A method for generating personalized question practice plans with intelligent integration of wrong questions

Abstract

The invention discloses a method for generating a personalized question practice plan for intelligent integration of wrong questions, comprising: collecting students' wrong question text data, dividing the wrong question text data into different wrong question categories by adaptive clustering; extracting knowledge points involved in different wrong question text data based on co-occurrence information of word segmentation results of the wrong question text data in different wrong question categories; calculating the similarity of knowledge points between wrong question text data in combination with knowledge points and difficulty statistics of the wrong question text data, and constructing a wrong question knowledge point similarity matrix; constructing an answer tendency matrix according to students' historical answer question data and students' unanswered question data in a question bank and performing matrix decomposition to obtain a semantic feature vector of each question; constructing a personalized question practice plan generation model to calculate students' answer tendency on different questions, and selecting the first several unanswered questions from high to low according to the answer tendency to generate a personalized question practice plan.

Description

Technical Field

[0001] The present invention relates to the field of intelligent information technology, and in particular to a method for generating a personalized question practice plan by intelligently integrating wrong questions. Background Art

[0002] In the current educational environment, students are faced with more and more exams and assessments, which have a huge impact on their scores and academic performance. Therefore, it is crucial for students to master knowledge and skills quickly and effectively, which often requires a lot of exercise training.

[0003] Since different students have different weaknesses in their studies, it is not easy for them to find exercises that suit them. Currently, most exercise books on the market are universal, highly homogeneous, lacking in specificity, and cannot be adjusted individually based on the actual situation of the students. This is not conducive to the growth of students, nor is it suitable for the current personalized education concept. Summary of the invention

[0004] In view of this, the present invention provides a method for generating a personalized question practice plan with intelligent integration of wrong questions, the purpose of which is to generate a personalized question practice plan for wrong questions based on the students' actual learning situation by using artificial intelligence technology to improve students' learning efficiency and learning interest.

[0005] In order to achieve the above object, the present invention provides a method for generating a personalized question practice plan by intelligently integrating wrong questions, comprising the following steps:

[0006] S1: Collect students' wrong question text data and perform word segmentation and vector encoding processing, and divide the wrong question text data into different wrong question categories through adaptive clustering;

[0007] S2: Extract knowledge points based on the co-occurrence information of the word segmentation results of the wrong question text data in different wrong question categories, and obtain the knowledge points involved in different wrong question text data;

[0008] S3: Combine the knowledge points and difficulty statistics of the wrong question text data to calculate the similarity of the knowledge points between the wrong question text data, and construct a wrong question knowledge point similarity matrix;

[0009] S4: Based on the students’ historical answer data and the students’ unanswered questions in the question bank, a question answering tendency matrix is ​​constructed and matrix decomposition is performed to obtain the semantic feature vector of each question;

[0010] S5: Construct a personalized question practice plan generation model, which takes the semantic feature vector of each question and the similarity matrix of wrong question knowledge points as input, integrates graph convolution and attention mechanism to calculate the students' answering tendency on different questions, and selects the first several unanswered questions from high to low according to the answering tendency to generate a personalized question practice plan.

[0011] Preferably, the step S1 comprises:

[0012] S11: Collect students’ wrong questions text data:

[0013] ;

[0014] Where: represents the wrong question text data of the student's k-th wrong answer, It indicates the total number of questions that students answered incorrectly;

[0015] S12: Perform word segmentation on the wrong question text data, where the wrong question text data The segmentation result is:

[0016] ;

[0017] Where: Indicates wrong question text data The i-th word segmentation result, Indicates wrong question text data The total number of word segmentation results;

[0018] The word segmentation results of the wrong question text data are encoded by word vectors, and the encoding vectors of the wrong question text data are constructed based on the word vector encoding, where the wrong question text data The corresponding encoding vector is:

[0019] ;

[0020] Where: Indicates wrong question text data The corresponding encoding vector, Indicates word segmentation results The word vector encoding of ;

[0021] S13: Calculate the distance between any two wrong question text data, where the wrong question text data and The distance between them is:

[0022] ;

[0023] Where: Indicates wrong question text data and The distance between , represents the L2 norm, represents the feature transformation matrix;

[0024] S14: Use K-means clustering algorithm to cluster K wrong question text data and obtain U cluster centers, where the uth cluster center is , , cluster center The corresponding encoding vector is ;

[0025] S15: Filter the misclassified wrong question text data adjacent to multiple cluster centers from the students' wrong question text data, where the set of misclassified wrong question text data is:

[0026] ;

[0027] Where: represents the mth misclassified wrong question text data obtained by screening, and M represents the total number of misclassified wrong question text data obtained by screening;

[0028] S16: Based on manual judgment, the misclassified wrong question text data is classified into the cluster where the correct cluster center is located, and the misclassification association information matrix is ​​constructed:

[0029] ;

[0030] Where: Q represents the misclassification association information matrix, is the element value of the element in the kth row and eth column in the misclassified association information matrix, representing the wrong question text data and If the misclassification association information between and Both are misclassified wrong question text data, and they are manually judged to be in the same cluster. ,otherwise ;

[0031] S17: Reconstruct the encoding vector of the wrong question text data based on the misclassification association information matrix:

[0032] ;

[0033] Where: T represents transpose, Represents the encoding vector The reconstruction result of

[0034] S18: Recalculate the distance between wrong question text data based on the reconstructed encoding vector, cluster K wrong question text data using K-means clustering algorithm, and obtain U cluster centers and cluster clusters. The wrong question categories of the wrong question text data in each cluster are the same, and the wrong question text data The error category is , , correspond Types of wrong questions.

[0035] Preferably, the step S2 comprises:

[0036] S21: Calculate wrong question text data The co-occurrence information of the word segmentation results in The co-occurrence information is:

[0037] ;

[0038] Where: Indicates wrong question text data Chinese word segmentation results The co-occurrence information of express A set of word segmentation results of wrong question text data, Represents any word segmentation result in the word segmentation result set; Indicates word segmentation results In the wrong category The number of co-occurrences in the word segmentation results In the wrong question text data that appears at the same time, the wrong question category is The number of wrong question text data; Indicates word segmentation results The number of co-occurrences in the remaining wrong question categories, that is, in the word segmentation results In the wrong question text data that appears at the same time, the wrong question category is not The number of wrong question text data; Indicates word segmentation results The adjacent word set of The set of other word segmentation results that appear in the same wrong question text data; Indicates any word segmentation result The set of other word segmentation results that appear in the same wrong question text data; Indicates the number of word segmentation results in the calculation set;

[0039] S22: Select the E word segmentation results with the largest co-occurrence information as the wrong question text data Knowledge points:

[0040] ;

[0041] Where: Indicates wrong question text data The Eth knowledge point.

[0042] Preferably, step S3 comprises:

[0043] S31: Obtaining difficulty statistics of wrong question text data, i.e., wrong question text data Difficulty of the corresponding question ;

[0044] S32: Calculate wrong question text data and Similarity between:

[0045] ;

[0046] Where: Indicates wrong question text data and The similarity between represents an exponential function with a natural constant as base, represents the L1 norm, Representing knowledge points The word vector encoding of ;

[0047] S33: Based on the similarity of knowledge points between wrong question text data, a wrong question knowledge point similarity matrix is ​​constructed.

[0048] Preferably, the step S33 includes:

[0049] S331: Extract all knowledge points of wrong question text data and remove duplicates:

[0050] ;

[0051] Where: represents the sth knowledge point after deduplication, and S represents the total number of knowledge points after deduplication;

[0052] S332: Calculate the similarity of knowledge points after deduplication and construct a similarity matrix of wrong question knowledge points:

[0053] ;

[0054] Where: Z represents the similarity matrix of wrong question knowledge points, Represents the element value of the element in the sth row and rth column of the wrong question knowledge point similarity matrix Z, corresponding to the knowledge point The wrong question text data and the knowledge points included The mean similarity between the wrong question text data, .

[0055] Preferably, step S4 comprises:

[0056] S41: Obtain the student's historical answer data and the student's unanswered question data in the question bank:

[0057] ;

[0058] Where: Represents the student's historical answer data. represents the j-th history question data answered by the student, Indicates the total number of questions answered by students. ; Represents the data of students’ unanswered questions in the question bank. It represents the data of the pth question that the student has not answered in the question bank. Indicates the total number of questions in the question bank that students have not answered. ;

[0059] S42: Construct a corresponding answer tendency matrix for each question data, where the historical answer question data The answer tendency matrix is:

[0060] ;

[0061] Where: Indicates historical answer question data The answer tendency matrix, Indicates the data of students answering questions in history The answer situation, Indicates the data of students answering historical questions correctly , Indicates the data of students' wrong answers to historical questions , Indicates historical answer question data The knowledge point information of the corresponding question text data, Indicates historical answer question data The corresponding question text data contains knowledge points , Indicates historical answer question data The corresponding question text data does not contain any knowledge point ;

[0062] Unanswered Questions Data The answer tendency matrix is:

[0063] ;

[0064] Where: Indicates data for unanswered questions The answer tendency matrix, Indicates data for unanswered questions The knowledge point information of the corresponding question text data, Indicates data for unanswered questions The corresponding question text data contains knowledge points , Indicates data for unanswered questions The corresponding question text data does not contain any knowledge point ;

[0065] S43: Perform matrix decomposition on the answer tendency matrix to obtain the semantic feature vector of each question, where the historical answer question data The semantic feature vector of , unanswered questions data The semantic feature vector of .

[0066] Preferably, the step S43 includes:

[0067] Question answering tendency matrix Perform matrix decomposition and calculate ;

[0068] right Perform eigendecomposition and select the eigenvector corresponding to the maximum eigenvalue as the answer tendency matrix The decomposition vector is used as the question data The semantic feature vector of .

[0069] Preferably, step S5 comprises:

[0070] S51: Build a personalized question practice plan generation model, the model receives unanswered questions The semantic feature vector of Similarity matrix with wrong question knowledge points ;

[0071] S52: If the knowledge point similarity matrix of the wrong question Element value If it is greater than the preset threshold, then the knowledge point and There are edges between them, and the similarity matrix of wrong question knowledge points is calculated The corresponding degree matrix :

[0072] ;

[0073] Where: In the form of a diagonal matrix, Representation and knowledge points The number of associated edges;

[0074] S53: True or False Question Knowledge Point Similarity Matrix Unanswered questions The semantic feature vector of Perform graph convolution to obtain unanswered questions The fused semantic feature vector:

[0075] ;

[0076] Where: represents the convolution matrix, represents the convolution process, represents the activation function, Indicates that the question was not answered The fused semantic feature vector of

[0077] S54: Calculate the fusion semantic feature vector Attention distribution:

[0078] ;

[0079] Where: Represents the fused semantic feature vector The attention distribution of , T represents transposition, represents the attention mapping matrix;

[0080] S55: Calculate the number of questions that students have not answered The answer tendency of

[0081] ;

[0082] ;

[0083] Where: Indicates that the student did not answer the question The answer tendency of Represents the fused semantic feature vector The corresponding attention weights;

[0084] S56: Count the students' answering tendencies for all unanswered questions, and select the first several unanswered questions from high to low according to the answering tendencies to form a personalized question practice plan.

[0085] The present invention further provides an electronic device, comprising:

[0086] A memory storing at least one instruction;

[0087] Communication interface, enabling electronic equipment to communicate; and

[0088] The processor executes the instructions stored in the memory to implement the above-mentioned method for generating personalized question practice plans by intelligently integrating wrong questions.

[0089] The present invention also provides a computer-readable storage medium, in which at least one instruction is stored. The at least one instruction is executed by a processor in an electronic device to implement the above-mentioned method for generating a personalized question practice plan with intelligent integration of wrong questions.

[0090] Compared with the prior art, the present invention has at least the following beneficial effects:

[0091] The present invention collects students' wrong question text data, performs initial clustering by using a low-sample-based clustering method, extracts misclassification information after clustering, recodes and re-clusters the wrong question text data, improves clustering accuracy under low wrong question samples, extracts knowledge points from wrong question text data of different wrong question categories based on co-occurrence information of word segmentation results in the wrong question text data in the same category of wrong question text data and all wrong question text data, comprehensively calculates the similarity of knowledge points between wrong question text data in combination with knowledge points and difficulty statistics of the wrong question text data, constructs a wrong question knowledge point similarity matrix, and realizes the similarity calculation of wrong question knowledge points.

[0092] The present invention constructs an answering tendency matrix and performs matrix decomposition based on students' historical answering question data and students' unanswered question data in a question bank, obtains a semantic feature vector representing the knowledge point information of each question, fuses the wrong question knowledge point similarity matrix and the semantic feature vector of the question by using a graph convolution method, represents the wrong question knowledge point information in the semantic feature vector, obtains a fused semantic feature vector of each question, realizes the mapping representation of question knowledge points and wrong question knowledge points, calculates the attention weight of each unanswered question by combining the attention mechanism, weights the fused semantic feature vector, focuses on the wrong question knowledge point information in the fused semantic feature vector, and maps it to an answering tendency, wherein the higher the answering tendency, the richer the student wrong question knowledge point information contained in the question, and realizes the generation of personalized question practice plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Figure 1 The present invention is a flowchart of a method for generating a personalized question practice plan by intelligently integrating wrong questions in Embodiment 1 of the present invention.

[0094] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0095] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0096] The embodiment of the present application provides a method for generating a personalized question practice solution with intelligent integration of wrong questions. The execution subject of the method for generating a personalized question practice solution with intelligent integration of wrong questions includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for generating a personalized question practice solution with intelligent integration of wrong questions can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0097] Embodiment 1:

[0098] S1: Collect students' wrong question text data and perform word segmentation and vector encoding. Use adaptive clustering to divide the wrong question text data into different wrong question categories, including:

[0099] S11: Collect students’ wrong questions text data:

[0100] ;

[0101] Where: represents the wrong question text data of the student's k-th wrong answer, It indicates the total number of questions that students answered incorrectly;

[0102] S12: Perform word segmentation on the wrong question text data, where the wrong question text data The segmentation result is:

[0103] ;

[0104] Where: Indicates wrong question text data The i-th word segmentation result, Indicates wrong question text data The total number of word segmentation results;

[0105] The word segmentation results of the wrong question text data are encoded by word vectors, and the encoding vectors of the wrong question text data are constructed based on the word vector encoding, where the wrong question text data The corresponding encoding vector is:

[0106] ;

[0107] Where: Indicates wrong question text data The corresponding encoding vector, Indicates word segmentation results In this embodiment, the word segmentation processing tool selected is the jieba word segmentation tool, and the word vector encoding method is the word2vec model;

[0108] S13: Calculate the distance between any two wrong question text data, where the wrong question text data and The distance between them is:

[0109] ;

[0110] Where: Indicates wrong question text data and The distance between , represents the L2 norm, represents the feature transformation matrix;

[0111] S14: Use K-means clustering algorithm to cluster K wrong question text data and obtain U cluster centers, where the uth cluster center is , , cluster center The corresponding encoding vector is ;

[0112] S15: Filter the misclassified wrong question text data adjacent to multiple cluster centers from the students' wrong question text data, where the set of misclassified wrong question text data is:

[0113] ;

[0114] Where: represents the mth misclassified wrong question text data obtained by screening, and M represents the total number of misclassified wrong question text data obtained by screening;

[0115] The screening process for misclassified wrong question text data is as follows:

[0116] S151: Calculate the misclassification information of the wrong question text data:

[0117] ;

[0118] Where: Indicates wrong question text data misclassification information, Indicates wrong question text data With cluster center distance;

[0119] S152: Filtering M wrong question text data with the highest misclassification information as misclassified wrong question text data;

[0120] S16: Based on manual judgment, the misclassified wrong question text data is classified into the cluster where the correct cluster center is located, and the misclassification association information matrix is ​​constructed:

[0121] ;

[0122] Where: Q represents the misclassification association information matrix, is the element value of the element in the kth row and eth column in the misclassified association information matrix, representing the wrong question text data and If the misclassification association information between and Both are misclassified wrong question text data, and they are manually judged to be in the same cluster. ,otherwise ;

[0123] S17: Reconstruct the encoding vector of the wrong question text data based on the misclassification association information matrix:

[0124] ;

[0125] Where: T represents transpose, Represents the encoding vector The reconstruction result of

[0126] S18: Recalculate the distance between wrong question text data based on the reconstructed encoding vector, cluster K wrong question text data using K-means clustering algorithm, and obtain U cluster centers and cluster clusters. The wrong question categories of the wrong question text data in each cluster are the same, and the wrong question text data The error category is , , correspond Types of wrong questions.

[0127] S2: Based on the co-occurrence information of the word segmentation results of the wrong question text data in different wrong question categories, knowledge points are extracted to obtain the knowledge points involved in different wrong question text data, including:

[0128] S21: Calculate wrong question text data The co-occurrence information of the word segmentation results in The co-occurrence information is:

[0129] ;

[0130] Where: Indicates wrong question text data Chinese word segmentation results The co-occurrence information of express A set of word segmentation results of wrong question text data, Represents any word segmentation result in the word segmentation result set; Indicates word segmentation results In the wrong category The number of co-occurrences in the word segmentation results In the wrong question text data that appears at the same time, the wrong question category is The number of wrong question text data; Indicates word segmentation results The number of co-occurrences in the remaining wrong question categories, that is, in the word segmentation results In the wrong question text data that appears at the same time, the wrong question category is not The number of wrong question text data; Indicates word segmentation results The adjacent word set of The set of other word segmentation results that appear in the same wrong question text data; Indicates any word segmentation result The set of other word segmentation results that appear in the same wrong question text data; Indicates the number of word segmentation results in the calculation set;

[0131] S22: Select the E word segmentation results with the largest co-occurrence information as the wrong question text data Knowledge points:

[0132] ;

[0133] Where: Indicates wrong question text data The Eth knowledge point.

[0134] S3: Combine the knowledge points and difficulty statistics of the wrong question text data to calculate the similarity of the knowledge points between the wrong question text data, and construct a wrong question knowledge point similarity matrix, including:

[0135] S31: Obtaining difficulty statistics of wrong question text data, i.e., wrong question text data Difficulty of the corresponding question ;

[0136] S32: Calculate wrong question text data and Similarity between:

[0137] ;

[0138] Where: Indicates wrong question text data and The similarity between represents an exponential function with a natural constant as base, represents the L1 norm, Representing knowledge points The word vector encoding of ;

[0139] S33: Based on the similarity of knowledge points between wrong question text data, a wrong question knowledge point similarity matrix is ​​constructed, including:

[0140] S331: Extract all knowledge points of wrong question text data and remove duplicates:

[0141] ;

[0142] Where: represents the sth knowledge point after deduplication, and S represents the total number of knowledge points after deduplication;

[0143] S332: Calculate the similarity of knowledge points after deduplication and construct a similarity matrix of wrong question knowledge points:

[0144] ;

[0145] Where: Z represents the similarity matrix of wrong question knowledge points, Represents the element value of the element in the sth row and rth column of the wrong question knowledge point similarity matrix Z, corresponding to the knowledge point The wrong question text data and the knowledge points included The mean similarity between the wrong question text data, .

[0146] S4: Based on the students’ historical answer data and the students’ unanswered questions in the question bank, we construct the answer tendency matrix and perform matrix decomposition to obtain the semantic feature vector of each question, including:

[0147] S41: Obtain the student's historical answer data and the student's unanswered question data in the question bank:

[0148] ;

[0149] Where: Represents the student's historical answer data. represents the j-th history question data answered by the student, Indicates the total number of questions answered by students. ; Represents the data of students’ unanswered questions in the question bank. It represents the data of the pth question that the student has not answered in the question bank. Indicates the total number of questions in the question bank that students have not answered. ;

[0150] S42: Construct a corresponding answer tendency matrix for each question data, where the historical answer question data The answer tendency matrix is:

[0151] ;

[0152] Where: Indicates historical answer question data The answer tendency matrix, Indicates the data of students answering questions in history The answer situation, Indicates the data of students answering historical questions correctly , Indicates the data of students' wrong answers to historical questions , Indicates historical answer question data The knowledge point information of the corresponding question text data, Indicates historical answer question data The corresponding question text data contains knowledge points , Indicates historical answer question data The corresponding question text data does not contain any knowledge point ;

[0153] Unanswered Questions Data The answer tendency matrix is:

[0154] ;

[0155] Where: Indicates data for unanswered questions The answer tendency matrix, Indicates data for unanswered questions The knowledge point information of the corresponding question text data, Indicates data for unanswered questions The corresponding question text data contains knowledge points , Indicates data for unanswered questions The corresponding question text data does not contain any knowledge point ;

[0156] S43: Perform matrix decomposition on the answer tendency matrix to obtain the semantic feature vector of each question, where the historical answer question data The semantic feature vector of , unanswered questions data The semantic feature vector of ,include:

[0157] Question answering tendency matrix Perform matrix decomposition and calculate ;

[0158] right Perform eigendecomposition and select the eigenvector corresponding to the maximum eigenvalue as the answer tendency matrix The decomposition vector is used as the question data The semantic feature vector of .

[0159] S5: Construct a personalized question practice plan generation model, which takes the semantic feature vector of each question and the similarity matrix of wrong question knowledge points as input, integrates graph convolution and attention mechanism to calculate the students' answering tendency on different questions, and selects several unanswered questions from high to low according to the answering tendency to generate personalized question practice plans, including:

[0160] S51: Build a personalized question practice plan generation model, the model receives unanswered questions The semantic feature vector of Similarity matrix with wrong question knowledge points ;

[0161] S52: If the knowledge point similarity matrix of the wrong question Element value If it is greater than the preset threshold, then the knowledge point and There are edges between them, and the similarity matrix of wrong question knowledge points is calculated The corresponding degree matrix :

[0162] ;

[0163] Where: In the form of a diagonal matrix, Representation and knowledge points The number of associated edges;

[0164] S53: True or False Question Knowledge Point Similarity Matrix Unanswered questions The semantic feature vector of Perform graph convolution to get unanswered questions The fused semantic feature vector:

[0165] ;

[0166] Where: represents the convolution matrix, represents the convolution process, represents the activation function, Indicates that the question was not answered The fused semantic feature vector of

[0167] S54: Calculate the fusion semantic feature vector Attention distribution:

[0168] ;

[0169] Where: Represents the fused semantic feature vector The attention distribution of , T represents transposition, represents the attention mapping matrix;

[0170] S55: Calculate the number of questions that students have not answered The answer tendency of

[0171] ;

[0172] ;

[0173] Where: Indicates that the student did not answer the question The answer tendency of Represents the fused semantic feature vector The corresponding attention weights;

[0174] S56: Count the students' answering tendencies for all unanswered questions, and select the first several unanswered questions from high to low according to the answering tendencies to form a personalized question practice plan.

[0175] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for generating a personalized question practice plan by intelligently integrating wrong questions, characterized in that: The method comprises the following steps: S1: Collect students’ wrong question text data and perform word segmentation and vector encoding processing, and divide the wrong question text data into different wrong question categories through adaptive clustering; including: S11: Collect students’ wrong questions text data: ； Where: represents the wrong question text data of the student's k-th wrong answer, It indicates the total number of questions that students answered incorrectly; S12: Perform word segmentation on the wrong question text data, where the wrong question text data The segmentation result is: ； Where: Indicates wrong question text data The i-th word segmentation result, Indicates wrong question text data The total number of word segmentation results; The word segmentation results of the wrong question text data are encoded by word vectors, and the encoding vectors of the wrong question text data are constructed based on the word vector encoding, where the wrong question text data The corresponding encoding vector is: ； Where: Indicates wrong question text data The corresponding encoding vector, Indicates word segmentation results The word vector encoding of ; S13: Calculate the distance between any two wrong question text data, where the wrong question text data and The distance between them is: ； Where: Indicates wrong question text data and The distance between , represents the L2 norm, represents the feature transformation matrix; S14: Use K-means clustering algorithm to cluster K wrong question text data and obtain U cluster centers, where the uth cluster center is , , cluster center The corresponding encoding vector is ; S15: Filter the misclassified wrong question text data adjacent to multiple cluster centers from the students' wrong question text data, where the set of misclassified wrong question text data is: ； Where: represents the mth misclassified wrong question text data obtained by screening, and M represents the total number of misclassified wrong question text data obtained by screening; S16: Based on manual judgment, the misclassified wrong question text data is classified into the cluster where the correct cluster center is located, and the misclassification association information matrix is ​​constructed: ； Where: Q represents the misclassification association information matrix, is the element value of the element in the kth row and eth column in the misclassified association information matrix, representing the wrong question text data and If the misclassification association information between and Both are misclassified wrong question text data, and they are manually judged to be in the same cluster. ,otherwise ; S17: Reconstruct the encoding vector of the wrong question text data based on the misclassification association information matrix: ； Where: T represents transpose, Represents the encoding vector The reconstruction result of S18: Recalculate the distance between wrong question text data based on the reconstructed encoding vector, cluster K wrong question text data using K-means clustering algorithm, and obtain U cluster centers and cluster clusters. The wrong question categories of the wrong question text data in each cluster are the same, and the wrong question text data The error category is , , correspond Types of wrong questions; S2: Extract knowledge points based on the co-occurrence information of the word segmentation results of the wrong question text data in different wrong question categories, and obtain the knowledge points involved in different wrong question text data; S3: Combine the knowledge points and difficulty statistics of the wrong question text data to calculate the similarity of the knowledge points between the wrong question text data, and construct a wrong question knowledge point similarity matrix; S4: Based on the students’ historical answer data and the students’ unanswered questions in the question bank, a question answering tendency matrix is ​​constructed and matrix decomposition is performed to obtain the semantic feature vector of each question; S5: Construct a personalized question practice plan generation model, which takes the semantic feature vector of each question and the similarity matrix of wrong question knowledge points as input, integrates graph convolution and attention mechanism to calculate the students' answering tendency on different questions, and selects the first several unanswered questions from high to low according to the answering tendency to generate a personalized question practice plan.

2. The method for generating a personalized question practice plan by intelligently integrating wrong questions according to claim 1, characterized in that: The step S2 comprises: S21: Calculate wrong question text data The co-occurrence information of the word segmentation results in The co-occurrence information is: ； Where: Indicates wrong question text data Chinese word segmentation results The co-occurrence information of express A set of word segmentation results of wrong question text data, Represents any word segmentation result in the word segmentation result set; Indicates word segmentation results In the wrong category The number of co-occurrences in the word segmentation results In the wrong question text data that appears at the same time, the wrong question category is The number of wrong question text data; Indicates word segmentation results The number of co-occurrences in the remaining wrong question categories, that is, in the word segmentation results In the wrong question text data that appears at the same time, the wrong question category is not The number of wrong question text data; Indicates word segmentation results The adjacent word set of The set of other word segmentation results that appear in the same wrong question text data; Indicates any word segmentation result The set of other word segmentation results that appear in the same wrong question text data; Indicates the number of word segmentation results in the calculation set; S22: Select the E word segmentation results with the largest co-occurrence information as the wrong question text data Knowledge points: ； Where: Indicates wrong question text data The Eth knowledge point.

3. The method for generating a personalized question practice plan by intelligently integrating wrong questions according to claim 2, characterized in that: The step S3 comprises: S31: Obtaining difficulty statistics of wrong question text data, i.e., wrong question text data Difficulty of the corresponding question ; S32: Calculate wrong question text data and Similarity between: ； Where: Indicates wrong question text data and The similarity between represents an exponential function with a natural constant as base, represents the L1 norm, Representing knowledge points The word vector encoding of ; S33: Based on the similarity of knowledge points between wrong question text data, a wrong question knowledge point similarity matrix is ​​constructed.

4. The method for generating a personalized question practice plan by intelligently integrating wrong questions according to claim 3, characterized in that: The step S33 comprises: S331: Extract all knowledge points of wrong question text data and remove duplicates: ； Where: represents the sth knowledge point after deduplication, and S represents the total number of knowledge points after deduplication; S332: Calculate the similarity of knowledge points after deduplication and construct a similarity matrix of wrong question knowledge points: ； Where: Z represents the similarity matrix of wrong question knowledge points, Represents the element value of the element in the sth row and rth column of the wrong question knowledge point similarity matrix Z, corresponding to the knowledge point The wrong question text data and the knowledge points included The mean similarity between the wrong question text data, .

5. The method for generating a personalized question practice plan by intelligently integrating wrong questions according to claim 4, characterized in that: The step S4 comprises: S41: Obtain the student's historical answer data and the student's unanswered question data in the question bank: ； Where: Represents the student's historical answer data. represents the j-th history question data answered by the student, Indicates the total number of questions answered by students. ; Represents the data of students’ unanswered questions in the question bank. It represents the data of the pth question that the student has not answered in the question bank. Indicates the total number of questions in the question bank that students have not answered. ; S42: Construct a corresponding answer tendency matrix for each question data, where the historical answer question data The answer tendency matrix is: ； Where: Indicates historical answer question data The answer tendency matrix, Indicates the data of students answering questions in history The answer situation, Indicates the data of historical questions answered by students , Indicates the data of students' wrong answers to historical questions , Indicates historical answer question data The knowledge point information of the corresponding question text data, Indicates historical answer question data The corresponding question text data contains knowledge points , Indicates historical answer question data The corresponding question text data does not contain any knowledge point ; Unanswered Questions Data The answer tendency matrix is: ； Where: Indicates data for unanswered questions The answer tendency matrix, Indicates data for unanswered questions The knowledge point information of the corresponding question text data, Indicates unanswered questions data The corresponding question text data contains knowledge points , Indicates unanswered questions data The corresponding question text data does not contain any knowledge point ; S43: Perform matrix decomposition on the answer tendency matrix to obtain the semantic feature vector of each question, where the historical answer question data The semantic feature vector of , unanswered questions data The semantic feature vector of .

6. The method for generating a personalized question practice plan by intelligently integrating wrong questions according to claim 5, characterized in that: The step S43 comprises: Question answering tendency matrix Perform matrix decomposition and calculate ; right Perform eigendecomposition and select the eigenvector corresponding to the maximum eigenvalue as the answer tendency matrix The decomposition vector is used as the question data The semantic feature vector of .

7. The method for generating a personalized question practice plan by intelligently integrating wrong questions according to claim 5, characterized in that: The step S5 comprises: S51: Build a personalized question practice plan generation model, the model receives unanswered questions The semantic feature vector of Similarity matrix with wrong question knowledge points ; S52: If the knowledge point similarity matrix of the wrong question Element value If it is greater than the preset threshold, then the knowledge point and There are edges between them, and the similarity matrix of wrong question knowledge points is calculated The corresponding degree matrix : ； Where: In the form of a diagonal matrix, Representation and knowledge points The number of associated edges; S53: True or False Question Knowledge Point Similarity Matrix Unanswered questions The semantic feature vector of Perform graph convolution to obtain unanswered questions The fused semantic feature vector: ； Where: represents the convolution matrix, represents the convolution process, represents the activation function, Indicates that the question was not answered The fused semantic feature vector of S54: Calculate the fusion semantic feature vector Attention distribution: ； Where: Represents the fused semantic feature vector The attention distribution of , T represents transposition, represents the attention mapping matrix; S55: Calculate the number of questions that students have not answered The answer tendency of ； ； Where: Indicates that the student did not answer the question The answer tendency of Represents the fused semantic feature vector The corresponding attention weights; S56: Count the students' answering tendencies for all unanswered questions, and select the first several unanswered questions from high to low according to the answering tendencies to form a personalized question practice plan.

8. An electronic device, comprising: A memory storing at least one instruction; Communication interface, to realize communication between electronic devices; and The processor executes the instructions stored in the memory to implement the method for generating a personalized question practice plan by intelligently integrating wrong questions as described in any one of claims 1-7.

9. A computer-readable storage medium, wherein at least one instruction is stored in the computer-readable storage medium, and the at least one instruction is executed by a processor in an electronic device to implement the method for generating a personalized question practice plan by intelligently integrating wrong questions as described in any one of claims 1-7.

Images