Academic analysis method and system based on semantic question and answer sequence clustering

Abstract

The invention discloses an academic analysis method and system based on semantic question and answer sequence clustering, and relates to the technical field of online learning analysis and education, and the method comprises the steps: constructing a multi-dimensional semantic question and answer sequence based on the multi-dimensional data of students; based on the multi-dimensional semantic question and answer sequence, using a Word2Vec technology to perform feature representation learning, and generating a semantic feature vector; clustering by adopting a K-means clustering algorithm based on the generated semantic feature vector to obtain a clustering result so as to identify academic performance features of the student group; and based on the clustering result, analyzing different student groups by adopting a statistical method to obtain an analysis result. According to the method, the limitations of an existing method in the aspects of insufficient heterogeneous data processing, single analysis dimension, incomplete student feature understanding and the like are overcome, and targeted teaching strategies and improvement suggestions can be provided for educators.

Description

Technical Field

[0001] The present invention relates to the field of online learning analysis and educational technology, and in particular to an academic analysis method and system based on semantic question-answer sequence clustering. Background Art

[0002] Due to the lack of face-to-face teaching conditions and enrollment restrictions in online learning environments, the academic performance of different student groups varies significantly. Current research on student academic performance in online learning environments mainly focuses on overall grades or question-related answer scores, while ignoring the analysis of multi-dimensional data such as question-related knowledge concepts and answering behaviors. This limits the in-depth understanding of students' answering process and academic performance, and cannot fully capture students' performance in knowledge concept mastery and answering behavior. When analyzing the academic performance of different student groups, there are two problems: First, most methods only use limited statistical data features and fail to fully consider the knowledge concepts and semantic information contained in the questions. This leads to insufficient depth in existing analysis and makes it difficult to fully reveal students' knowledge application and logical reasoning ability in the process of answering questions. Second, existing methods lack effective combination of structured data and unstructured sequence data, especially when processing complex data (such as semantic question-answering sequences). This deficiency limits the analysis of student groups from multiple perspectives and hinders educators' ability to design targeted teaching strategies.

[0003] Therefore, there is an urgent need for an academic analysis method and system based on semantic question-answer sequence clustering that can improve the accuracy of student academic performance analysis. Summary of the invention

[0004] The purpose of the present invention is to provide an academic analysis method and system based on semantic question and answer sequence clustering, which can more accurately reflect the students' mastery of various knowledge points in the learning process, so as to overcome the shortcomings of the prior art that fail to fully consider the knowledge concepts contained in the questions and the lack of depth in the analysis of semantic information, which limits the analysis of student groups from multiple angles.

[0005] In order to achieve the above object, the present invention provides the following technical solutions: In a first aspect, the present invention provides an academic analysis method based on semantic question-answer sequence clustering, comprising the following steps: Based on the students' multidimensional data, a multidimensional semantic question-answering sequence is constructed; wherein the students' multidimensional data includes the knowledge concept number, the text vector of the knowledge concept, the score of each type of question of the knowledge concept, the total score of the knowledge concept test, the score of each type of question of the final exam and the total score of the final exam; Based on multi-dimensional semantic question-answering sequences, Word2Vec technology is used for feature representation learning to generate semantic feature vectors. By capturing the semantic relationship between words, the association between students’ answers and knowledge concepts is converted into semantic feature vectors, which enhances the expressiveness and usability of the data. Based on the generated semantic feature vectors, K-means clustering algorithm is used to cluster and obtain clustering results to identify the academic performance characteristics of the student group; Based on the clustering results, different student groups were analyzed using statistical methods to obtain analytical results.

[0006] Furthermore, based on the students' multidimensional data, the process of constructing a multidimensional semantic question-answering sequence is as follows: Building a dataset , , , ,in, is a dataset of knowledge concept numbers, is a text vector dataset of knowledge concepts, The scores of each type of questions for knowledge concepts and the scores of each type of questions for the final exam are datasets. This is the total score data set for the knowledge concept test. This is a dataset of the total score of the final exam; among them, the score of each type of question of the knowledge concept and the score of each type of question of the final exam Includes the score vectors for each type of question in the knowledge concept and the score vectors for each type of question in the final exam; Construct a multidimensional semantic question-answering sequence based on the vectors in each constructed data set S , ; The construction order is: add the knowledge concept number vector, the text vector of the current knowledge concept, the text vector of the previous knowledge concept, the score vector of each type of question of the knowledge concept, the total score vector of the knowledge concept test, the score vector of each type of question in the final exam, and the total score vector of the final exam score.

[0007] Furthermore, based on the multidimensional semantic question and answer sequence, Word2Vec technology is used for feature representation learning to generate semantic feature vectors. The continuous bag-of-words method based on the Word2Vec model regards the multidimensional semantic question and answer sequence as a sentence, and converts the sentence into a semantic feature vector representation, which is the semantic feature vector.

[0008] Furthermore, based on the continuous bag-of-words method of the Word2Vec model, the multi-dimensional semantic question-answer sequence is regarded as a sentence, and the process of converting the sentence into a semantic feature vector representation is as follows: Multidimensional semantic question answering sequence Each element in is considered as a target word , and one-hot encode it to generate a one-hot encoded sparse vector, which passes through the weight matrix from the input layer to the hidden layer Generate embedding vectors ; The embedding vector of the context word Aggregate by averaging operation to get the context vector : ,in, is the number of context words; The context vector Mapped to the output layer, through the weight matrix from the hidden layer to the output layer Multiply and calculate the target word Prediction score over the entire vocabulary , ; Using negative sampling optimization, the target word is regarded as a positive sample, and j negative samples are randomly selected, among which the prediction score Including the prediction scores of negative samples And the prediction score of the positive sample ; According to the prediction score of negative samples And the prediction score of the positive sample , calculate the target loss function: , where j is the number of negative samples, It is the sigmoid function; according to the target loss function Calculate the gradient and use stochastic gradient ascent to update the weight matrix from the input layer to the hidden layer and the weight matrix from hidden layer to output layer , get the row vectors of all words in the sequence ; Take the mean of the vectors of all words in the sequence: ,in, for A dense vector representing the entire sequence The semantic feature of is the semantic feature vector.

[0009] Furthermore, based on the generated semantic feature vectors, the K-means clustering algorithm is used to cluster and obtain clustering results to identify the academic performance characteristics of the student group. The specific process is as follows: Calculate the cosine similarity between the semantic feature vectors generated by different students and construct a distance matrix; Perform K-means clustering based on the constructed distance matrix to obtain cluster samples; The silhouette coefficient is used to evaluate the clustering samples and obtain the clustering results.

[0010] Furthermore, by calculating the cohesion and separation of the cluster samples, a silhouette coefficient is obtained, and the range of the silhouette coefficient is -1 to 1.

[0011] Furthermore, different student groups are analyzed based on statistical methods, including: analysis of academic performance of different student groups on different knowledge concepts; analysis of the degree of mastery of knowledge concepts by different student groups; analysis of performance behaviors of different student groups; analysis of the correlation between knowledge concepts based on grades; and analysis of academic performance and learning background of different student groups based on questionnaires.

[0012] Furthermore, the text vector of the knowledge concept is directly generated by a pre-trained BERT model.

[0013] In a second aspect, a system for applying an academic analysis method based on semantic question-answer sequence clustering is provided, including: constructing a sequence unit: constructing a multidimensional semantic question-answer sequence; wherein the multidimensional data of the student includes a knowledge concept number, a text vector of the knowledge concept, a score for each type of question of the knowledge concept, a total score of a knowledge concept test, a score for each type of question of a final exam, and a total score of the final exam results; Vector generation unit: Based on the multi-dimensional semantic question-answer sequence, Word2Vec technology is used to learn feature representation and generate semantic feature vectors; Clustering unit: Based on the generated semantic feature vectors, K-means clustering algorithm is used to perform clustering and obtain clustering results to identify the academic performance characteristics of the student group; Analysis unit: Based on the clustering results, different student groups are analyzed using statistical methods to obtain analytical results.

[0014] Furthermore, the analysis unit conducts analysis on different student groups based on statistical methods, including: analysis of academic performance of different student groups on different knowledge concepts; analysis of the degree of mastery of knowledge concepts by different student groups; analysis of performance behaviors of different student groups; analysis of the correlation between knowledge concepts based on grades; and analysis of academic performance and learning background of different student groups based on questionnaires.

[0015] Compared with the prior art, the present invention has the following beneficial technical effects: The present invention provides an academic analysis method based on semantic question-answer sequence clustering, which constructs a multidimensional semantic question-answer sequence by numbering knowledge concepts, text vectors of knowledge concepts, scores of each type of questions of knowledge concepts, total scores of knowledge concept tests, scores of each type of questions in final exams, and total scores of final exam scores, and uses Word2Vec technology to generate semantic vectors of fixed length, and combines K-means clustering algorithm to group student groups. The method can capture the semantic association between knowledge concepts and scores, and effectively improve the ability to identify the characteristics of student learning groups. In particular, in the analysis of processing multidimensional data, the shortcomings of existing methods in data integration and sequence modeling are overcome. The method can capture the contextual semantic association in the question-answer sequence, and effectively reveal the characteristic differences of student groups in terms of knowledge concept mastery, answering behavior, and learning background; the method overcomes the limitations of existing methods in terms of insufficient heterogeneous data processing, single analysis dimension, and incomplete understanding of student characteristics, and can provide educators with targeted teaching strategies and improvement suggestions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Schematic diagram of the process of the academic analysis method based on semantic question and answer sequence clustering in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0018] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0019] The present invention is based on students' multi-dimensional data, including knowledge concept numbers, corresponding text vectors, scores of various questions, tests and final scores, etc., to form multiple data sets, which are integrated into a multi-dimensional semantic question-answering sequence S in a specific order. By associating students' answer data with knowledge concepts, it can more accurately reflect the students' mastery of various knowledge points in the learning process, wherein the text vectors of knowledge concepts are generated by the pre-trained BERT model. Using the continuous bag-of-words method of Word2Vec, the multi-dimensional semantic question-answering sequence is treated as a sentence to generate a semantic vector of fixed length. This method captures the semantic relationship between words and converts the association between students' answer content and knowledge concepts into a low-dimensional dense feature vector, thereby enhancing the expressive power and usability of the data. For example, the sequence elements are first one-hot encoded to obtain a sparse vector, which is then transmitted from the input layer to the hidden layer weight matrix. Generate embedding vectors , aggregate the context embedding vector to get the context vector m, and add m to the hidden layer to output layer weight matrix Multiply them to get the predicted score u, select the target word as the positive sample, randomly select j negative samples first, and calculate the target loss function based on this , according to the chain derivation rule, calculate the target loss function The weight matrix from the input layer to the hidden layer and the weight matrix from hidden layer to output layer The gradient of , using stochastic gradient ascent, updates the weight matrix Weight Matrix , after continuous iteration, the final weight matrix The row vector of That is, the vector of all words in the sequence, and finally the vector of all words in the sequence is obtained Take the mean and get the semantic feature vector According to the obtained semantic feature vector Clustering is performed to obtain clustering results, thereby identifying the characteristics of student groups with different academic performance. In this way, the student group can be subdivided into multiple different categories, and students in each category may have differences in academic performance, knowledge concept mastery, etc.; based on the clustering results, statistical methods are used to analyze different student groups and obtain analysis results.

[0020] Example 1 See also Figure 1 The present invention provides a method for academic analysis based on semantic question-answer sequence clustering, the method comprising: Based on the students' multidimensional data, a multidimensional semantic question-answering sequence is constructed; among them, the students' multidimensional data includes the knowledge concept number, the text vector of the knowledge concept, the score of each type of question of the knowledge concept, the total score of the knowledge concept test, the score of each type of question in the final exam and the total score of the final exam. Compared with the single-dimensional performance analysis, this method can outline the overall picture of the students' learning situation, capture the knowledge mastery, application ability, and even the subtle performance in the learning process, and accurately locate the students' learning pain points and advantages.

[0021] Constructing a multidimensional semantic question-answering sequence is as follows: Constructing a dataset , , , ,in, is a dataset of knowledge concept numbers, is a text vector dataset of knowledge concepts, The scores of each type of questions for knowledge concepts and the scores of each type of questions for the final exam are datasets. This is the total score data set for the knowledge concept test. The total score dataset for the final exam results; Construct a multidimensional semantic question-answering sequence based on the vectors in each constructed data set S , ; The construction order is: add the knowledge concept number vector, the text vector of the current knowledge concept, the text vector of the previous knowledge concept, the score vector of each type of question of the knowledge concept, the total score vector of the knowledge concept test, the score vector of each type of question of the final exam, and the total score vector of the final exam score. Among them, the text vector of the knowledge concept is directly generated by the pre-trained BERT model.

[0022] It should be noted that, taking a student's multidimensional semantic question-answering sequence as an example, first add the knowledge concept number, then add the text vector of the current knowledge concept, the text vector of the previous knowledge concept, the score of each type of question of the knowledge concept and the total score vector of the knowledge concept test in sequence, construct the above vectors i times, where i is the number of knowledge concepts, and finally add the score vector of each type of question of the final exam and the total score vector of the final exam score to obtain the student's multidimensional semantic question-answering sequence. If the text vector of the current knowledge concept is the text vector of the first knowledge concept, then the text vector of its previous knowledge concept is itself.

[0023] Based on multi-dimensional semantic question-answering sequences, Word2Vec technology is used for feature representation learning to generate semantic feature vectors. The semantic feature vectors generated by Word2Vec technology for feature representation learning are not only low-dimensional representations of data, but more importantly, they retain contextual semantic relationships. In this way, the inherent logic between knowledge concepts will not be lost during analysis, avoiding the one-sided understanding caused by simple numerical analysis, thereby achieving a more accurate and realistic portrayal of students' academic performance.

[0024] In some preferred embodiments of the present invention, based on multidimensional semantic question and answer sequences, Word2Vec technology is used for feature representation learning to generate semantic feature vectors. A continuous bag-of-words method based on the Word2Vec model is used to regard the multidimensional semantic question and answer sequences as sentences, and the sentences are converted into semantic feature vector representations.

[0025] The continuous bag-of-words method based on the Word2Vec model regards the multi-dimensional semantic question-answer sequence as a sentence. The specific process of converting the sentence into a semantic feature vector representation is as follows: Multidimensional semantic question answering sequence Each element in is considered as a target word , and one-hot encode it to generate a one-hot encoded sparse vector, which passes through the weight matrix from the input layer to the hidden layer Generate embedding vectors ; The embedding vector of the context word Aggregate by averaging operation to get the context vector : ,in, is the number of context words; The context vector Mapped to the output layer, through the weight matrix from the hidden layer to the output layer Multiply and calculate the target word Prediction score over the entire vocabulary , ; Using negative sampling optimization, the target word is regarded as a positive sample, and j negative samples are randomly selected, among which the prediction score Including the prediction scores of negative samples And the prediction score of the positive sample ; According to the prediction score of negative samples And the prediction score of the positive sample , calculate the target loss function: , where j is the number of negative samples, It is the sigmoid function; according to the target loss function Calculate the gradient and update the weight matrix using stochastic gradient ascent Weight Matrix , after continuous iteration, the final weight matrix The row vector of That is, the vector of all words in the sequence; Take the mean of the vectors of all words in the sequence: ,in, for A dense vector representing the entire sequence The semantic feature of is the semantic feature vector.

[0026] For example, the present invention adopts a stochastic gradient ascent optimization algorithm to update the model parameters to minimize the loss function.

[0027] First, we use negative sampling technology to calculate the target loss function based on the target word (positive sample) and the scores of j randomly selected negative samples. , and then calculate the loss function The weight matrix from the hidden layer to the output layer The partial derivative of , and then gradually reversely deduce from the output layer according to the chain derivation rule, and then find the target loss function About the weight matrix from input layer to hidden layer The chain rule can decompose the derivative of a complex composite function into the product of multiple simple function derivatives, accurately track the flow of gradients between the layers of the neural network, and use stochastic gradient ascent to update the model parameters.

[0028] Based on the generated semantic feature vectors, the K-means clustering algorithm is used to cluster and obtain clustering results to identify the academic performance characteristics of the student group; the generated semantic feature vectors are low-dimensional and dense vectors; they can capture and retain the contextual semantic relationship between data points and accurately convey the semantic information of the object. It can automatically classify massive student data according to academic performance characteristics and dig out different types of student groups hidden behind the data without manually setting complex rules in advance, which greatly saves labor costs and time costs and improves the objectivity of classification.

[0029] In some preferred embodiments of the present invention, based on the generated semantic feature vectors, a K-means clustering algorithm is used to perform clustering to obtain clustering results to identify the academic performance characteristics of the student group. The specific process is as follows: Calculate the cosine similarity between the semantic feature vectors generated by different students and construct a distance matrix; Perform K-means clustering based on the constructed distance matrix to obtain cluster samples; The silhouette coefficient is used to evaluate the clustering samples and obtain the clustering results. The silhouette coefficient can evaluate the clustering quality, and by calculating the cohesion and separation of the clustering samples, the silhouette coefficient value is obtained, ranging from -1 to 1. Using the silhouette coefficient to evaluate the clustering effect can scientifically measure the cohesion and separation of the clustering samples. Educators can obtain high-quality clustering results, accurately distinguish different groups, and avoid misleading subsequent analysis due to inaccurate clustering.

[0030] Based on the clustering results, statistical methods are used to analyze different student groups and obtain analytical results to provide educators with targeted teaching strategies and improvement suggestions.

[0031] For example, different student groups are analyzed based on statistical methods, including: analysis of the academic performance of different student groups on different knowledge concepts; analysis of the degree of mastery of knowledge concepts by different student groups; analysis of performance behaviors of different student groups; analysis of the correlation between knowledge concepts based on grades; analysis of academic performance and learning background of different student groups based on questionnaires.

[0032] Example 2 The present invention also provides a system for applying the academic analysis method based on semantic question-answer sequence clustering, comprising: Construct sequence unit: construct a multi-dimensional semantic question-answering sequence; wherein the multi-dimensional data of students include knowledge concept number, text vector of knowledge concept, score of each type of question of knowledge concept, total score of knowledge concept test, score of each type of question of final examination and total score of final examination; Vector generation unit: Based on the multi-dimensional semantic question-answer sequence, Word2Vec technology is used to learn feature representation and generate semantic feature vectors; Clustering unit: Based on the generated semantic feature vectors, K-means clustering algorithm is used to perform clustering and obtain clustering results to identify the academic performance characteristics of the student group; Analysis unit: Based on the clustering results, different student groups are analyzed using statistical methods to obtain analytical results.

[0033] For example, the analysis unit analyzes different student groups based on statistical methods, including: analysis of the academic performance of different student groups on different knowledge concepts; analysis of the degree of mastery of knowledge concepts by different student groups; analysis of performance behaviors of different student groups; analysis of the correlation between knowledge concepts based on grades; and analysis of the academic performance and learning background of different student groups based on questionnaires.

[0034] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved.

Claims

1. The academic analysis method based on semantic question-answer sequence clustering is characterized by: The following steps are involved: Based on the students' multidimensional data, a multidimensional semantic question-answering sequence is constructed; wherein the students' multidimensional data includes the knowledge concept number, the text vector of the knowledge concept, the score of each type of question of the knowledge concept, the total score of the knowledge concept test, the score of each type of question of the final exam and the total score of the final exam; Based on multi-dimensional semantic question-answering sequences, Word2Vec technology is used to learn feature representation and generate semantic feature vectors; Based on the generated semantic feature vectors, K-means clustering algorithm is used to cluster and obtain clustering results to identify the academic performance characteristics of the student group; Based on the clustering results, statistical methods were used to analyze different student groups and obtain the analysis results.

2. The academic analysis method based on semantic question-answer sequence clustering according to claim 1 is characterized in that: Based on the students' multidimensional data, the process of constructing a multidimensional semantic question-answering sequence is as follows: Building a dataset , , , ,in, is a dataset of knowledge concept numbers, is a text vector dataset of knowledge concepts, The scores of each type of questions for knowledge concepts and the scores of each type of questions for the final exam are datasets. This is the total score data set for the knowledge concept test. This is a dataset of the total score of the final exam; among them, the score of each type of question of the knowledge concept and the score of each type of question of the final exam Includes the score vectors for each type of question in the knowledge concept and the score vectors for each type of question in the final exam; Construct a multidimensional semantic question-answering sequence based on the vectors in each constructed data set S , ; The construction order is: add the knowledge concept number vector, the text vector of the current knowledge concept, the text vector of the previous knowledge concept, the score vector of each type of question of the knowledge concept, the total score vector of the knowledge concept test, the score vector of each type of question in the final exam, and the total score vector of the final exam results.

3. The academic analysis method based on semantic question-answer sequence clustering according to claim 2 is characterized in that: Based on the multi-dimensional semantic question-answer sequence, Word2Vec technology is used for feature representation learning to generate semantic feature vectors. Based on the continuous bag-of-words method of the Word2Vec model, the multi-dimensional semantic question-answer sequence is regarded as a sentence, and the sentence is converted into a semantic feature vector representation.

4. The academic analysis method based on semantic question-answer sequence clustering according to claim 3 is characterized in that: The continuous bag-of-words method based on the Word2Vec model regards the multi-dimensional semantic question-answer sequence as a sentence. The specific process of converting the sentence into a semantic feature vector representation is as follows: Multidimensional semantic question answering sequence Each element in is considered as a target word , and one-hot encode it to generate a one-hot encoded sparse vector, which passes through the weight matrix from the input layer to the hidden layer Generate embedding vectors ; The embedding vector of the context word Aggregate by averaging operation to get the context vector : ,in, is the number of context words; The context vector Mapped to the output layer, through the weight matrix from the hidden layer to the output layer Multiply and calculate the target word Prediction score over the entire vocabulary , ; Use negative sampling optimization, treat the target word as a positive sample, and randomly select j negative samples, where the prediction score Including the prediction scores of negative samples And the prediction score of the positive sample ; According to the prediction score of negative samples And the prediction score of the positive sample , calculate the target loss function: , where j is the number of negative samples, It is the sigmoid function; according to the target loss function Calculate the gradient and use stochastic gradient ascent to update the weight matrix from the input layer to the hidden layer and the weight matrix from hidden layer to output layer , the final weight matrix The row vector of That is, the vector of all words in the sequence; For all words in the sequence Take the mean: ,in, for A dense vector representing the entire sequence The semantic feature of is the semantic feature vector.

5. The academic analysis method based on semantic question-answer sequence clustering according to claim 1 is characterized in that: Based on the generated semantic feature vectors, the K-means clustering algorithm is used to cluster and obtain clustering results to identify the academic performance characteristics of the student group. The specific process is as follows: Calculate the cosine similarity between the semantic feature vectors generated by different students and construct a distance matrix; Perform K-means clustering based on the constructed distance matrix to obtain cluster samples; The silhouette coefficient is used to evaluate the clustering samples and obtain the clustering results.

6. The academic analysis method based on semantic question-answer sequence clustering according to claim 5 is characterized in that: The silhouette coefficient is obtained by calculating the cohesion and separation of cluster samples, and the range of the silhouette coefficient is -1 to 1.

7. The academic analysis method based on semantic question-answer sequence clustering according to claim 1 is characterized in that: The analysis of different student groups is based on statistical methods, including: analysis of the academic performance of different student groups on different knowledge concepts; analysis of the degree of mastery of knowledge concepts by different student groups; analysis of performance behaviors of different student groups; analysis of the correlation between knowledge concepts based on grades; and analysis of the academic performance and learning background of different student groups based on questionnaires.

8. The academic analysis method based on semantic question-answer sequence clustering according to claim 1 is characterized in that: The text vector of the knowledge concept is directly generated by the pre-trained BERT model.

9. A system for use in the analysis method according to any one of claims 1 to 8, characterized in that: include: Construct sequence unit: construct a multi-dimensional semantic question-answering sequence; wherein the multi-dimensional data of students include knowledge concept number, text vector of knowledge concept, score of each type of question of knowledge concept, total score of knowledge concept test, score of each type of question of final examination and total score of final examination; Vector generation unit: Based on the multi-dimensional semantic question-answer sequence, Word2Vec technology is used to learn feature representation and generate semantic feature vectors; Clustering unit: Based on the generated semantic feature vectors, K-means clustering algorithm is used to perform clustering and obtain clustering results to identify the academic performance characteristics of the student group; Analysis unit: Based on the clustering results, different student groups are analyzed using statistical methods to obtain analytical results.

10. A system for applying a learning analysis method based on semantic question-answer sequence clustering according to claim 9, characterized in that: The analysis unit analyzes different student groups based on statistical methods, including: analysis of the academic performance of different student groups on different knowledge concepts; analysis of the degree of mastery of knowledge concepts by different student groups; analysis of performance behaviors of different student groups; analysis of the correlation between knowledge concepts based on grades; and analysis of the academic performance and learning background of different student groups based on questionnaires.

Images