content can be extracted: the main content of the course can be analyzed using topic models (such as LDA, Latent Dirichlet Allocation) to extract the core topics and related topic keywords of the course; keywords of the course content can be extracted through technologies such as TF-IDF (Term Frequency-Inverse DocumentFrequency) or Word2Vec; and a brief summary of the course content can be generated using a text summarization algorithm.

[0091] The courses in the smart learning platform are divided into different chapters or modules, and the hierarchical features of the course content can be extracted: hierarchical information such as chapters, units, and sub-modules of the course can be extracted; the knowledge points involved in each chapter or unit in the course content can be extracted, and a correlation matrix between knowledge points and chapters can be established; and a learning path map can be generated by analyzing the learning order of each chapter of the course.

[0092] Course difficulty features can help recommend appropriate learning content based on the learner's level. Course difficulty features can be extracted: NLP tools can be used to analyze the complexity of course texts, such as sentence length, vocabulary difficulty, grammatical complexity, etc., to determine the difficulty of the course language; based on the knowledge points involved in the course, combined with the test scores or feedback of other learners, the difficulty coefficient of the knowledge points can be calculated, and the difficulty of the course content can be determined based on the difficulty coefficient; if the course contains exercises or tests, the difficulty level of the course can be calculated by statistically analyzing the accuracy of the answers, the time it takes to answer the questions, and other data.

[0093] The course content in the smart learning cloud platform is not limited to text, but can also include video, audio, pictures, interactive content, etc. Multimedia content features can be extracted: extract video duration, playback speed, chapter division and other features; the explanation content in the video can be converted into text through speech recognition technology (such as ASR, Automatic Speech Recognition), and the converted text content can be further analyzed and keywords extracted; image processing technology can be used to extract the features of visual elements such as pictures and charts in the course, and analyze the explanatory function of visual element features in the course.

[0094] The text content features, hierarchical structure features, course difficulty features and multimedia content features can be integrated to form learning content features.

[0095] Through this step, the multi-dimensional features of the course learning content in the smart learning cloud platform can be fully extracted, which can provide basic support for subsequent personalized recommendations, learning path planning, and adaptive learning content push.

[0096] In one possible implementation, see Figure 3 , S300, extract features of the course learning content in the smart learning cloud platform to obtain learning content features, including: S310, obtaining the learner's course type information according to the learner's course learning data.

[0097] It can be understood that the course type information may include subjects (such as mathematics, computer science), course levels (such as elementary, intermediate, advanced), etc., which are used to filter out matching learning resources from the resource library of the smart learning cloud platform.

[0098] For example, the description of the learner's learning content can be analyzed through natural language processing (NLP) technology (such as TF-IDF, keyword extraction or topic modeling), and high-frequency words and subject keywords can be extracted to identify the field to which the course belongs. The extracted keywords are matched with a preset course type library (such as mathematics, computer science, physics, etc.) to determine the course type.

[0099] If the learner's course learning data contains labels or category information, these labels are directly used for classification. For example, if the learner's course label is data science or machine learning, the course type can be directly classified as computer science or artificial intelligence.

[0100] S320, obtaining course learning content that matches the learner's course type information from the learning resource library of the smart learning cloud platform to obtain a course information package.

[0101] It can be understood that the course information package may include the course name to facilitate learners to identify each course; the course content module to extract the specific chapters, knowledge points and learning objectives contained in the course; the resource type to identify the content form of the course, such as video, text, exercises, interactive content, etc.; the learning path, the recommended learning order or chapter order of the course, to help learners learn in sequence.

[0102] For example, keywords in the course type information (such as machine learning, programming basics, etc.) can be used to match in the learning resource library of the smart learning cloud platform to find courses containing these keywords; if the learning resources have pre-labeled tags (such as field tags, knowledge point tags, difficulty tags, etc.), you can directly use the course type tags to filter. For example, if the learner's course type information is data science, then filter out course content with the data science tag; if the learning resource library is classified by subject or theme, you can directly query the matching courses in the corresponding subject catalog based on the course type information.

[0103] All course content that matches the course type information can be aggregated into a course information package, which is a collection of multiple chapters and resource modules, which is convenient for subsequent feature extraction and learning content generation. Each matching course resource (chapter, video, test, etc.) can be added to the course information package. If there are duplicate or similar resources in the queried course content, the content can be optimized by indicators such as course rating, resource popularity, and completion rate, and high-quality and popular resources can be retained.

[0104] The course information package, after screening, combination and organization, contains the course content that matches the learner's course type information and has been organized according to the structured information of the learning resources (such as resource type, knowledge points, and difficulty).

[0105] S330, using a second graph neural network to extract graph structure features of the course learning content in the course information package to obtain course graph structure features.

[0106] For example, the course learning content in the course information package can be constructed into a learning content graph structure. After the course graph is constructed, the graph convolutional network (GCN) or graph attention network (GAT) in the second graph neural network is used to update the features of each node in the learning content graph. Through graph convolution, the features of the node will be combined with the information of the node's neighboring nodes, so as to learn the role of each node in the graph structure. After multiple layers of graph convolution, the deep features of each node are obtained, and the deep features of each node are summarized into an overall course graph structure feature.

[0107] Optionally, S330, using a second graph neural network to extract graph structure features of the course learning content in the course information package to obtain course graph structure features, including: S331, dividing the course learning content in the course information package into knowledge modules to obtain a knowledge module collection, and determining each knowledge module in the knowledge module collection as a node of the course knowledge graph. The course knowledge graph is used to represent the relationship between each knowledge module.

[0108] For example, the learning content can be divided into multiple independent knowledge modules based on the structure of the course learning content (such as chapter titles, knowledge point descriptions). Each knowledge module can have a single theme or core concept. For example, knowledge points such as linear regression, data cleaning, and probability distribution can be used as knowledge modules. The attributes that each knowledge module may contain may include title, content summary, key knowledge points, module difficulty, etc. After segmentation, a collection of knowledge modules is obtained, in which each knowledge module represents an independent knowledge point in the course learning content.

[0109] Each knowledge module in the knowledge module set can be used as a node in the course knowledge graph, and the node name can be the theme or knowledge point of the corresponding knowledge module. The attributes that each node can carry can include the knowledge module title, content summary, difficulty coefficient, related exercises, etc. The node set of the course knowledge graph is equivalent to the knowledge module set.

[0110] S332, determining the edges of the course knowledge graph according to the relationship information of each knowledge module in the knowledge module collection and the nodes of the course knowledge graph.

[0111] Exemplarily, in order to construct a course knowledge graph, edges may be added between nodes to represent the relationship between knowledge modules. The relationship information of the knowledge modules may include prerequisite relationships, dependency relationships, and the like.

[0112] Prerequisites: Some knowledge modules are prerequisites for other knowledge modules, meaning that learners may need to master the former to better understand the latter. For example, linear algebra is a prerequisite for linear regression.

[0113] Dependency: The relationship between knowledge modules, for example, the probability basic module is closely related to the statistics module.

[0114] According to the relationship information of each knowledge module, an edge can be added for every two nodes with a relationship. The direction of the edge can indicate the order of dependency or the order of prerequisite relationships. For example, adding a directed edge from linear algebra to linear regression means that learners can learn linear algebra before learning linear regression. Attributes can be added to the edge, such as the relationship type (such as prerequisite relationship, dependency relationship) and the relationship weight. The weight can be defined according to the closeness of the relationship or the strength of the dependency.

[0115] S333, construct a course knowledge graph based on the nodes and edges of the course knowledge graph.

[0116] For example, after determining the nodes (knowledge modules) and edges (relationships between knowledge modules) of the course knowledge graph, a complete course knowledge graph can be constructed. The structure of the course knowledge graph is a directed graph or an undirected graph, which may depend on the relationship between the course content. The course knowledge graph can be represented by an adjacency matrix or an adjacency list, that is, , Represents the course knowledge graph; A set of nodes representing the course knowledge graph, including all modules in the knowledge module collection; The edge set representing the course knowledge graph contains all the relations between knowledge modules.

[0117] S334, using the second graph neural network to calculate the structural feature vector of each node in the course knowledge graph, and obtain the structural features of the course graph.

[0118] For example, after the course knowledge graph is constructed, the second graph neural network (GNN) can be used to extract structural features of each node in the course knowledge graph. The extraction process uses the graph convolution mechanism of the second graph neural network to learn the features of the nodes in the graph structure.

[0119] The node features and edge relationships of the course knowledge graph are input into the second graph neural network, and graph convolution (such as graph convolution network GCN or graph attention network GAT) is used for feature extraction. The graph convolution operation aggregates the feature vector of each node with the features of its neighboring nodes to learn the role of the node in the knowledge graph structure. Through multi-layer convolution operations, the node features will fuse the information of its multi-layer neighboring nodes to form a higher-dimensional feature representation. After calculation by the second graph neural network, the feature vector of each node is updated to an aggregated feature containing the information of the neighboring nodes. The aggregated feature reflects the position and relative importance of each knowledge module in the course knowledge graph. By globally pooling the feature vector of each node (such as average pooling or maximum pooling), the overall feature representation of the course graph, that is, the structural features of the course graph, can be obtained.

[0120] S340, using a recurrent neural network to extract sequence features of the course learning content in the course information package to obtain course sequence features.

[0121] For example, in addition to the graph structure features, the course content also has a sequential nature (such as the order of course chapters, the order of learning knowledge points, etc.), so the sequence features can be extracted through a recurrent neural network (RNN).

[0122] The sequence order of course learning content can be extracted from the course information package, for example, a content sequence is formed according to the order of chapters or knowledge points, and each sequence element represents a course content unit, such as chapters, videos, etc. The course content sequence is input into the RNN for sequence feature extraction. LSTM (Long Short-Term Memory Network) or GRU (Gated Recurrent Unit) can be used to extract course sequence features, and finally the recurrent neural network (RNN) will output the course sequence features.

[0123] Optionally, S340, extracting sequence features of the course learning content in the course information package using a recurrent neural network to obtain course sequence features, including: S341, extracting logical information of the course learning content from the course information package, wherein the logical information is used to represent the arrangement order of the course learning content.

[0124] It can be understood that the logical information can represent the arrangement order of the course learning content, reflecting the learning order or dependency relationship between knowledge modules.

[0125] For example, the chapters and contents of a course are organized in the order of learning, from elementary to advanced, from basic to advanced, so the order of the course chapters can be used as logical information; some courses will design recommended learning paths, indicating the optimal order for learners in the learning content, and the recommended path can be used to determine the logical information. The chapter order and recommended path in the course information package are extracted to form a logical information table for subsequent knowledge module sorting.

[0126] S342, sorting each knowledge module in the knowledge module collection according to the logic information to obtain a sorted knowledge module collection.

[0127] For example, an empty list can be created to arrange the knowledge modules in order, and the knowledge modules can be added to the sorted list one by one according to the order in the logic information. For example, if the logic information indicates that the order is basic knowledge->intermediate knowledge->advanced knowledge, the knowledge modules are arranged in this order. The collection of knowledge modules after sorting is a collection of knowledge modules arranged in a logical order.

[0128] S343, using a recurrent neural network to extract sequence feature vectors from each knowledge module in the sorted knowledge module collection to obtain course sequence features.

[0129] For example, the sorted knowledge module set can be input into the recurrent neural network, and each knowledge module can be represented as a feature vector, which can include information such as the module content summary and knowledge points. If the feature vector of the knowledge module is based on text information (such as knowledge point description), the text features can be converted into a dense feature vector through the embedding layer, so that the feature vector of the knowledge module can adapt to the input format of the RNN.

[0130] In the RNN layer, LSTM (Long Short-Term Memory Network) or GRU (Gated Recurrent Unit) is used to extract features from sequence data. LSTM and GRU can memorize long-distance dependent information in the sequence, which is very suitable for sequences of knowledge modules with dependencies. At each time step, RNN generates a hidden state based on the current knowledge module and the previous learning content, reflecting the characteristics of the current knowledge module in the sequence. RNN passes the information of the previous module to the next module, so that the output features contain the learning path and order information in the entire sequence.

[0131] After the RNN has processed the sequence, the last hidden state output by the RNN can be used as the sequence feature vector of the entire course, or the hidden state of each time step can be aggregated using a global pooling operation to form a complete course sequence feature representation.

[0132] S350, integrating the course graph structure features and the course sequence features to obtain learning content features.

[0133] For example, feature concatenation can be used to directly concatenate the course graph structure features and course sequence features into a complete vector; weighted fusion can be used. If a feature has a higher weight, the feature vector can be weighted and then fused. The fused feature vector represents the learning content features of the course, and contains the graph structure information and sequence information of the course content.

[0134] Optionally, S350, the course graph structure feature and the course sequence feature are integrated to obtain learning content features, including: S351, the structural feature vector of each knowledge module in the course graph structural feature and the sequence feature vector of each knowledge module in the course sequence feature are merged to obtain the comprehensive feature of each knowledge module.

[0135] Exemplarily, the structural feature vector and the course sequence feature vector of each knowledge module are directly spliced ​​into a larger feature vector to represent the comprehensive features of the knowledge module. For example, if the structural feature vector of a knowledge module is [0.5, 0.3, 0.7] and the sequence feature vector is [0.6, 0.2, 0.8], then the comprehensive features of the knowledge module after splicing are [0.5, 0.3, 0.7, 0.6, 0.2, 0.8].

[0136] In addition to the feature splicing method, the weighted fusion method can also be used. The weight values ​​of the course graph structural features and course sequence features can be set manually. According to the weights corresponding to the course graph structural features and the course sequence features, the structural feature vector and course sequence feature vector of each knowledge module are weighted to obtain the comprehensive features of each knowledge module. For example, the structural feature vector of a knowledge module is [0.5, 0.3, 0.7], the weight value corresponding to the course graph structural features is 0.4, the sequence feature vector is [0.6, 0.2, 0.8], and the weight value corresponding to the course sequence features is 0.6. Then, the comprehensive features of the knowledge module after weighted fusion are 0.4×[0.5, 0.3, 0.7]+0.6×[0.6, 0.2, 0.8]=[0.56, 0.24, 0.76].

[0137] S352, integrating the comprehensive features of each knowledge module to obtain learning content features.

[0138] For example, the comprehensive feature vectors of all knowledge modules can be average pooled to obtain the average feature representation of the entire course (learning content features). Average pooling can balance the information of all knowledge modules so that the features of each module have the same weight in the integration.

[0139] In some cases, different knowledge modules may have different importances, and the characteristics of each knowledge module can be aggregated in a weighted manner to obtain more representative learning content characteristics.

[0140] Through the above steps, the system is provided with a way to deeply represent learning content, which can provide a basis for the subsequent system to accurately generate personalized learning resources and learning paths that are most suitable for learners.

[0141] S400, determining the learner's adaptive learning content from the course learning content of the smart learning cloud platform according to the learner's course learning characteristics and learning content characteristics.

[0142] Exemplarily, the learner's course learning characteristics and learning content characteristics may be input into the prediction model, and the prediction model may output learning content characteristics that are adapted to the course learning characteristics.

[0143] Prediction model training process: The course learning data and corresponding course learning content of different learners can be collected from the smart learning cloud platform as sample data, and based on the learner's historical learning records, which content is suitable for the learner in the course learning content. The definition of suitable can be determined based on the learning effect (such as test scores, completion status). For example, if the learner performs well in a certain content or has a high interest score, it can be considered as learning content suitable for the learner. Extract features from the course learning data and corresponding course learning content of different learners to obtain course learning sample features and learning content sample features. The sample features can be divided into training sets and test sets. The training set is used for prediction model training, and the test set is used for the final performance evaluation of the prediction model.

[0144] You can choose a suitable model, for example, decision tree / random forest, which is suitable for relatively clear classification tasks and can capture the hierarchical relationship between learners' course learning characteristics and learning content characteristics; neural network (such as DNN), which is suitable for tasks with complex nonlinear relationships and can model complex feature relationships through multiple layers of neurons; support vector machine (SVM), which is suitable for situations where the relationship between features and labels is clear and there is a small amount of high-dimensional data, can well separate different learning content feature categories. Use the training set to iterate the prediction model multiple times, adjust the model weights or parameters, so that the model can capture the relationship between learners' course learning characteristics and learning content characteristics, and use the test set to evaluate the model through cross-validation or leave-one-out validation method to check the performance of the model on different sample sets, which is conducive to the generalization ability of the model.

[0145] After the prediction model is trained, the learner's course learning characteristics and learning content characteristics can be input into the prediction model. The prediction model calculates the degree of fit between the learner characteristics and each content characteristic, and outputs the content characteristics with the highest degree of match with the learner characteristics, which can be expressed as a score or category. For example, the prediction model may output suitable topics: data analysis; suitable content format: video; suitable difficulty: medium. For multi-dimensional outputs (such as topic, format, difficulty), you can sort them according to the fit score and select the most suitable feature combination.

[0146] According to the predicted learning content features that are most suitable for learners, specific learning content that matches the learning content features that are most suitable for learners is retrieved from the resource library of the smart learning cloud platform to generate the final personalized learning content. For example, all learning content related to the topic predicted by the prediction model, such as data analysis, can be screened out, and the content can be further filtered according to the predicted difficulty level (such as medium), which is conducive to the content difficulty being suitable for the current level of the learner, and the recommendation format can be selected according to the learner's preferences, such as recommending more video content or interactive exercises to increase the adaptability and experience of learning. If the generated learning content has a clear order of chapters and knowledge points, a learning path can be designed for learners, for example, learning the basic content first, and then gradually advancing to ensure the coherence and logic of the content.

[0147] This step uses a predictive model to analyze the relationship between learners' course learning characteristics and learning content characteristics, which can achieve accurate prediction of learning content characteristics and retrieve specific learning content that matches it from the smart learning cloud platform, providing learners with a highly adaptive learning experience.

[0148] In one possible implementation, see Figure 4 , S400, according to the learner's course learning characteristics and learning content characteristics, determine the learner's adaptive learning content from the course learning content of the smart learning cloud platform, including: S410, performing a matching operation on the learner's course learning characteristics and the comprehensive characteristics of each knowledge module in the learning content characteristics to obtain a matching degree of each knowledge module.

[0149] It can be understood that the matching degree can represent the similarity between the learner's course learning characteristics and the comprehensive characteristics of each knowledge module.

[0150] For example, the cosine similarity or Euclidean distance can be used to calculate the matching degree. Cosine similarity is a vector similarity measurement method suitable for high-dimensional feature vectors. The range of cosine similarity is 0 to 1. The closer the value is to 1, the higher the similarity. The calculation formula of cosine similarity is: ,in, Indicates The matching degree of knowledge modules, represents the learner’s course learning characteristics, Indicates The comprehensive characteristics of each knowledge module, and They are and The length of the vector, i.e. the norm.

[0151] Euclidean distance is suitable for measuring the distance between two feature vectors. The smaller the distance, the higher the similarity. In order to make the matching value consistent with the similarity, the reciprocal of the distance can be taken. The calculation formula based on Euclidean distance is: ,in, and The course learning characteristics of learners are and Comprehensive characteristics of knowledge modules The elements.

[0152] The selected similarity measurement method can be used to calculate the matching degree of each knowledge module one by one, and the calculated matching degree values ​​are stored in a matching degree list for subsequent screening.

[0153] S420, determining whether the matching degree of each knowledge module is greater than a preset threshold, and determining the learning content in the knowledge module with a matching degree greater than the preset threshold as the learner's adaptive learning content. The preset threshold is the minimum matching degree value.

[0154] It can be understood that the preset threshold can be used to filter out knowledge modules that are highly matched with the learner's course learning characteristics.

[0155] For example, a fixed matching threshold value can be set manually, such as 0.7 or 0.8 (if cosine similarity is used, the value is between 0 and 1), indicating that knowledge modules with a matching degree of at least 70% or 80% are selected. The threshold value can be set dynamically according to the data distribution, such as taking the average matching degree of all knowledge modules as the threshold value, or selecting the lowest acceptable matching degree.

[0156] For each knowledge module, if the matching degree If it is greater than or equal to the preset threshold, the matching degree The learning content in the knowledge module that is greater than or equal to a preset threshold is determined as adaptive learning content to form a personalized recommended content set.

[0157] Assume that the learner's course learning feature vector is , the comprehensive feature vectors of the knowledge module are: , , , the preset threshold is set to 0.7, and the cosine similarity can be used to calculate the matching degree. Matching degree 1 ( and ): ; Matching degree 2 ( and ): ; Matching degree 3 ( and ): The final matching results include matching degree 1 of 0.97, matching degree 2 of 0.57, and matching degree 3 of 0.99. Matching degree 1, matching degree 2, and matching degree 3 are compared with the preset threshold. If matching degree 1 and matching degree 3 are greater than the preset threshold, and matching degree 2 is less than the preset threshold, the learning content in knowledge module 1 and knowledge module 3 is determined as the learner's adaptive learning content.

[0158] By calculating the matching degree between the learner's course learning characteristics and the comprehensive characteristics of the knowledge modules, and screening out the knowledge modules that meet the threshold requirements, we can generate highly personalized adaptive learning content for learners, which is conducive to the learning content being in line with the learner's interests and learning level, and helping learners to learn efficiently on suitable knowledge modules.

[0159] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0160] Corresponding to the adaptive learning content generation method of the smart learning cloud platform described in the above embodiment, the embodiment of the present application also provides an adaptive learning content generation device of the smart learning cloud platform, and each unit of the device can implement each step of the adaptive learning content generation method of the smart learning cloud platform. Figure 5 A structural block diagram of an adaptive learning content generation device of a smart learning cloud platform provided in an embodiment of the present application is shown. For ease of explanation, only the parts related to the embodiment of the present application are shown.

[0161] Reference Figure 5 , the device comprises: The acquisition unit is used to acquire the course learning data of the learner. The course learning data of the learner is used to represent the learning data of the learner in a certain course; the course learning data of the learner includes learning interest data, course progress data, learning behavior data and test evaluation data.

[0162] The course learning feature obtaining unit is used to determine the course learning features of the learner according to the course learning data of the learner.

[0163] The learning content feature obtaining unit is used to extract features of the course learning content in the smart learning cloud platform to obtain learning content features.

[0164] The adaptive learning content determination unit is used to determine the learner's adaptive learning content from the course learning content of the smart learning cloud platform according to the learner's course learning characteristics and learning content characteristics.

[0165] It should be noted that the information interaction, execution process and other contents between the above-mentioned units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0166] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0167] The present application also provides an adaptive learning content generation device for a smart learning cloud platform. Figure 6 This is a schematic diagram of the structure of an adaptive learning content generation device of a smart learning cloud platform provided in an embodiment of the present application. Figure 6 As shown, the adaptive learning content generating device 6 of the smart learning cloud platform of this embodiment includes: at least one processor 60 ( Figure 6 Only one is shown), at least one memory 61 ( Figure 6 Only one is shown in the figure) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60. When the processor 60 executes the computer program 62, the adaptive learning content generation device 6 of the smart learning cloud platform implements the steps in any of the above-mentioned adaptive learning content generation method embodiments of the smart learning cloud platform, or the adaptive learning content generation device 6 of the smart learning cloud platform implements the functions of each unit in the above-mentioned device embodiments.

[0168] Exemplarily, the computer program 62 may be divided into one or more modules / units, which are stored in the memory 61 and executed by the processor 60 to complete the present application. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program 62 in the adaptive learning content generation device 6 of the smart learning cloud platform.

[0169] The adaptive learning content generation device 6 of the smart learning cloud platform can be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The adaptive learning content generation device of the smart learning cloud platform may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art can understand that Figure 6 It is only an example of the adaptive learning content generation device 6 of the smart learning cloud platform and does not constitute a limitation of the adaptive learning content generation device 6 of the smart learning cloud platform. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components, for example, it may also include input and output devices, network access devices, buses, etc.

[0170] The processor 60 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0171] In some embodiments, the memory 61 may be an internal storage unit of the adaptive learning content generating device 6 of the smart learning cloud platform, such as a hard disk or memory of the adaptive learning content generating device 6 of the smart learning cloud platform. In other embodiments, the memory 61 may also be an external storage device of the adaptive learning content generating device 6 of the smart learning cloud platform, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the adaptive learning content generating device 6 of the smart learning cloud platform. Further, the memory 61 may also include both an internal storage unit and an external storage device of the adaptive learning content generating device 6 of the smart learning cloud platform. The memory 61 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory 61 may also be used to temporarily store data that has been output or is to be output.

[0172] An embodiment of the present application further 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 steps in any of the above method embodiments are implemented.

[0173] An embodiment of the present application provides a computer program product. When the computer program product runs on an adaptive learning content generation device of a smart learning cloud platform, the adaptive learning content generation device of the smart learning cloud platform implements the steps in any of the above-mentioned method embodiments.

[0174] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device, recording medium, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium that can carry computer program code to the adaptive learning content generation device of the smart learning cloud platform. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electric carrier signals and telecommunication signals.

[0175] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0176] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0177] In the embodiments provided in the present application, it should be understood that the disclosed adaptive learning content generation device, equipment and method of the intelligent learning cloud platform can be implemented in other ways. For example, the adaptive learning content generation device and equipment embodiments of the intelligent learning cloud platform described above are merely schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0178] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0179] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for generating adaptive learning content for a smart learning cloud platform. It is characterized in that include: Acquire the course learning data of the learner; wherein the course learning data of the learner is used to represent the learning data of the learner in a certain course; the course learning data of the learner includes learning interest data, course progress data, learning behavior data and test evaluation data; Determining the course learning characteristics of the learner according to the course learning data of the learner; Extract features of course learning content in the smart learning cloud platform to obtain learning content features; According to the course learning characteristics of the learner and the learning content characteristics, the learner's adaptive learning content is determined from the course learning content of the smart learning cloud platform.

2. The method for generating adaptive learning content for a smart learning cloud platform according to claim 1, It is characterized in that The determining the course learning characteristics of the learner according to the course learning data of the learner includes: Performing feature extraction on the learning interest data to obtain a first feature; Numerically processing the course progress data to obtain a second feature; Performing feature extraction on the learning behavior data and the test evaluation data to obtain a third feature; The first feature, the second feature and the third feature are fused to obtain the course learning feature of the learner.

3. The method for generating adaptive learning content for a smart learning cloud platform according to claim 2, It is characterized in that The extracting features of the learning behavior data and the test evaluation data to obtain a third feature includes: constructing a learning timing diagram according to the learning behavior data and the test evaluation data; The third feature is obtained by extracting features from the learning timing diagram using a first graph neural network; wherein the first graph neural network is trained based on learning behavior data and test evaluation data of different learners.

4. The method for generating adaptive learning content for a smart learning cloud platform according to claim 3, It is characterized in that The step of constructing a learning timing diagram according to the learning behavior data and the test evaluation data includes: Extracting time information of the learning behavior data and time information of the test evaluation data; Based on the learning behavior data, determining a learning behavior node of the learning timing diagram; Based on the test evaluation data, determining a test evaluation node of the learning timing diagram; Based on the time information of the learning behavior data, the learning behavior nodes of the learning timing diagram, the time information of the test evaluation data, and the test evaluation nodes of the learning timing diagram, the time sequence edges of the learning timing diagram are determined; wherein the time sequence edges of the learning timing diagram are used to represent the sequence between different learning behaviors of the learner and the test evaluation results; Determining edge attributes of the learning timing graph based on the time information of the learning behavior data and the time information of the test evaluation data; The learning timing graph is constructed according to the learning behavior nodes of the learning timing graph, the test evaluation nodes of the learning timing graph, the time sequence edges of the learning timing graph and the edge attributes of the learning timing graph.

5. The method for generating adaptive learning content for a smart learning cloud platform according to claim 2, It is characterized in that The step of fusing the first feature, the second feature, and the third feature to obtain the course learning feature of the learner includes: Obtain a first weight, a second weight, and a third weight; wherein the first weight is used to characterize the weight value corresponding to the first feature, the second weight is used to characterize the weight value corresponding to the second feature, and the third weight is used to characterize the weight value corresponding to the third feature; Calculate a weighted sum of the first feature, the second feature, and the third feature according to the first weight, the second weight, and the third weight to obtain a fused feature vector; The fused feature vector is determined as the course learning feature of the learner.

6. The method for generating adaptive learning content for a smart learning cloud platform according to claim 1, It is characterized in that The feature extraction of the course learning content in the smart learning cloud platform to obtain the learning content features includes: According to the course learning data of the learner, obtaining the course type information of the learner; Acquire course learning content matching the course type information of the learner from the learning resource library of the smart learning cloud platform to obtain a course information package; Using a second graph neural network to extract graph structure features of the course learning content in the course information package to obtain course graph structure features; Using a recurrent neural network to extract sequence features of the course learning content in the course information package to obtain course sequence features; The course map structure feature and the course sequence feature are integrated to obtain the learning content feature.

7. The method for generating adaptive learning content for a smart learning cloud platform according to claim 6, It is characterized in that The step of extracting graph structure features of the course learning content in the course information package using the second graph neural network to obtain the course graph structure features includes: Segment the course learning content in the course information package into knowledge modules to obtain a knowledge module collection, and determine each knowledge module in the knowledge module collection as a node of a course knowledge graph; wherein the course knowledge graph is used to represent the relationship between each knowledge module; Determine the edge of the course knowledge graph according to the relationship information of each knowledge module in the knowledge module collection and the node of the course knowledge graph; Constructing the course knowledge graph based on the nodes of the course knowledge graph and the edges of the course knowledge graph; The second graph neural network is used to calculate the structural feature vector of each node in the course knowledge graph to obtain the structural features of the course graph.

8. The method for generating adaptive learning content for a smart learning cloud platform according to claim 7, It is characterized in that The process of extracting sequence features of the course learning content in the course information package using a recurrent neural network to obtain course sequence features includes: Extracting the logical information of the course learning content from the course information package; wherein the logical information is used to represent the arrangement order of the course learning content; Sorting each knowledge module in the knowledge module collection according to the logic information to obtain a sorted knowledge module collection; The recurrent neural network is used to extract sequence feature vectors from each knowledge module in the sorted knowledge module collection to obtain the course sequence features.

9. The method for generating adaptive learning content for a smart learning cloud platform according to claim 8, It is characterized in that The step of fusing the course graph structure feature and the course sequence feature to obtain the learning content feature includes: The structural feature vector of each knowledge module in the course graph structural feature and the sequence feature vector of each knowledge module in the course sequence feature are merged to obtain the comprehensive feature of each knowledge module; The comprehensive features of each knowledge module are integrated to obtain the learning content features.

10. The method for generating adaptive learning content for a smart learning cloud platform according to claim 9, It is characterized in that The step of determining the learner's adaptive learning content from the course learning content of the smart learning cloud platform according to the learner's course learning characteristics and the learning content characteristics includes: Performing a matching operation on the learner's course learning characteristics and the comprehensive characteristics of each knowledge module in the learning content characteristics to obtain a matching degree of each knowledge module; It is determined whether the matching degree of each knowledge module is greater than a preset threshold, and the learning content in the knowledge module with a matching degree greater than the preset threshold is determined as the adaptive learning content of the learner; wherein the preset threshold is the minimum matching degree value.

Images