A method and device for accurately locating weak knowledge points

Abstract

The present application discloses a method and device for accurately locating weak knowledge points. The method comprises: obtaining wrong question set data; using an error cause classification model to classify the wrong question set data into error causes, generating wrong question set data classified by error causes; using a knowledge point association model to associate the wrong question set data classified by error causes with relevant knowledge points, and determining the association between wrong questions and knowledge points in the wrong question set data classified by error causes; based on the association between wrong questions and knowledge points in the wrong question set data classified by error causes, determining weak knowledge points. Through the application, the error causes of wrong questions are identified, and the degree of association between the question or answer text and the knowledge points is measured, so as to determine which knowledge points are the students' weak points. In addition, targeted learning suggestions and guidance can be provided to students to help them strengthen their grasp of weak knowledge points.

Description

Technical Field

[0001] The present application relates to the field of intelligent education, and in particular to a method and device for accurately locating weak knowledge points. Background Art

[0002] In the field of educational technology, with the continuous advancement of information technology, learning assistance products have become an important tool to improve students' learning efficiency and quality. These products can provide students with personalized learning experiences by integrating advanced algorithms and data analysis technologies, including but not limited to intelligent tutoring robots and educational assistant applications. These systems usually have functions such as error analysis, learning trajectory tracking and personalized recommendations, aiming to provide customized learning suggestions and resources based on students' learning situation and needs.

[0003] However, many learning assistance products try to identify weaknesses by analyzing students' wrong questions and learning trajectories, but these methods often fail to dig deep into the hidden problems in students' learning process, such as cognitive biases, improper learning methods, etc. Therefore, although these products can help students and teachers identify learning difficulties to a certain extent, it is still challenging to accurately locate students' weak knowledge points. The breadth and depth of subject knowledge means that students' weaknesses may involve multiple areas, and individual differences and the diversity of learning methods mean that even for the same knowledge point, the reasons for its weakness may vary.

[0004] In summary, although existing learning assistance technologies have played an important role in personalized teaching and improving learning efficiency, their ability to accurately locate students' weak knowledge points still needs to be improved. Therefore, developing an algorithm and device that can accurately identify students' weak knowledge points is of great significance for achieving a more efficient and personalized learning experience. Summary of the invention

[0005] The embodiments of the present application provide an algorithm and a device that can accurately identify students' weak knowledge points, and are used to accurately locate students' weak knowledge points.

[0006] A method for accurately locating weak knowledge points provided in this application includes the following steps:

[0007] Get wrong question set data;

[0008] Using an error cause classification model to perform error cause classification processing on the wrong question set data to generate error cause classified wrong question set data;

[0009] The knowledge point association model is used to associate relevant knowledge points with the wrong question set data classified by the error causes, and the association between the wrong questions and the knowledge points in the wrong question set data classified by the error causes is determined;

[0010] Based on the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes, weak knowledge points are determined.

[0011] Furthermore, the wrong question set data at least includes the student's online wrong question data and offline wrong question data;

[0012] The online wrong question data or the offline wrong question data include at least one of the wrong question title text, the wrong answer text of the wrong question, and the wrong question standard answer text.

[0013] Further, the error cause classification model is used to classify the error causes of the wrong question set data to generate the wrong question set data classified by error causes, specifically including:

[0014] Based on the wrong question set data, retrieve the corresponding BERT subject model in the error cause classification model;

[0015] Classifying the causes of errors in the wrong question set data by using the BERT subject model to generate wrong question set data classified by the causes of errors;

[0016] The classification of reasons for wrong questions includes at least one of a concept understanding error category, a calculation error category, and a question meaning understanding error category.

[0017] Furthermore, the knowledge point association model is used to associate the wrong questions set data classified by the error causes with relevant knowledge points, and the association between the wrong questions and the knowledge points in the wrong questions set data classified by the error causes is determined, specifically including:

[0018] At least associating knowledge points in the wrong question title text and the wrong answer text of the wrong question to generate a first knowledge point association matrix;

[0019] At least associating the wrong question text and the knowledge points in the wrong question text to generate a second knowledge point association matrix;

[0020] Function operations are performed on the first knowledge point association matrix and the second knowledge point association matrix to determine the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes.

[0021] Further, at least the knowledge points in the wrong question title text and the wrong answer text are associated to generate a first knowledge point association matrix, which specifically includes:

[0022] Convert the knowledge points in the wrong question text into the knowledge point one-hot vector K;

[0023] The knowledge point one-hot vector K is processed for similarity with the wrong answer text of the wrong question to generate the first knowledge point association matrix.

[0024] Furthermore, after the knowledge point one-hot vector K is processed for similarity with the wrong answer text, it also includes:

[0025] The knowledge point one-hot vector K and the wrong answer text after similarity processing are normalized to generate the normalized knowledge point one-hot vector K and the wrong answer text;

[0026] The normalized knowledge point one-hot vector K and the wrong answer text of the wrong question are indexed to the M-dimensional all-0 vector with the preset knowledge point M to generate the first knowledge point association matrix.

[0027] Furthermore, after the normalized knowledge point one-hot vector K and the wrong answer text of the wrong question are indexed to the M-dimensional all-0 vector with the preset knowledge point M, the method further includes performing a function operation on the M-dimensional all-0 vector indexed to the preset knowledge point M and the category coefficient of the wrong question reason to generate a first knowledge point association matrix;

[0028] The category coefficients of the reasons for wrong questions include at least a, b, and c, where a, b, and c correspond to the categories of concept understanding error, calculation error, and question meaning understanding error, respectively;

[0029] The relationship between the category coefficients a, b, and c of the reasons for wrong questions is that a is greater than b, and b is greater than c.

[0030] Further, performing a function operation on the first knowledge point association matrix and the second knowledge point association matrix to determine the association between the wrong questions and the knowledge points in the wrong question set data classified by the error causes specifically includes:

[0031] Subtracting the first knowledge point association matrix from the second knowledge point association matrix to generate an intermediate knowledge point matrix;

[0032] The correlation between wrong questions and knowledge points in the wrong question set data classified by error causes is determined by multiplying the N-dimensional all-1 vector with the intermediate knowledge point matrix.

[0033] Furthermore, based on the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes, weak knowledge points are determined, specifically including:

[0034] Normalize the correlation between wrong questions and knowledge points in the wrong question set data;

[0035] When the correlation between wrong questions and knowledge points in the normalized wrong question set data is closer to 1, the weak knowledge points are determined.

[0036] A device for accurately locating weak knowledge points, comprising:

[0037] An acquisition module is used to obtain wrong question set data;

[0038] A first processing module is used to perform error cause classification processing on the wrong question set data using an error cause classification model to generate error cause classified wrong question set data;

[0039] The second processing module is used to associate relevant knowledge points with the wrong question set data classified by error causes by using a knowledge point association model, and determine the association between wrong questions and knowledge points in the wrong question set data classified by error causes;

[0040] The determination module is used to determine the weak knowledge points based on the correlation between the wrong questions and the knowledge points in the wrong question set data classified by the error causes.

[0041] A computer-readable storage medium stores a computer program / instruction, which, when executed by a processor, implements any of the steps of the method for accurately locating weak knowledge points.

[0042] A computer program product includes a computer program / instruction, which, when executed by a processor, implements any of the steps of the method for accurately locating weak knowledge points.

[0043] The beneficial effects of this application are:

[0044] Through the method and device for accurately locating weak knowledge points provided by this application, the causes of wrong questions can be identified, and the degree of correlation between the question or answer text and the knowledge points can be measured, so as to determine which knowledge points are the students' weak points. In addition, targeted learning suggestions and guidance can be provided to students to help them strengthen their grasp of weak knowledge points. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0046] Figure 1 A workflow diagram of a method for accurately locating weak knowledge points provided in an embodiment of the present application;

[0047] Figure 2 A schematic diagram of the structure of a device for accurately locating weak knowledge points provided in an embodiment of the present application; DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present application clearer, the technical solution of the present application will be clearly and completely described below in combination with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present application.

[0049] Please refer to Figure 1 , a method for accurately locating weak knowledge points provided in an embodiment of the present application, specifically comprising the following steps:

[0050] S100: Obtain wrong question set data.

[0051] The specific wrong question set data is mainly obtained by collecting students' online or offline wrong question data. Online wrong question data is directly pulled from the database, and offline wrong question data can be obtained by taking photos and uploading for identification. It should be pointed out that each wrong question in the obtained wrong question set data should at least include the wrong question title text and the wrong answer text of the wrong question. It is understandable that in order to improve the subsequent accurate positioning of weak knowledge points, the online wrong question data can simultaneously pull the wrong question standard answer text when pulling the wrong question, and the offline wrong question data can also upload the correct answer to the corresponding wrong question. That is, the wrong question set data may only include the wrong question title text and the wrong answer text of the wrong question, but the standard answer text of the wrong question can also be uploaded to increase the accuracy of subsequent positioning of weak knowledge points.

[0052] S110: Using an error cause classification model to perform error cause classification processing on the wrong question set data, and generating error cause classified wrong question set data.

[0053] The error cause classification model is a supervised learning model based on the BERT model, which is mainly used to classify students according to their error causes. In this embodiment, the error cause classification model is a pre-trained BERT model and corresponding BERT subject models are provided for different subjects. Specifically, the establishment of the pre-trained BERT subject model is mainly as follows: collecting and organizing previous wrong question set data: combining the wrong question title text and the wrong answer text of the wrong question in the wrong question set data into a text pair, and marking the wrong question with the wrong cause, and classifying the marked wrong question cause and converting it into a digital code; dividing the sorted wrong question set data into a training set and a test set, training the BERT model with the marked wrong question data, and using the classification label of the wrong question cause as a supervisory signal to guide the model on how to classify the wrong question; evaluating the trained BERT model through testing, and calculating the accuracy and recall rate of the model in the task of classifying the cause of the wrong question.

[0054] Furthermore, an error cause classification model is used to classify the causes of errors on the wrong question set data to generate error cause classified wrong question set data, specifically including: retrieving the corresponding BERT subject model in the error cause classification model based on the wrong question set data; classifying the causes of errors on the wrong question set data through the BERT subject model to generate error cause classified wrong question set data; the categories of the error cause classification include at least one of concept understanding error class, calculation error class, and question meaning understanding error class.

[0055] Specifically, the error cause classification model obtains the BERT subject model corresponding to the wrong question set data through manual or automatic identification, and then classifies the error causes of the wrong question set data through the BERT subject model to generate the wrong question set data classified by the error causes.

[0056] It should be noted that the BERT subject model classifies the wrong question set data into the concept understanding error class, calculation error class, and question meaning understanding error class, and also includes sorting the different types of error causes. So that weak knowledge points can be accurately located later. In a preferred embodiment provided by the present application, the concept understanding error class is set as a priority, and then the wrong question set data classified by the error cause is generated.

[0057] S120: Use the knowledge point association model to associate relevant knowledge points with the wrong question set data classified by error causes, and determine the association between the wrong questions and the knowledge points in the wrong question set data classified by error causes.

[0058] The knowledge point relevance model is used to measure the degree of relevance between the wrong question text, wrong answer text, etc. and the knowledge points. This model is also pre-trained, and uses natural language processing technology, combined with knowledge graphs or other knowledge bases in the education field, to analyze the semantic relationship between text and knowledge points, thereby determining which knowledge points are involved in the text and the degree of relevance between these knowledge points and the text.

[0059] Specifically, the model contains a list of all possible knowledge points in the corresponding subject, and these lists can query all possible knowledge point lists according to different subjects and grades. The knowledge points are encoded in one-hot. The model can also be the wrong question set data used for model establishment in S110. Of course, the wrong question set data preferably contains the wrong question standard answer text of the wrong question, and then manually annotate the knowledge points involved in the wrong question title text, wrong question wrong answer text, and wrong question standard answer text in the wrong question set data. For a determined list of knowledge points, assign a unique index to each knowledge point and use a dictionary to map it. For each knowledge point annotation in a sample, convert it into a corresponding one-hot vector. The specific steps are as follows: create an all-zero vector with a length equal to the length of the knowledge point list; find the index of the knowledge point corresponding to the sample according to the dictionary, and change the 0 at the index position in the vector to 1.

[0060] Specifically, the knowledge points involved in the labeled wrong question set data are divided into a training set and a test set, and then trained through a knowledge point association model and tested after training to determine whether the model can identify knowledge points from the wrong question title text, the wrong answer text, and the wrong question standard answer text, and then associate them.

[0061] Furthermore, a knowledge point association model is used to associate relevant knowledge points with the wrong question set data classified by error causes, and the association between the wrong questions and the knowledge points in the wrong question set data classified by error causes is determined, specifically including: at least associating the knowledge points in the wrong question title text and the wrong answer text of the wrong question to generate a first knowledge point association matrix; at least associating the knowledge points in the wrong question title text and the wrong question title text to generate a second knowledge point association matrix; performing function operations on the first knowledge point association matrix and the second knowledge point association matrix to determine the association between the wrong questions and the knowledge points in the wrong question set data classified by error causes.

[0062] Furthermore, at least the knowledge points in the wrong question title text and the wrong answer text are associated to generate a first knowledge point association matrix, specifically including: converting the knowledge points in the wrong question title text into a knowledge point one-hot vector K; performing similarity processing on the knowledge point one-hot vector K and the wrong answer text to generate a first knowledge point association matrix.

[0063] Specifically, the similarity processing of the knowledge point one-hot vector K and the wrong answer text of the wrong question includes: inputting the wrong answer text and the knowledge point vector K into the model to obtain the similarity sim(Answer,K) between the wrong answer text and the question knowledge point vector K. Decompose the knowledge point vector K involved in the wrong question text into one vector for each knowledge point. If the question involves N knowledge points, decompose it into K1, K2...K n,N vectors. Then the wrong answer text Answer and the decomposed knowledge points are input into the model in turn to obtain the similarity between the answer text and each knowledge point sim(Answer,K1), sim(Answer,K2)…sim(Answer,K n ).

[0064] Furthermore, after the knowledge point one-hot vector K and the wrong question and wrong answer text are processed for similarity, it also includes: normalizing the knowledge point one-hot vector K and the wrong question and wrong answer text after the similarity processing to generate the normalized knowledge point one-hot vector K and the wrong question and wrong answer text; indexing the normalized knowledge point one-hot vector K and the wrong question and wrong answer text to the M-dimensional all-0 vector with the preset knowledge point as M, to generate a first knowledge point association matrix.

[0065] Specifically, the similarity of each knowledge point is normalized, and the calculation formula is as follows:

[0066]

[0067] Get the normalized knowledge point associations nsim(Answer,K1), nsim(Answer,K2)…nsim(Answer,K n ).

[0068] The normalized knowledge point one-hot vector K and the wrong answer text of the wrong question are indexed to an M-dimensional all-0 vector with a preset knowledge point M to generate a first knowledge point association matrix, specifically including setting the total number of knowledge points to M, creating an M-dimensional all-0 vector (0,0…,0), filling the vector with the knowledge point index corresponding to the association degree of the N knowledge points involved in the question, and forming an M-dimensional wrong question answer text and knowledge point association degree vector AK(0,0…0,nsim(Answer,K1),nsim(Answer,K2),0…0,nsim(Answer,Kn),0).

[0069] It should be noted that each wrong question in the wrong question set data corresponds to an M-dimensional wrong question answer text and knowledge point association vector, so the wrong question set data can form the final first knowledge point association matrix. Specifically, the first knowledge point association matrix is ​​as follows

[0070]

[0071] Furthermore, after the normalized knowledge point one-hot vector K and the wrong answer text of the wrong question are indexed to the M-dimensional all-0 vector with the preset knowledge point M, the method further includes performing a function operation on the M-dimensional all-0 vector indexed to the preset knowledge point M and the category coefficient of the reason for the wrong question to generate a first knowledge point association matrix. The category coefficient of the reason for the wrong question includes at least a, b, and c, and a, b, and c correspond to the concept understanding error class, calculation error class, and question meaning understanding error class respectively; the relationship between the category coefficients a, b, and c of the reason for the wrong question is that a is greater than b, and b is greater than c.

[0072] After the final first knowledge point association matrix is ​​generated, a function operation is performed with the category coefficient of the reason for the wrong question. It should be understood that if the category coefficient of the reason for the wrong question is not set, the association degree of the error reason to the knowledge point is the same, and by setting different error reason category coefficients, the error reasons that are more important and have a greater impact on the mastery of the knowledge point can be associated, so as to accurately locate the weak knowledge points.

[0073] Similarly, at least the knowledge points in the wrong question title text and the wrong question title text are associated to generate the second knowledge point association matrix, which is the same as the establishment of the first knowledge point association matrix. It should be noted that the first knowledge point association matrix is ​​the association of the wrong question title text and the knowledge points in the wrong answer text of the wrong question, and the second knowledge point association matrix is ​​the association of the wrong question title text and the knowledge points in the wrong question title text. It should be noted that when there is no standard answer text in the wrong question obtained by the wrong question set data, the association can only be performed by the wrong question title text and the knowledge points in the wrong question title text. When the wrong question set data obtains the wrong question, there is a wrong question standard answer text, and the second knowledge point association matrix can be the association of the wrong question title text and the knowledge points in the wrong question standard answer text. It can be understood that the association performed only by the wrong question title text and the knowledge points in the wrong question title text is compared with the association performed by the wrong question title text and the knowledge points in the wrong question standard answer text, and the latter's association degree is more accurate than the former, and it is also more convenient to achieve the purpose of association. In a preferred embodiment provided by the present application, the second knowledge point association matrix preferably associates the knowledge points in the wrong question title text and the wrong question standard answer text.

[0074] Furthermore, function operations are performed on the first knowledge point association matrix and the second knowledge point association matrix to determine the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes, specifically including: subtracting the first knowledge point association matrix from the second knowledge point association matrix to generate an intermediate knowledge point matrix; multiplying the intermediate knowledge point matrix with an N-dimensional all-1 vector to determine the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes.

[0075] Specifically, by subtracting the knowledge point correlation between the wrong question title text and the wrong answer text in the first knowledge point correlation matrix and the knowledge point correlation between the wrong question title text and the wrong answer text in the second knowledge point correlation matrix, the difference in the knowledge point correlation between the wrong answer text and the wrong answer text can be clearly seen, that is, obtained through the intermediate knowledge point matrix. Later, the correlation between the wrong questions and the knowledge points in the wrong question set data classified by the error cause is determined by multiplying the N-dimensional all-1 vector with the intermediate knowledge point matrix.

[0076] S130: Determine weak knowledge points based on the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes.

[0077] Furthermore, the correlation between wrong questions and knowledge points in the wrong question set data is normalized; when the correlation between wrong questions and knowledge points in the wrong question set data after normalization is closer to 1, the weak knowledge point is determined.

[0078] Specifically, after determining the correlation between the wrong questions and the knowledge points in the wrong question set data classified by the error cause, normalization processing is performed again. The specific formula is:

[0079]

[0080] Through this formula, the knowledge point correlation vector value of each wrong question in the wrong question set can be obtained. It should be pointed out that the closer the knowledge point correlation vector is to 0, the better the knowledge point is mastered, and the closer it is to 1, the weaker the knowledge point is. In this way, the weak knowledge points in the wrong question set can be determined.

[0081] Please refer to Figure 2 , a device for accurately locating weak knowledge points provided in an embodiment of the present application, comprising:

[0082] An acquisition module is used to obtain wrong question set data;

[0083] A first processing module is used to perform error cause classification processing on the wrong question set data using an error cause classification model to generate error cause classified wrong question set data;

[0084] The second processing module is used to associate relevant knowledge points with the wrong question set data classified by error causes by using a knowledge point association model, and determine the association between wrong questions and knowledge points in the wrong question set data classified by error causes;

[0085] The determination module is used to determine the weak knowledge points based on the correlation between the wrong questions and the knowledge points in the wrong question set data classified by the error causes.

[0086] A computer-readable storage medium stores a computer program / instruction, which, when executed by a processor, implements any of the steps of the method for accurately locating weak knowledge points.

[0087] A computer program product includes a computer program / instruction, which, when executed by a processor, implements any of the steps of the method for accurately locating weak knowledge points.

[0088] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0089] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0090] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for accurately locating weak knowledge points, characterized in that: The following steps are involved: Get wrong question set data; Using an error cause classification model to perform error cause classification processing on the wrong question set data to generate error cause classified wrong question set data; The knowledge point association model is used to associate relevant knowledge points with the wrong question set data classified by the error causes, and the association between the wrong questions and the knowledge points in the wrong question set data classified by the error causes is determined; Based on the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes, weak knowledge points are determined; The wrong question set data at least includes online wrong question data and offline wrong question data of the student, and the online wrong question data or offline wrong question data includes at least one of the wrong question title text, the wrong answer text of the wrong question, and the standard answer text of the wrong question; The error cause classification model is used to classify the error causes of the wrong question set data to generate the error cause classified wrong question set data, specifically including: retrieving the corresponding BERT subject model in the error cause classification model based on the wrong question set data; classifying the error causes of the wrong question set data by the BERT subject model to generate the error cause classified wrong question set data; the error cause classification category includes at least one of a concept understanding error category, a calculation error category, and a question meaning understanding error category; The knowledge point association model is used to associate relevant knowledge points with the wrong question set data classified by the error cause, and the association between the wrong questions and the knowledge points in the wrong question set data classified by the error cause is determined, specifically including: at least associating the knowledge points in the wrong question title text and the wrong answer text of the wrong question to generate a first knowledge point association matrix; at least associating the wrong question title text and the knowledge points in the wrong question title text to generate a second knowledge point association matrix; performing function operations on the first knowledge point association matrix and the second knowledge point association matrix to determine the association between the wrong questions and the knowledge points in the wrong question set data classified by the error cause; At least associating knowledge points in the wrong question title text and the wrong answer text to generate a first knowledge point association matrix, specifically including: converting the knowledge points in the wrong question title text into a knowledge point one-hot vector K; performing similarity processing on the knowledge point one-hot vector K and the wrong answer text to generate a first knowledge point association matrix; After the knowledge point one-hot vector K and the wrong answer text are processed for similarity, the method further includes: normalizing the knowledge point one-hot vector K and the wrong answer text after the similarity processing to generate the normalized knowledge point one-hot vector K and the wrong answer text; indexing the normalized knowledge point one-hot vector K and the wrong answer text to an M-dimensional all-0 vector with a preset knowledge point as M to generate a first knowledge point association matrix; After the normalized knowledge point one-hot vector K and the wrong answer text of the wrong question are indexed to the M-dimensional all-0 vector with the preset knowledge point M, the method further includes performing a function operation on the M-dimensional all-0 vector indexed to the preset knowledge point M and the category coefficient of the reason for the wrong question to generate a first knowledge point association matrix; the category coefficient of the reason for the wrong question includes at least a, b, and c, and the a, b, and c correspond to the concept understanding error class, the calculation error class, and the question meaning understanding error class respectively; the relationship between the category coefficients a, b, and c of the reason for the wrong question is that a is greater than b, and b is greater than c; The M is the total number of preset knowledge points.

2. The method for accurately locating weak knowledge points according to claim 1, characterized in that: Performing a function operation on the first knowledge point association matrix and the second knowledge point association matrix to determine the association between the wrong questions and the knowledge points in the wrong question set data classified by the error causes specifically includes: Subtracting the first knowledge point association matrix from the second knowledge point association matrix to generate an intermediate knowledge point matrix; The correlation between wrong questions and knowledge points in the wrong question set data classified by error causes is determined by multiplying the N-dimensional all-1 vector with the intermediate knowledge point matrix; The N refers to the number of knowledge points involved in the question.

3. The method for accurately locating weak knowledge points according to claim 1, characterized in that: Based on the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes, weak knowledge points are determined, including: Normalize the correlation between wrong questions and knowledge points in the wrong question set data; When the correlation between wrong questions and knowledge points in the normalized wrong question set data is closer to 1, the weak knowledge points are determined.

4. A device for accurately locating weak knowledge points, characterized in that: Specifically include: An acquisition module is used to obtain wrong question set data; A first processing module is used to perform error cause classification processing on the wrong question set data using an error cause classification model to generate error cause classified wrong question set data; The second processing module is used to associate relevant knowledge points with the wrong question set data classified by error causes by using a knowledge point association model, and determine the association between wrong questions and knowledge points in the wrong question set data classified by error causes; A determination module, for determining weak knowledge points based on the correlation between wrong questions and knowledge points in the wrong question set data classified by error causes; The wrong question set data at least includes online wrong question data and offline wrong question data of the student, and the online wrong question data or offline wrong question data includes at least one of the wrong question title text, the wrong answer text of the wrong question, and the standard answer text of the wrong question; The error cause classification model is used to classify the error causes of the wrong question set data to generate the error cause classified wrong question set data, specifically including: retrieving the corresponding BERT subject model in the error cause classification model based on the wrong question set data; classifying the error causes of the wrong question set data by the BERT subject model to generate the error cause classified wrong question set data; the error cause classification category includes at least one of a concept understanding error category, a calculation error category, and a question meaning understanding error category; The knowledge point association model is used to associate relevant knowledge points with the wrong question set data classified by the error cause, and the association between the wrong questions and the knowledge points in the wrong question set data classified by the error cause is determined, specifically including: at least associating the knowledge points in the wrong question title text and the wrong answer text of the wrong question to generate a first knowledge point association matrix; at least associating the wrong question title text and the knowledge points in the wrong question title text to generate a second knowledge point association matrix; performing function operations on the first knowledge point association matrix and the second knowledge point association matrix to determine the association between the wrong questions and the knowledge points in the wrong question set data classified by the error cause; At least associating knowledge points in the wrong question title text and the wrong answer text to generate a first knowledge point association matrix, specifically including: converting the knowledge points in the wrong question title text into a knowledge point one-hot vector K; performing similarity processing on the knowledge point one-hot vector K and the wrong answer text to generate a first knowledge point association matrix; After the knowledge point one-hot vector K and the wrong answer text are processed for similarity, the method further includes: normalizing the knowledge point one-hot vector K and the wrong answer text after the similarity processing to generate the normalized knowledge point one-hot vector K and the wrong answer text; indexing the normalized knowledge point one-hot vector K and the wrong answer text to an M-dimensional all-0 vector with a preset knowledge point as M to generate a first knowledge point association matrix; After the normalized knowledge point one-hot vector K and the wrong answer text of the wrong question are indexed to the M-dimensional all-0 vector with the preset knowledge point M, the method further includes performing a function operation on the M-dimensional all-0 vector indexed to the preset knowledge point M and the category coefficient of the reason for the wrong question to generate a first knowledge point association matrix; the category coefficient of the reason for the wrong question includes at least a, b, and c, and the a, b, and c correspond to the concept understanding error class, the calculation error class, and the question meaning understanding error class respectively; the relationship between the category coefficients a, b, and c of the reason for the wrong question is that a is greater than b, and b is greater than c; The M is the total number of preset knowledge points.

Images