A note generation method based on an intelligent learning terminal

Abstract

The present disclosure provides a note generation method based on an intelligent learning terminal. A dependency graph is constructed by presetting the learning relationship of the first knowledge points, and based on the answer samples of the knowledge points and the dependency graph, an answer count matrix and a dependency error matrix are constructed to correct the errors existing in the dependency graph, improving the accuracy of determining the association relationship between the first knowledge points. Furthermore, according to the answer conditions of the answer samples of the knowledge points, a set of weak knowledge points corresponding to the answer samples of the knowledge points is determined, and thus the set of weak knowledge points is matched with the corrected dependency graph, and personalized learning notes corresponding to the answer samples of the knowledge points are generated by using the preset learning materials, improving the flexibility and accuracy of learning note generation.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, and in particular, to a note generation method based on an intelligent learning terminal. Background Art

[0002] In recent years, with the continuous advancement and development of computer technologies and educational informatization, computer and artificial intelligence technologies have gradually been applied to various daily educational teaching activities. More and more intelligent devices have been gradually applied to teaching scenarios, and students can learn through intelligent devices to provide more convenient learning services for students. During the process of students using intelligent devices for learning, the intelligent devices can provide the function of learning notes for students.

[0003] Currently, usually only preset learning notes can be provided for students according to the chapters of textbooks, the types of textbooks, etc., which may lead to the learning notes provided not being adaptable to the needs of students. Or, by adopting the method of testing students, the knowledge points that students have a poor grasp of are determined, and then learning notes are provided for students. However, only providing learning notes for the knowledge points that students have a poor grasp of may lead to students only being able to master individual knowledge points and unable to achieve the serial learning of knowledge points. For the solution that can provide learning notes for the serial connection of multiple knowledge points, usually the serial connection of knowledge points is carried out in a preset manner, and the serial connection relationship cannot be updated in a timely manner, and personalized learning notes cannot be provided for students. Summary of the Invention

[0004] The present disclosure provides a note generation method based on an intelligent learning terminal.

[0005] According to a first aspect of the present disclosure, there is provided a note generation method based on an intelligent learning terminal, the method including: based on a preset first knowledge point learning relationship, using an attention network model to extract the associated features of a first knowledge point to construct a dependency graph of the first knowledge point, where the first knowledge point includes a pre-first knowledge point and a dependent first knowledge point; in response to a dependency graph update instruction, obtaining a knowledge point answering sample to determine the average dependency error of the first knowledge point in the dependency graph and the answering count matrix of the first knowledge point; based on the average dependency error and the answering situation of the knowledge point answering sample, determining a dependency error matrix of the dependent first knowledge point; based on the answering count matrix and the dependency error matrix, determining the error of the preset first knowledge point learning relationship to update the dependency graph; based on the set of first knowledge points corresponding to a single knowledge point answering sample, updating the preset mastery probability of the set of first knowledge points to determine a set of weak knowledge points corresponding to the knowledge point answering sample; based on the set of weak knowledge points and the updated dependency graph, using preset learning materials to generate learning notes.

[0006] In some embodiments of the present disclosure, based on a preset learning relationship of the first knowledge point, an attention network model is used to extract the associated features of the first knowledge point to construct a dependency graph of the first knowledge point, including: based on the preset learning relationship of the first knowledge point, extracting the associated feature vector of the first knowledge point; based on the associated feature vector, using the triple method to construct an initial relationship graph of the first knowledge point; vectorizing the initial relationship graph to obtain the vectorized encoding of the first knowledge point corresponding to adjacent nodes; based on the vectorized encoding, determining the implicit features between the first knowledge points; based on the implicit features, updating the initial relationship graph to construct a dependency graph.

[0007] In some embodiments of the present disclosure, obtaining a knowledge point answering sample to determine the average dependency error of the first knowledge point and the answering count matrix of the first knowledge point in the dependency graph includes: based on the dependency graph, determining the first correct answering sample of the pre - set first knowledge point corresponding to the knowledge point answering sample, the first wrong answering sample depending on the first knowledge point, the second correct answering sample depending on the first knowledge point, and the number of the first knowledge points depending on the pre - set first knowledge point; based on the first correct answering sample, the second correct answering sample and the first wrong answering sample, determining the abnormal samples and the answering count matrix in the knowledge point answering sample; based on the set of dependent knowledge points corresponding to the knowledge point answering sample, the second correct answering sample, the abnormal samples and the number of the first knowledge points depending on the pre - set first knowledge point, determining the average dependency error, and the set of dependent knowledge points is composed of the pre - set first knowledge point and the first knowledge point depending on it.

[0008] In some embodiments of the present disclosure, determining the abnormal samples and the answering count matrix in the knowledge point answering sample based on the first correct answering sample and the second correct answering sample includes: based on the first correct answering sample and the first wrong answering sample, when it is determined that the answering of the pre - set first knowledge point corresponding to the same sample source is wrong and the answering depending on the first knowledge point is correct, determining the first correct answering sample corresponding to the same sample source as an abnormal sample; based on the number of the first correct answering samples, the number of the second correct answering samples, and the number of the abnormal samples, determining the answering count matrix.

[0009] In some embodiments of the present disclosure, determining the average dependency error based on the set of dependent knowledge points corresponding to the knowledge point answering sample, the second correct answering sample, the abnormal samples and the number of the first knowledge points depending on the pre - set first knowledge point includes: based on the set of dependent knowledge points corresponding to the knowledge point answering sample, the second correct answering sample, the abnormal samples and the number of the first knowledge points depending on the pre - set first knowledge point, combining the following formula 1 to determine the average dependency error:

[0010] Formula 1,

[0011] Wherein, M represents the average dependence error, N(E) represents the number of dependence on the first knowledge points corresponding to the first knowledge points in front, i represents the first knowledge points in front, j represents the dependence on the first knowledge points, and E represents the set of dependence knowledge points. represents the number of abnormal samples. represents the number of second correct answer samples.

[0012] In some embodiments of the present disclosure, based on the average dependence error and the answer conditions of the knowledge point answer samples, determining the dependence error matrix of the dependence on the first knowledge points includes: based on the knowledge point answer samples, determining the number of first correct answer samples, the number of second correct answer samples, and the total number of knowledge point answer samples; based on the number of first correct answer samples, the number of second correct answer samples, the total number, and the average dependence error, and combining the following formula 2, determining the dependence error matrix:

[0013] Formula 2

[0014] Wherein represents the dependence error matrix represents the number of first correct answer samples, and n represents the total number.

[0015] In some embodiments of the present disclosure, based on the answer counting matrix and the dependence error matrix, determining the error of the preset first knowledge point learning relationship to update the dependence relationship graph includes: based on the answer counting matrix and the dependence error matrix, and combining the following formula 3, determining the dependence error of the set of dependence knowledge points includes:

[0016] Formula 3

[0017] Wherein, Diff represents the dependence error of the set of dependence knowledge points, m represents the diagonal elements in the answer counting matrix and the dependence error matrix, represents the answer counting matrix;

[0018] When the dependence error is greater than the preset error threshold, it is determined that there is an error in the dependence relationship graph for the first knowledge points included in the set of dependence knowledge points; based on the first knowledge points included in the set of dependence knowledge points, using the attention network model, the dependence relationship graph is updated.

[0019] In some embodiments of the present disclosure, based on the first knowledge point set corresponding to the single knowledge point answer sample, the preset mastery probability of the first knowledge point set is updated to determine the weak knowledge point set corresponding to the knowledge point answer sample, including: determining the correct answer probability of the first knowledge point set based on the knowledge point answer sample; updating the preset mastery probability of the first knowledge point set based on the first knowledge point set corresponding to the single knowledge point answer sample and the correct answer probability to obtain the sample mastery probability; determining the weak knowledge point set of the knowledge point answer sample based on the sample mastery probability and the preset knowledge point mastery threshold.

[0020] In some embodiments of the present disclosure, updating the preset mastery probability of the first knowledge point set based on the first knowledge point set corresponding to the single knowledge point answer sample and the correct answer probability to obtain the sample mastery probability includes: determining the update weight of the preset mastery probability based on the correct answer probability; determining the updated mastery probability based on the answer situation of the knowledge point answer sample, the first knowledge point set corresponding to the single knowledge point answer sample, the preset mastery probability, and the update weight; performing normalization processing on the updated mastery probability to obtain the sample mastery probability.

[0021] In some embodiments of the present disclosure, determining the weak knowledge point set of the knowledge point answer sample based on the sample mastery probability and the preset knowledge point mastery threshold includes: determining the comprehensive mastery probability of the first knowledge point in the knowledge point answer sample based on the sample mastery probability and the preset knowledge point weight, in combination with the following formula 4:

[0022] Formula 4,

[0023] where, , represents the comprehensive mastery probability, represents the preset knowledge point weight, represents the first knowledge point, represents the first knowledge point set, represents the sample mastery probability; comparing the comprehensive mastery probability with the preset knowledge point mastery threshold to determine the weak knowledge point set.

[0024] In some embodiments of the present disclosure, based on the weak knowledge point set and the updated dependency graph, using the preset learning materials to generate learning notes includes: matching the first knowledge point in the weak knowledge point set with the updated dependency graph to determine the knowledge points required to be included in the learning notes; extracting the corresponding knowledge content from the preset learning materials based on the knowledge points required to be included in the learning notes to generate the learning notes.

[0025] In summary, a note generation method based on an intelligent learning terminal proposed by the present disclosure includes: based on a preset first knowledge point learning relationship, using an attention network model to extract the associated features of the first knowledge point to construct a dependency graph of the first knowledge point, where the first knowledge point includes a pre-first knowledge point and a dependent first knowledge point; in response to a dependency graph update instruction, obtaining a knowledge point answering sample to determine the average dependency error of the first knowledge point in the dependency graph and the answering count matrix of the first knowledge point; based on the average dependency error and the answering situation of the knowledge point answering sample, determining a dependency error matrix of the dependent first knowledge point; based on the answering count matrix and the dependency error matrix, determining the error of the preset first knowledge point learning relationship to update the dependency graph; based on the set of first knowledge points corresponding to a single knowledge point answering sample, updating the preset mastery probability of the set of first knowledge points to determine a set of weak knowledge points corresponding to the knowledge point answering sample; based on the set of weak knowledge points and the updated dependency graph, using preset learning materials to generate learning notes. The method of the present disclosure constructs a dependency graph through a preset first knowledge point learning relationship, and based on the knowledge point answering sample and the dependency graph, constructs an answering count matrix and a dependency error matrix to correct the errors existing in the dependency graph, improving the accuracy of determining the association relationship between the first knowledge points; and then, according to the answering situation of the knowledge point answering sample, determining a set of weak knowledge points corresponding to the knowledge point answering sample, thereby matching the set of weak knowledge points with the corrected dependency graph, and using preset learning materials to generate personalized learning notes corresponding to the knowledge point answering sample, improving the flexibility and accuracy of learning note generation.

[0026] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The drawings are used to better understand the solution and do not constitute a limitation to the present disclosure. Among them:

[0028] Figure 1 is a schematic flowchart of a note generation method based on an intelligent learning terminal provided by an embodiment of the present disclosure;

[0029] Figure 2 is a schematic flowchart of a second note generation method based on an intelligent learning terminal provided by an embodiment of the present disclosure;

[0030] Figure 3 is a schematic flowchart of a third note generation method based on an intelligent learning terminal provided by an embodiment of the present disclosure;

[0031] Figure 4Schematic flowchart of the fourth note generation method based on an intelligent learning terminal provided by an embodiment of the present disclosure;

[0032] Figure 5 Schematic flowchart of the fifth note generation method based on an intelligent learning terminal provided by an embodiment of the present disclosure;

[0033] Figure 6 Schematic structural diagram of a note generation device based on an intelligent learning terminal provided by an embodiment of the present disclosure;

[0034] Figure 7 Schematic hardware structure diagram of an electronic device provided by an embodiment of the present disclosure. Detailed implementation manners

[0035] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to assist in understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted below.

[0036] In recent years, with the continuous advancement and development of computer technology and educational informatization, computer and artificial intelligence technologies have gradually been applied to various daily educational and teaching activities. More and more intelligent devices have been gradually applied to teaching scenarios, and students can use intelligent devices for learning to provide more convenient learning services for students. During the process of students using intelligent devices for learning, the intelligent devices can provide the function of learning notes for students.

[0037] Currently, usually only preset learning notes can be provided for students according to the chapters of textbooks, types of textbooks, etc., which may lead to the learning notes provided not being adaptable to the needs of students. Or, by adopting the method of testing students, the knowledge points that students master poorly are determined, and then learning notes are provided for students; however, only providing learning notes for the knowledge points that students master poorly may lead to students only being able to master individual knowledge points and unable to achieve the serial learning between knowledge points; for the solution that can provide learning notes with multiple knowledge points in series, usually the preset method is used to connect the knowledge points in series, and the serial relationship cannot be updated in a timely manner, and personalized learning notes cannot be provided for students.

[0038] To solve the problems in the related art, a note generation method based on an intelligent learning terminal proposed by the present disclosure constructs a dependency graph through a preset first knowledge point learning relationship, and constructs an answer count matrix and a dependency error matrix based on the knowledge point answer samples and the dependency graph to correct the errors existing in the dependency graph, improving the accuracy of determining the association relationship between the first knowledge points; and then determines a set of weak knowledge points corresponding to the knowledge point answer samples according to the answer conditions of the knowledge point answer samples, thereby matching the set of weak knowledge points with the corrected dependency graph, and using the preset learning materials to generate personalized learning notes corresponding to the knowledge point answer samples, improving the flexibility and accuracy of learning note generation.

[0039] The following describes a note generation method based on an intelligent learning terminal according to an embodiment of the present disclosure with reference to the accompanying drawings.

[0040] Figure 1 It is a flowchart of a note generation method based on an intelligent learning terminal provided by an embodiment of the present disclosure. As Figure 1 shown, this method can be applied to an intelligent learning terminal, including:

[0041] Step 101: Based on a preset first knowledge point learning relationship, use an attention network model to extract the association features of the first knowledge points to construct a dependency graph of the first knowledge points.

[0042] In some embodiments, the intelligent learning terminal can be implemented based on a hardware device (such as a tablet computer, an intelligent whiteboard), or can be a software application installed on a general computer or a mobile device. Usually, these intelligent learning terminals can be connected to a cloud service to implement functions such as accessing rich educational resources and data synchronization. In other words, the intelligent learning terminal can record the process of the first student doing questions, learning, and reviewing during the first student's question-solving and learning process, so as to provide personalized learning notes for the user.

[0043] Optionally, the intelligent learning terminal can also synchronously display the operations of the teacher, knowledge point annotations (such as on the intelligent learning terminal), explanation content, dynamic demonstrations, etc. for the students.

[0044] In some embodiments, the first knowledge point can be any knowledge point in any subject, and the present disclosure does not limit this, as shown in Table 1 for example:

[0045] Table 1

[0046]

[0047] In some embodiments, the preset first knowledge point learning relationship may be the logical relationship and learning order between predefined first knowledge points. For example, the first knowledge point 1 is a prerequisite for the first knowledge point 2; the first knowledge point 3 needs to be mastered before understanding the first knowledge point 4.

[0048] In some embodiments, through the attention network model, the correlation features between the preset first knowledge point learning relationships are extracted to construct a dependency graph, where each node in the dependency graph is a first knowledge point, and the edge between each node represents the existence of a dependency relationship between two first knowledge points.

[0049] Furthermore, according to the preset first knowledge point learning relationship, the first knowledge points can be divided into prerequisite first knowledge points and dependent first knowledge points. Among them, the prerequisite first knowledge point is the prerequisite learning condition for the dependent first knowledge point, that is, the learning order is to learn the prerequisite first knowledge point first and then the dependent first knowledge point. For example: the trigonometric function knowledge point can be a prerequisite first knowledge point, and the inverse trigonometric function knowledge point can be a dependent first knowledge point of the trigonometric function knowledge point.

[0050] In some embodiments, a prerequisite first knowledge point can correspond to one or more dependent first knowledge points, and the present disclosure does not limit this.

[0051] In some embodiments, the relationship between the prerequisite first knowledge point and the dependent first knowledge point is relative, that is, a first knowledge point can be both a prerequisite first knowledge point and a dependent first knowledge point. For example, the trigonometric function knowledge point can be a prerequisite first knowledge point for the inverse trigonometric function knowledge point, and can also be a dependent first knowledge point of the four arithmetic operations knowledge point.

[0052] In some embodiments, the specific model of the attention network model is not limited, and it is, for example, a Neighborhood Calibration Graph Attention Network (NCGAT) and other models.

[0053] Specifically, based on the preset first knowledge point learning relationship, the correlation feature vector of the first knowledge point can be extracted; based on the correlation feature vector, using the triple method, the initial relationship graph of the first knowledge point can be constructed; the initial relationship graph is vectorized to obtain the vectorized encoding of the first knowledge point corresponding to the adjacent nodes; based on the vectorized encoding, the implicit features between the first knowledge points are determined; based on the implicit features, the initial relationship graph is updated to construct a dependency graph.

[0054] Exemplarily, the preset first knowledge point learning relationship can be initialized and embedded, and each first knowledge point is initialized as an embedding vector, and the dimension of the embedding vector is 64 (but not limited to this, it can also be other dimensions);

[0055] Use NCGAT to capture the features and relationships of the first knowledge points, so that the representation of each first knowledge point can reflect its position and importance in the entire knowledge system;

[0056] Use the NCGAT model to calculate the embedded representation of each first knowledge point, where these embedded representations capture the relationships and dependencies between knowledge points.

[0057] Furthermore, the specific structure and execution process of the NCGAT model are as follows:

[0058] Input layer: Extract knowledge points from the preset relationships of the first knowledge points (such as teaching objectives, curriculum outlines, chapter contents, summaries and reviews, and expert teaching knowledge bases, etc.), and filter out the corresponding first knowledge points and associated relationship features as the model input.

[0059] Graph construction layer: Based on the knowledge points and their associated features extracted above, use the Subject-Action-Object (SAO) triple method to construct the knowledge point network association graph structure. The SAO triple is used to construct the initial knowledge point network structure for knowledge points and associated features. The triple can be composed of levels such as basic, advanced, and difficult. Among them, the triple classification of the first knowledge point is based on examining the set of knowledge points of the same type, and its complexity and difficulty increase with the change of levels. Furthermore, it can be used to extract the implicit dependence relationship between the first knowledge points.

[0060] Embedding layer: Perform vectorization conversion on the knowledge point association network graph constructed based on the SAO triple. Among them, the vectorization encoding is generated sequentially from basic, advanced, and difficult according to the hierarchical order of the SAO triple architecture, and the vectorization encodings of multiple single knowledge point sets are obtained.

[0061] Graph Attention Network Layer: Using the graph attention network, a hidden masked self-attention layer is used to capture the importance of neighboring nodes in graph-structured data. For example, in a single knowledge point set, there may be a situation where knowledge point units are combined, and the same examination point may involve the combined examination of multiple knowledge points, that is, there is a situation of intersection of knowledge point sets. This belongs to the implicit feature of the knowledge point association network. Each layer of the graph attention network serves as a feature extractor, aggregates information from adjacent nodes, and generates a new node feature representation by learning the convolutional kernel weights. Furthermore, the attention mechanism is used to assign a unique attention score to each adjacent node to identify more important neighboring nodes. The knowledge point association network is updated through the graph attention network to generate the above-mentioned dependency graph. Since the graph attention network is a local model that allows exploration of adjacent nodes without accessing the entire graph, it can ensure that the implicit knowledge point association feature is incorporated from a single knowledge point set without damaging the initial knowledge point association network graph, thereby improving the accuracy of the graph attention network construction.

[0062] Step 102: In response to the dependency graph update instruction, obtain the knowledge point answering samples to determine the average dependency error of the first knowledge point in the dependency graph and the answering count matrix of the first knowledge point.

[0063] In some embodiments, when the user is ready to generate a learning note, a dependency graph update instruction can be sent to the intelligent learning terminal to update the dependency graph and improve the accuracy of the generated learning note.

[0064] In some embodiments, the intelligent learning terminal, in response to the dependency graph update instruction, obtains the knowledge point answering samples, where the knowledge point answering samples can be knowledge point answering samples in units of classes, study groups, specific student groups, etc. The scope of the knowledge point answering samples is not limited in this disclosure.

[0065] In some instances, based on the dependency graph, the first correct answering sample of the pre-posed first knowledge point corresponding to the knowledge point answering sample, the first incorrect answering sample depending on the first knowledge point, the second correct answering sample depending on the first knowledge point, and the number of first knowledge points depending on the pre-posed first knowledge point can be determined; then, based on the first correct answering sample, the second correct answering sample, and the first incorrect answering sample, the abnormal samples and the answering count matrix in the knowledge point answering sample can be determined; and based on the dependency knowledge point set corresponding to the knowledge point answering sample, the second correct answering sample, the abnormal samples, and the number of first knowledge points depending on the pre-posed first knowledge point, the average dependency error can be determined. For specific details, please refer to Figure 3 the embodiments shown, which will not be elaborated here.

[0066] In some embodiments, the average dependence error is used to indicate the error of the association relationship between different first knowledge points in the dependence relationship graph.

[0067] In some embodiments, the response count matrix is used to indicate the number of correct responses for combinations between different first knowledge points in the large sample of the knowledge points.

[0068] Step 103: Determine the dependence error matrix that depends on the first knowledge point based on the average dependence error and the response situation of the knowledge point response sample.

[0069] In some instances, based on the knowledge point response sample, the number of first correct response samples, the number of second correct response samples, and the total number of knowledge point response samples can be determined; and then, based on the number of first correct response samples, the number of second correct response samples, the total number, and the average dependence error, the dependence error matrix is determined. For specific reference, see Figure 4 the embodiments shown, which will not be elaborated here.

[0070] In some embodiments, the dependence error matrix can be used to indicate the number of errors between the pre - first knowledge point and the dependence first knowledge point in the knowledge point response sample.

[0071] Step 104: Determine the error of the preset first knowledge point learning relationship based on the response count matrix and the dependence error matrix to update the dependence relationship graph.

[0072] In some embodiments, based on the response count matrix and the dependence error matrix, the error value of the dependence knowledge point set composed of the pre - first knowledge point and the dependence first knowledge point can be determined. Thus, when the dependence error is greater than the preset error threshold, it is determined that there is an error in the first knowledge point included in the dependence knowledge point set in the dependence relationship graph; and based on the first knowledge point included in the dependence knowledge point set, the dependence relationship graph is updated using the attention network model.

[0073] Step 105: Update the preset mastery probability of the first knowledge point set based on the first knowledge point set corresponding to a single knowledge point response sample to determine the weak knowledge point set corresponding to the knowledge point response sample.

[0074] In some embodiments, based on the knowledge point response sample, the correct response probability of the first knowledge point set can be determined; based on the first knowledge point set corresponding to a single knowledge point response sample and the correct response probability, the preset mastery probability of the first knowledge point set is updated to obtain the sample mastery probability; based on the sample mastery probability and the preset knowledge point mastery threshold, the weak knowledge point set of the knowledge point response sample is determined.

[0075] In other words, the probability of correct answers in the knowledge point answer samples is used to determine the first knowledge points with relatively poor mastery of the corresponding first knowledge points in the knowledge point answer samples, so as to determine the set of weak knowledge points. For specific examples, please refer to Figure 6 the embodiments shown, which will not be elaborated here.

[0076] Step 106: Based on the set of weak knowledge points and the updated dependency graph, generate study notes using the preset learning materials.

[0077] In some embodiments, the first knowledge points in the set of weak knowledge points can be matched with the updated dependency graph to determine the knowledge points that the study notes need to include; then, based on the knowledge points that the study notes need to include, the corresponding knowledge content is extracted from the preset learning materials to generate study notes.

[0078] For example, the preset learning materials can be stored in the intelligent learning terminal, and the corresponding knowledge points in the preset learning materials are extracted through technologies such as semantic analysis and clustering analysis. Based on the above-determined set of weak knowledge points and the updated dependency graph for matching, finally, study notes suitable for the user corresponding to the knowledge point answer samples are automatically generated on the intelligent learning terminal device, such as key knowledge point annotations and material links under the knowledge point annotations, so as to realize the generation of personalized notes and adapt to different users' learning strategy plans.

[0079] In summary, the present disclosure proposes a note generation method based on an intelligent learning terminal, including: based on a preset first knowledge point learning relationship, using an attention network model to extract the associated features of the first knowledge point to construct a dependency graph of the first knowledge point, where the first knowledge point includes a pre - first knowledge point and a dependent first knowledge point; in response to a dependency graph update instruction, obtaining a knowledge point answer sample to determine the average dependency error of the first knowledge point in the dependency graph and the answer count matrix of the first knowledge point; based on the average dependency error and the answer situation of the knowledge point answer sample, determining a dependency error matrix of the dependent first knowledge point; based on the answer count matrix and the dependency error matrix, determining the error of the preset first knowledge point learning relationship to update the dependency graph; based on the set of first knowledge points corresponding to a single knowledge point answer sample, updating the preset mastery probability of the set of first knowledge points to determine a set of weak knowledge points corresponding to the knowledge point answer sample; based on the set of weak knowledge points and the updated dependency graph, using preset learning materials to generate learning notes. The method of the present disclosure constructs a dependency graph through a preset first knowledge point learning relationship, and based on the knowledge point answer sample and the dependency graph, constructs an answer count matrix and a dependency error matrix to correct the errors existing in the dependency graph, improving the accuracy of determining the association relationship between the first knowledge points; and then, according to the answer situation of the knowledge point answer sample, determining a set of weak knowledge points corresponding to the knowledge point answer sample, thereby matching the set of weak knowledge points with the corrected dependency graph, and using preset learning materials to generate personalized learning notes corresponding to the knowledge point answer sample, improving the flexibility and accuracy of learning note generation.

[0080] As a possible implementation, as Figure 2 shown in the schematic flowchart of the second note generation method based on an intelligent learning terminal, on the basis of the above - mentioned embodiments, step 102 is further explained and includes the following steps:

[0081] Step 201: Based on the dependency graph, determine the first correct answer sample of the pre - first knowledge point corresponding to the knowledge point answer sample, the first wrong answer sample of the dependent first knowledge point, the second correct answer sample of the dependent first knowledge point, and the number of dependent first knowledge points of the pre - first knowledge point.

[0082] In some embodiments, the first knowledge points included in the knowledge point answer sample can be determined according to the association relationship between the first knowledge points in the dependency graph, and the pre - first knowledge point and the dependent first knowledge point can be divided according to the included first knowledge points.

[0083] For example, if the knowledge point answer sample includes three first knowledge points V1, V2, and V3, then according to the dependency graph, it can be determined that V1 is the pre - first knowledge point of V2 and V3, and V2 can be the pre - first knowledge point of V3.

[0084] Step 202: Based on the first correct answer sample, the second correct answer sample, and the first incorrect answer sample, determine the abnormal samples and the answer count matrix in the knowledge point answer samples.

[0085] In some embodiments, based on the first correct answer sample and the first incorrect answer sample, when it is determined that the pre - first knowledge point corresponding to the same sample source is answered incorrectly and the answer to the first knowledge point is correct, the first correct answer sample corresponding to the same sample source is determined as an abnormal sample. Furthermore, based on the number of the first correct answer samples, the number of the second correct answer samples, and the number of abnormal samples, the answer count matrix is determined.

[0086] For example, if in the knowledge point answer samples, user 1 answers the knowledge point V1 incorrectly but answers the knowledge point V2 correctly, then the answer samples of user 1 for the knowledge point V1 and the knowledge point V2 can be determined as abnormal samples.

[0087] For example, the first correct answer samples, the second correct answer samples, the first incorrect answer samples, and the second incorrect answer samples depending on the first knowledge point can be statistically counted in the following table format, where 1 represents a correct answer and 0 represents an incorrect answer:

[0088] Table 2

[0089]

[0090] Furthermore, as shown in Table 2, construct the answer count matrix, where the answer count matrix can be a matrix, is the number of the first knowledge points, and each element in the answer count matrix represents the number of cases where the knowledge point is answered incorrectly but the knowledge point is answered correctly. Among them, the values of each can be statistically counted according to Table 2, and the values of

[0091] can be as shown in Table 3:

[0092]

[0093] It should be understood that the pre - first knowledge point and the knowledge point depending on the first knowledge point can cover all the first knowledge points.

[0094] It should be understood that and may have a dependency relationship or may not have a dependency relationship, and the present disclosure does not limit this.

[0095] Step 203: Determine the average dependence error based on the set of dependent knowledge points corresponding to the knowledge point answer samples, the second correct answer samples, the abnormal samples, and the number of dependent first knowledge points of the preceding first knowledge point.

[0096] In some embodiments, the set of dependent knowledge points consists of the preceding first knowledge point and the dependent first knowledge point.

[0097] In some embodiments, based on the set of dependent knowledge points corresponding to the knowledge point answer samples, the second correct answer samples, the abnormal samples, and the number of dependent first knowledge points of the preceding first knowledge point, the average dependence error can be determined by combining the following formula 1:

[0098] Formula 1

[0099] where M represents the average dependence error, N(E) represents the number of dependent first knowledge points corresponding to the preceding first knowledge point, i represents the preceding first knowledge point, j represents the dependent first knowledge point, and E represents the set of dependent knowledge points. represents the number of abnormal samples. represents the number of second correct answer samples.

[0100] As a possible implementation, as Figure 3 shown in the flowchart of the third note generation method based on an intelligent learning terminal, on the basis of the above embodiments, step 103 is further explained, including the following steps:

[0101] Step 301: Based on the knowledge point answer samples, determine the number of first correct answer samples, the number of second correct answer samples, and the total number of knowledge point answer samples.

[0102] In some embodiments, by counting the knowledge point answer samples, the number of first correct answer samples, the number of second correct answer samples, and the total number of knowledge point answer samples can be determined.

[0103] Step 302: Based on the number of first correct answer samples, the number of second correct answer samples, the total number, and the average dependence error, determine the dependence error matrix.

[0104] In some embodiments, based on the number of first correct answer samples, the number of second correct answer samples, the total number, and the average dependence error, the dependence error matrix can be determined by combining the following formula 2:

[0105] Formula 2

[0106] where represents the dependence error matrix. represents the number of first correct answer samples, and n represents the total number.

[0107] Exemplarily, according to the dependency graph, when it is determined that there is a dependency relationship between the first knowledge point combination (i, j), the average dependency error can be multiplied by the number of correct answer samples of the knowledge point to determine the dependency error matrix, where the first knowledge point combination is a combination formed by any two first knowledge points in the knowledge point answer samples.

[0108] Exemplarily, according to the dependency graph, when it is determined that neither the first knowledge point combination (i, j) nor its inverse pair (j, i) has a dependency relationship, it indicates that these two first knowledge points have no correlation. Therefore, the dependency error matrix can be determined by multiplying the error probability of the knowledge point by the number of correct samples of the knowledge point to determine the dependency error matrix.

[0109] Exemplarily, according to the dependency graph, when it is determined that there is a dependency relationship between the first knowledge point combination (j, i), the dependency error matrix is determined by subtracting the number of correct samples of the knowledge point from the number of correct samples of the knowledge point and then adding the number of correct samples of the knowledge point multiplied by the average dependency error of the knowledge point dependency.

[0110] As a possible implementation manner, as shown in Figure 4 the schematic flowchart of the fourth note generation method based on an intelligent learning terminal, on the basis of the above embodiments, step 104 is further explained, including the following steps:

[0111] Step 401, based on the answer count matrix and the dependency error matrix, determine the dependency error of the set of dependent knowledge points.

[0112] In some embodiments, based on the answer count matrix and the dependency error matrix, combined with the following formula 3, the dependency error of the set of dependent knowledge points is determined, including:

[0113] Formula 3,

[0114] where Diff represents the dependency error of the set of dependent knowledge points, m represents the diagonal elements in the answer count matrix and the dependency error matrix, represents the answer count matrix.

[0115] Step 402, when the dependency error is greater than the preset error threshold, determine that there is an error in the first knowledge point included in the set of dependent knowledge points in the dependency graph.

[0116] In some embodiments, when the dependency error is greater than a preset error threshold, it is determined that there is an error in the first knowledge point included in the dependency knowledge point set in the dependency relationship graph; in other words, by judging whether the dependency error is greater than the preset error threshold, the dependency knowledge point set with a relatively large deviation in the dependency relationship graph is screened out.

[0117] Step 403: Based on the first knowledge point included in the dependency knowledge point set, use the attention network model to update the dependency relationship graph.

[0118] In some embodiments, screening out the dependency knowledge point set with a relatively large deviation in the dependency relationship graph can be the actual learning path in the knowledge point answer sample. Therefore, the features of the dependency knowledge point set with a dependency error greater than the preset threshold can be extracted and input into the attention network model to iteratively update the dependency relationship graph, so as to obtain the updated dependency relationship graph.

[0119] As a possible implementation, as Figure 5 shown in the flowchart of the fifth note generation method based on an intelligent learning terminal, on the basis of the above embodiments, step 105 is further explained, including the following steps:

[0120] Step 501: Based on the knowledge point answer sample, determine the correct answer probability of the first knowledge point set.

[0121] In some embodiments, the knowledge point answer sample can be statistically analyzed to determine the correct answer probability of the first knowledge point set.

[0122] In some embodiments, the correct answer probability of the first knowledge point set can also be calculated and determined according to the preset mastery probability of the first knowledge point set.

[0123] In some embodiments, the first knowledge point set corresponding to the knowledge point answer sample and the preset mastery probability of the first knowledge point set can be as shown in Table 4:

[0124] Table 4

[0125]

[0126] Among them, .

[0127] Furthermore, the correct answer probability can be calculated and determined through the following formula:

[0128] ,

[0129] where R(q) represents the correct answer probability, represents whether the current knowledge point answer sample q contains the first knowledge point set , and its calculation method is that if the current knowledge point answering sample q contains the first knowledge point set , then takes the value of 1; if the current knowledge point answering sample q does not contain the first knowledge point set , then takes the value of 0.

[0130] Step 502: Update the preset mastery probability of the first knowledge point set based on the first knowledge point set corresponding to a single knowledge point answering sample and the correct answering probability to obtain the sample mastery probability.

[0131] In some embodiments, the update weight of the preset mastery probability can be determined based on the correct answering probability; the updated mastery probability can be determined based on the answering situation of the knowledge point answering sample, the first knowledge point set corresponding to a single knowledge point answering sample, the preset mastery probability, and the update weight; and the updated mastery probability is normalized to obtain the sample mastery probability.

[0132] Specifically, the update weight can be determined by the following formula

[0133] ,

[0134] where θ and θ compl represent the update weight, and k represents a preset parameter.

[0135] Specifically, the mastery probability is updated by the following formula:

[0136] If the current knowledge point answering sample q is a sample with a correct answer, the first mastery probability can be determined by the following formula:

[0137] ,

[0138] where represents the first mastery probability;

[0139] If the current knowledge point answering sample q is a sample with a wrong answer, the second mastery probability can be determined by the following formula:

[0140] ,

[0141] where represents the second mastery probability;

[0142] Furthermore, the first mastery probability and the second mastery probability are normalized, and the result of the normalization is determined as the sample mastery probability.

[0143] Step 503: Determine the weak knowledge point set of the knowledge point answering sample based on the sample mastery probability and the preset knowledge point mastery threshold.

[0144] When the sample mastery probability is less than the preset knowledge point mastery threshold, one or more first knowledge points corresponding to the sample mastery probability can be determined as weak knowledge points, thereby constructing a weak knowledge point set.

[0145] Among them, the preset knowledge point mastery threshold can be multiple thresholds to distinguish the weakness degree of the first knowledge point. For example, when the sample mastery probability is less than 30, it can be considered that the first knowledge point corresponding to the sample mastery probability is not mastered; for example, when the sample mastery probability is less than 90 and greater than or equal to 30, it can be considered that the first knowledge point corresponding to the sample mastery probability needs to be consolidated.

[0146] Corresponding to the above-mentioned note generation method based on an intelligent learning terminal, the present invention also proposes a note generation device based on an intelligent learning terminal. Since the device embodiment of the present invention corresponds to the above-mentioned method embodiment, for the details not disclosed in the device embodiment, reference can be made to the above-mentioned method embodiment, and the present invention will not be elaborated herein.

[0147] Figure 6 It is a schematic structural diagram of a note generation device based on an intelligent learning terminal provided by an embodiment of the present disclosure, as Figure 6 shown, the device includes:

[0148] A construction unit 610, configured to extract the correlation features of the first knowledge points by using an attention network model based on a preset first knowledge point learning relationship, so as to construct a dependency graph of the first knowledge points, where the first knowledge points include a pre - first knowledge point and a dependent first knowledge point;

[0149] A first determination unit 620, configured to obtain a knowledge point answer sample in response to a dependency graph update instruction, so as to determine the average dependency error of the first knowledge points in the dependency graph and the answer count matrix of the first knowledge points;

[0150] A second determination unit 630, configured to determine a dependency error matrix of the dependent first knowledge points based on the average dependency error and the answer situation of the knowledge point answer sample;

[0151] An update unit 640, configured to determine the error of the preset first knowledge point learning relationship based on the answer count matrix and the dependency error matrix, so as to update the dependency graph;

[0152] A third determination unit 650, configured to update the preset mastery probability of the first knowledge point set based on the first knowledge point set corresponding to a single knowledge point answer sample, so as to determine a weak knowledge point set corresponding to the knowledge point answer sample;

[0153] A generation unit 660, configured to generate learning notes by using preset learning materials based on the weak knowledge point set and the updated dependency graph.

[0154] In some embodiments of the present disclosure, the construction unit 610 is further configured to: extract the associated feature vector of the first knowledge point based on a preset first knowledge point learning relationship; construct an initial relationship graph of the first knowledge point by using the triple method based on the associated feature vector; vectorize the initial relationship graph to obtain the vectorized encoding of the first knowledge point corresponding to adjacent nodes; determine the implicit features between the first knowledge points based on the vectorized encoding; and update the initial relationship graph based on the implicit features to construct a dependency graph.

[0155] In some embodiments of the present disclosure, the first determination unit 620 is further configured to: determine, based on the dependency graph, the first correct answer sample of the pre - first knowledge point corresponding to the knowledge point answer sample, the first wrong answer sample depending on the first knowledge point, the second correct answer sample depending on the first knowledge point, and the number of first knowledge points depending on the pre - first knowledge point; determine the abnormal samples and the answer count matrix in the knowledge point answer sample based on the first correct answer sample, the second correct answer sample, and the first wrong answer sample; and determine the average dependency error based on the set of dependent knowledge points corresponding to the knowledge point answer sample, the second correct answer sample, the abnormal samples, and the number of first knowledge points depending on the pre - first knowledge point, where the set of dependent knowledge points is composed of the pre - first knowledge point and the dependent first knowledge point.

[0156] In some embodiments of the present disclosure, the first determination unit 620 is further configured to: when it is determined, based on the first correct answer sample and the first wrong answer sample, that the pre - first knowledge point corresponding to the same sample source has a wrong answer and the first knowledge point depending on it has a correct answer, determine the first correct answer sample corresponding to the same sample source as an abnormal sample; and determine the answer count matrix based on the number of first correct answer samples, the number of second correct answer samples, and the number of abnormal samples.

[0157] In some embodiments of the present disclosure, the first determination unit 620 is further configured to: determine the average dependency error based on the set of dependent knowledge points corresponding to the knowledge point answer sample, the second correct answer sample, the abnormal samples, and the number of first knowledge points depending on the pre - first knowledge point, in combination with the following formula 1:

[0158] Formula 1

[0159] where M represents the average dependency error, N(E) represents the number of first knowledge points depending on the pre - first knowledge point, i represents the pre - first knowledge point, j represents the first knowledge point depending on it, E represents the set of dependent knowledge points, represents the number of abnormal samples, represents the number of second correct answer samples.

[0160] In some embodiments of the present disclosure, the second determination unit 630 is further configured to: based on the knowledge point answer samples, determine the number of first correct answer samples, the number of second correct answer samples, and the total number of knowledge point answer samples; based on the number of first correct answer samples, the number of second correct answer samples, the total number, and the average dependence error, and in combination with the following formula 2, determine the dependence error matrix:

[0161] Formula 2

[0162] wherein represents the dependence error matrix, represents the number of first correct answer samples, and n represents the total number.

[0163] In some embodiments of the present disclosure, the updating unit 640 is further configured to: based on the answer count matrix and the dependence error matrix, and in combination with the following formula 3, determine the dependence error of the dependence knowledge point set, including:

[0164] Formula 3

[0165] wherein Diff represents the dependence error of the dependence knowledge point set, m represents the diagonal element in the answer count matrix and the dependence error matrix, represents the answer count matrix;

[0166] When the dependence error is greater than the preset error threshold, it is determined that there is an error in the first knowledge point included in the dependence knowledge point set in the dependence relationship graph;

[0167] Based on the first knowledge point included in the dependence knowledge point set, use the attention network model to update the dependence relationship graph.

[0168] In some embodiments of the present disclosure, the third determination unit 650 is further configured to: based on the knowledge point answer samples, determine the correct answer probability of the first knowledge point set; based on the first knowledge point set corresponding to a single knowledge point answer sample and the correct answer probability, update the preset mastery probability of the first knowledge point set to obtain the sample mastery probability; based on the sample mastery probability and the preset knowledge point mastery threshold, determine the weak knowledge point set of the knowledge point answer samples.

[0169] In some embodiments of the present disclosure, the third determination unit 650 is further configured to: based on the correct answer probability, determine the update weight of the preset mastery probability; based on the answer situation of the knowledge point answer samples, the first knowledge point set corresponding to a single knowledge point answer sample, the preset mastery probability, and the update weight, determine the updated mastery probability; perform normalization processing on the updated mastery probability to obtain the sample mastery probability.

[0170] In some embodiments of the present disclosure, the third determination unit 650 is further configured to: based on the sample mastery probability and the preset knowledge point weight, and in combination with the following formula 4, determine the comprehensive mastery probability of the first knowledge point in the knowledge point answer sample:

[0171] Formula 4

[0172] where , represents the comprehensive mastery probability, represents the preset knowledge point weight, represents the first knowledge point, represents the first knowledge point set, represents the sample mastery probability; compare the comprehensive mastery probability with the preset knowledge point mastery threshold to determine the set of weak knowledge points.

[0173] In some embodiments of the present disclosure, the generation determination unit 660 is further configured to: match the first knowledge point in the set of weak knowledge points with the updated dependency graph to determine the knowledge points that the learning note needs to include; based on the knowledge points that the learning note needs to include, extract the corresponding knowledge content from the preset learning materials to generate a learning note.

[0174] It should be noted that the foregoing explanation of the method embodiments also applies to the apparatus in this embodiment. The principle is the same and will not be limited in this embodiment.

[0175] Based on the method as shown above Figures 1 to 5 In corresponding, this embodiment further provides a computer program product, including a computer program, which when executed by a processor, implements the method as shown above Figures 1 to 5 shown.

[0176] Based on the method as shown above Figures 1 to 5 In corresponding, this embodiment further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the method as shown above Figures 1 to 5 shown.

[0177] Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods in various implementation scenarios of the present application.

[0178] As Figure 7 shown is a schematic hardware structure diagram of an electronic device according to the present invention, including:

[0179] At least one processor 701; and,

[0180] A memory 702 communicatively connected to at least one of the processors 701; wherein,

[0181] The memory 702 stores instructions executable by at least one of the processors, and the instructions are executed by at least one of the processors to enable at least one of the processors to execute a note generation method based on an intelligent learning terminal as described above.

[0182] Figure 7 Taking one processor 701 as an example.

[0183] The electronic device may further include: an input device 703 and a display device 704.

[0184] The processor 701, the memory 702, the input device 703, and the display device 704 may be connected through a bus or other means. In the figure, connection through a bus is taken as an example.

[0185] The memory 702, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the review content generation method in the embodiments of the present application. For example, Figures 1 to 5 The method flow shown. The processor 701 executes various functional applications and data processing by running the non-volatile software programs, instructions, and modules stored in the memory 702, that is, implements a note generation method based on an intelligent learning terminal in the above embodiments.

[0186] The memory 702 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the review content generation method, etc. In addition, the memory 702 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory 702 may optionally include a memory remotely set relative to the processor 701, and these remote memories can be connected to the device executing the review content generation method through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0187] The input device 703 can receive input user clicks and generate signal inputs related to user settings and function controls of the review content generation method. The display device 704 may include a display screen and other display devices.

[0188] When the one or more modules are stored in the memory 702 and run by the one or more processors 701, a note generation method based on an intelligent learning terminal in any of the above method embodiments is executed.

[0189] Optionally, the above-mentioned entity device may further include a user interface, a network interface, a camera, a Radio Frequency (RF) circuit, sensors, an audio circuit, a WI-FI module, and so on. The user interface may include a display screen (Display) and an input unit such as a keyboard (Keyboard), etc. Optionally, the user interface may further include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0190] Those skilled in the art can understand that the above-mentioned entity device structure provided in this embodiment does not constitute a limitation on the entity device, and it may include more or fewer components, or combine certain components, or have different component arrangements.

[0191] The storage medium may further include an operating system and a network communication module. The operating system is a program for managing the hardware and software resources of the above-mentioned entity device, and supports the operation of information processing programs and other software and / or programs. The network communication module is used to implement communication between the components inside the storage medium, as well as communication between other hardware and software in the information processing entity device.

[0192] Through the description of the above embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus a necessary general hardware platform, or can also be implemented by hardware.

[0193] By applying the solution of this embodiment, compared with the current prior art, this solution constructs a dependency graph through a preset first knowledge point learning relationship, and constructs an answer counting matrix and a dependency error matrix based on the knowledge point answer samples and the dependency graph to correct the errors existing in the dependency graph, improving the accuracy of determining the association relationship between the first knowledge points; and then determines the set of weak knowledge points corresponding to the knowledge point answer samples according to the answer conditions of the knowledge point answer samples, so as to match the set of weak knowledge points with the corrected dependency graph, and uses the preset learning materials to generate personalized learning notes corresponding to the knowledge point answer samples, improving the flexibility and accuracy of learning note generation.

[0194] It should be noted that in this text, relational terms such as "front" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

[0195] The above are only specific embodiments of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments described herein, but rather to the broadest scope consistent with the principles and novel features claimed herein.

Claims

1. A note generation method based on an intelligent learning terminal, characterized in that: The method comprises: Based on the preset first knowledge point learning relationship, the attention network model is used to extract the associated features of the first knowledge point to construct a dependency graph of the first knowledge point, wherein the first knowledge point includes a preceding first knowledge point and a dependent first knowledge point; In response to a dependency graph update instruction, obtaining a knowledge point answer sample to determine an average dependency error of the first knowledge point in the dependency graph and an answer count matrix of the first knowledge point; Determine the dependency error matrix of the first knowledge point based on the average dependency error and the answer status of the knowledge point answer samples; Based on the answer count matrix and the dependency error matrix, determining the error of the preset first knowledge point learning relationship to update the dependency relationship graph; Based on a first knowledge point set corresponding to a single knowledge point answer sample, updating a preset mastering probability of the first knowledge point set to determine a weak knowledge point set corresponding to the knowledge point answer sample; Based on the weak knowledge point set and the updated dependency graph, using preset learning materials, generating learning notes; The step of obtaining knowledge point answer samples to determine the average dependency error of the first knowledge point and the answer count matrix of the first knowledge point in the dependency graph includes: Based on the dependency graph, determine the first correct answer sample of the preceding first knowledge point, the first incorrect answer sample that depends on the first knowledge point, the second correct answer sample that depends on the first knowledge point, and the number of the preceding first knowledge point that depends on the first knowledge point in the knowledge point answer sample; Based on the first correct answer sample, the second correct answer sample and the first incorrect answer sample, determining abnormal samples in the knowledge point answer samples and the answer count matrix; The average dependency error is determined based on the dependent knowledge point set corresponding to the knowledge point answer sample, the second correct answer sample, the abnormal sample and the number of dependent first knowledge points of the preceding first knowledge point, wherein the dependent knowledge point set is composed of the preceding first knowledge point and the dependent first knowledge point.

2. The method according to claim 1, characterized in that: The method of extracting the associated features of the first knowledge point based on the preset first knowledge point learning relationship and using the attention network model to construct a dependency graph of the first knowledge point includes: Based on the preset first knowledge point learning relationship, extracting a correlation feature vector of the first knowledge point; Based on the associated feature vector, construct an initial relationship graph of the first knowledge point using a triple method; Vectorizing the initial relationship graph to obtain vectorized codes of first knowledge points corresponding to adjacent nodes; Based on the vectorized encoding, determining invisible features between the first knowledge points; Based on the invisible features, the initial relationship graph is updated to construct a dependency relationship graph.

3. The method according to claim 1, characterized in that The determining, based on the first correct answer sample and the second correct answer sample, abnormal samples in the knowledge point answer sample and the answer count matrix comprises: Based on the first correct answer sample and the first incorrect answer sample, when it is determined that the answer to the preceding first knowledge point corresponding to the same sample source is incorrect and the answer to the dependent first knowledge point is correct, the first correct answer sample corresponding to the same sample source is determined as the abnormal sample; The answer count matrix is ​​determined based on the number of the first correct answer samples, the number of the second correct answer samples, and the number of the abnormal samples.

4. The method according to claim 1, characterized in that: The determining the average dependency error based on the dependent knowledge point set corresponding to the knowledge point answer sample, the second correct answer sample, the abnormal sample and the number of dependent first knowledge points of the preceding first knowledge point comprises: Based on the dependent knowledge point set corresponding to the knowledge point answer sample, the second correct answer sample, the abnormal sample and the number of dependent first knowledge points of the preceding first knowledge point, the average dependency error is determined in combination with the following formula 1: Formula 1, Wherein, M represents the average dependency error, N(E) represents the number of dependent first knowledge points corresponding to the preceding first knowledge point, i represents the preceding first knowledge point, j represents the dependent first knowledge point, and E represents the dependent knowledge point set. represents the number of abnormal samples, Represents the number of the second correct answer samples.

5. The method according to claim 1, characterized in that: The step of determining the dependency error matrix of the first knowledge point based on the average dependency error and the answer status of the knowledge point answer samples includes: Based on the knowledge point answer samples, determine the number of the first correct answer samples, the number of the second correct answer samples, and the total number of the knowledge point answer samples; Based on the number of the first correct answer samples, the number of the second correct answer samples, the total number and the average dependency error, the dependency error matrix is ​​determined in combination with the following formula 2: Formula 2, Among them, the represents the dependency error matrix, the represents the number of the first correct answer samples, and n represents the total number.

6. The method according to claim 1, characterized in that The step of determining the error of the preset first knowledge point learning relationship based on the answer count matrix and the dependency error matrix to update the dependency graph includes: Based on the answer count matrix and the dependency error matrix, combined with the following formula 3, determining the dependency error of the dependent knowledge point set includes: Formula 3, Wherein, Diff represents the dependency error of the dependent knowledge point set, m represents the diagonal elements in the answer count matrix and the dependency error matrix, and represents the answer count matrix; When the dependency error is greater than a preset error threshold, determining that a first knowledge point included in the dependent knowledge point set has an error in the dependency relationship graph; Based on the first knowledge point included in the dependent knowledge point set, the dependency graph is updated using the attention network model.

7. The method according to claim 1, characterized in that The updating of the preset mastering probability of the first knowledge point set based on the first knowledge point set corresponding to the single knowledge point answer sample to determine the weak knowledge point set corresponding to the knowledge point answer sample includes: Determining the probability of correct answers to the first knowledge point set based on the knowledge point answer samples; Based on the first knowledge point set corresponding to the single knowledge point answer sample and the correct answer probability, updating the preset mastery probability of the first knowledge point set to obtain a sample mastery probability; Based on the sample mastery probability and a preset knowledge point mastery threshold, a set of weak knowledge points of the knowledge point answer sample is determined.

8. The method according to claim 7, characterized in that The updating of the preset mastering probability of the first knowledge point set based on the first knowledge point set corresponding to the single knowledge point answer sample and the correct answer probability to obtain the sample mastering probability includes: Based on the correct answer probability, determining an update weight for the preset mastery probability; Determine an updated mastering probability based on the answering status of the knowledge point answering sample, the first knowledge point set corresponding to the single knowledge point answering sample, the preset mastering probability, and the updating weight; The updated mastering probability is normalized to obtain the sample mastering probability.

9. The method according to claim 7, characterized in that: The determining of the weak knowledge point set of the knowledge point answer sample based on the sample mastery probability and the preset knowledge point mastery threshold comprises: Based on the sample mastery probability and the preset knowledge point weight, combined with the following formula 4, the comprehensive mastery probability of the first knowledge point in the knowledge point answer sample is determined: Formula 4, in, , represents the comprehensive mastery probability, represents the preset knowledge point weight, represents the first knowledge point, represents the first knowledge point set, represents the probability of mastering the sample; The comprehensive mastery probability is compared with the preset knowledge point mastery threshold to determine the weak knowledge point set.

10. The method according to claim 1, characterized in that The generating of learning notes based on the weak knowledge point set and the updated dependency graph using preset learning materials includes: Matching the first knowledge point in the weak knowledge point set with the updated dependency graph to determine the knowledge points that the study notes need to include; Based on the knowledge points that the study notes need to contain, corresponding knowledge content is extracted from preset learning materials to generate the study notes.

Images