Auxiliary learning video generation and auxiliary learning method and device based on test question heavy and difficult point error causes

Abstract

The invention discloses an auxiliary learning video generation and auxiliary learning method and device based on test question key and difficulty error causes, and relates to the technical field of intelligent education. According to the method and the device, the heavy difficulty points and the error cause list of the target test question are determined, the personalized video script is generated based on the target test question, the heavy difficulty points and the error cause list, and the personalized video script comprises the basic script for explaining the heavy difficulty points of the target test question and the error cause branch script which is inserted into the basic script and is related to each heavy difficulty point; basic explanation video clips corresponding to the basic script and error cause branch video clips corresponding to the error cause branch scripts are generated respectively, a video file is obtained through integration, the mapping relation between identifications of the error cause branch video clips and error causes is embedded into the video file, and a personalized auxiliary learning video of the target test question is obtained. Targeted explanation contents can be provided for different errors, so that students can make up knowledge vulnerabilities after watching, and the learning assisting efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of smart education technology, and more specifically, to a method and device for generating supplementary learning videos based on the key points, difficulties and error causes of test questions. Background Technology

[0002] At the technological level, the application of artificial intelligence in education has evolved from early question bank matching and automatic grading to personalized content generation and precise diagnosis of learning progress. Currently, supplementary learning video generation technologies are mainly divided into two categories: one is based on teaching and research personnel manually editing scripts and recording explanatory videos. While this method can ensure the accuracy of knowledge point explanations, it suffers from long production cycles, high costs, and difficulty in scaling up to adapt to different students' learning situations. The other category is automatic video generation based on artificial intelligence technologies, such as using natural language processing (NLP) technology to convert knowledge point text into explanatory scripts and combining them with digital human-driven technology to generate videos.

[0003] Existing tutoring video production technologies revolve around generating general knowledge point text scripts and then producing videos. Whether scripts are manually written by curriculum researchers or automatically generated by AI, they fail to incorporate personalized content design based on students' individual learning needs. Script generation focuses solely on explaining knowledge points and general problem-solving steps, without linking to student answer data and error analysis. For example, if student A answers a quadratic equation problem incorrectly due to a formula misremembering, and student B answers the same problem incorrectly due to a calculation oversight, existing scripts will uniformly explain the correct solution, failing to provide customized explanations for the different reasons for each student's error. Consequently, students cannot fill their own knowledge gaps after watching the videos, resulting in limited improvement in learning efficiency. Summary of the Invention

[0004] In view of the above problems, this application proposes to provide a method and device for generating supplementary learning videos based on the key points, difficulties, and error causes of test questions, so as to realize the automatic generation of personalized supplementary learning videos. The specific solution is as follows:

[0005] Firstly, a method for generating supplementary learning videos based on the key points, difficulties, and error causes of test questions is provided, including:

[0006] For the target test questions to be used to generate supplementary learning videos, identify the key points and difficulties of the target test questions and a list of error causes for each key point and difficulty.

[0007] Based on the target test question, the key points and difficulties of the target test question, and the list of error causes, a personalized video script is generated. The personalized video script includes a basic script for explaining each key point and difficulty of the target test question, and one or more error cause branch scripts inserted into the basic script related to each key point and difficulty. Each error cause branch script is used to explain one error cause corresponding to the key point and difficulty.

[0008] Based on the personalized video script, basic explanation video clips corresponding to the basic script and error branch video clips corresponding to each error branch script are generated respectively.

[0009] By integrating the basic explanation video clips and the video clips for each of the error-causing branches, a video file is obtained. The mapping relationship between the identifiers of the error-causing branch video clips and the error causes is embedded in the video file to obtain the supplementary learning video for the target test question.

[0010] In one possible design, in another implementation of the first aspect of the embodiments of this application, the personalized video script further includes: an adaptation question inserted into the base script related to each of the key points and difficulties, the adaptation question being used to examine whether the user has made a mistake corresponding to the key points and difficulties.

[0011] In one possible design, in another implementation of the first aspect of the embodiments of this application, the process of generating a personalized video script based on the target test question, the key points and difficulties of the target test question, and the list of error causes includes:

[0012] A personalized video script is generated based on the target test question, the key points and difficulties of the target test question, and the list of reasons for error, using a configured personalized video script generation model.

[0013] The personalized video script generation model is trained using sample test questions, a list of key points, difficulties, and reasons for errors in the sample test questions, and personalized video scripts corresponding to the sample test questions as sample labels.

[0014] In one possible design, in another implementation of the first aspect of the embodiments of this application, the training process of the personalized video script generation model includes:

[0015] Using the sample test questions, the list of key points, difficulties, and reasons for errors of the sample test questions as training samples, and the personalized video scripts corresponding to the sample test questions as sample labels, the configured base model is fine-tuned by SFT to obtain the SFT model.

[0016] The SFT large model is used to perform multiple inferences on the configured typical test questions, the key points and difficulties of the typical test questions, and the list of reasons for errors, to obtain multiple personalized video scripts for the typical test questions, and to obtain human preference scores for the multiple personalized video scripts.

[0017] Using the typical test questions, the list of key points and difficulties of the typical test questions and the list of reasons for errors as training samples, and personalized video scripts labeled with human preference scores as sample labels, a reinforcement learning strategy is used to perform reinforcement learning on the base model or the SFT model to obtain a personalized video script generation model after reinforcement learning.

[0018] In one possible design, in another implementation of the first aspect of the embodiments of this application, the personalized video script corresponding to the sample test questions is obtained in the following manner:

[0019] For the configured sample test questions, generate basic scripts to explain the key and difficult points of the sample test questions;

[0020] Insert a highlight or annotation before the explanation of each key or difficult point in the basic script;

[0021] Locate the marked positions of each key and difficult point in the basic script, use the current position as a branch splitting node, take the explanation content of the key and difficult points related to the current position in the basic script as the main branch script, and create one or more error cause branch scripts parallel to the main branch script to obtain the personalized video script corresponding to the sample test question; wherein, the one or more error cause branch scripts correspond one-to-one with each error cause of the key and difficult point at the current position, and each error cause branch script is used to explain one error cause of the key and difficult point.

[0022] In one possible design, in another implementation of the first aspect of this application, the process of determining the key points and difficulties of the target test question and the list of error causes corresponding to each key point and difficulty includes:

[0023] The key points and difficulties of the target test question and the list of errors corresponding to each key point and difficulty are determined by the configured key point and difficulty prediction model.

[0024] The prediction model for key points, difficulties, and error causes uses sample test questions as training samples and the list of key points, difficulties, and error causes corresponding to the sample test questions as sample labels for training.

[0025] In one possible design, in another implementation of the first aspect of the embodiments of this application, the training samples for the training process of the key difficulties and error cause prediction model include two types:

[0026] The first type of training samples only contains sample test questions, so the error list in the sample labels corresponding to the first type of training samples is the list of common error reasons for the key and difficult points.

[0027] The second type of training samples contains both sample questions and corresponding sample user answers. Therefore, the error cause list in the sample labels corresponding to the second type of training samples is the actual error cause list of the sample user answers.

[0028] The process of determining the key points and difficulties of the target test questions and the corresponding list of error causes for each key point and difficulty through a key point and difficulty prediction model includes:

[0029] If there are user answers corresponding to the target test question, then the target test question and the corresponding user answers are input into the key points and error cause prediction model to obtain the key points and error cause list output by the model and the error cause list corresponding to each key point.

[0030] If no user answers the target question, the target question is input into the key points and error prediction model to obtain the key points and error list corresponding to each key point.

[0031] Secondly, it provides a supplementary learning method, including:

[0032] In response to a tutoring request from a tutoring requester for a target test question, the tutoring video for the target test question and the actual answer data of the tutoring requester for the target test question are obtained. The tutoring video is obtained using the tutoring video generation method based on the key points, difficulties and error causes of the test question as described in any of the first aspects.

[0033] Identify the key points and difficulties of the target test questions, and the actual reasons for errors in the actual answer data for each of the key points and difficulties;

[0034] Based on the mapping relationship embedded in the supplementary learning video, find the identifier of the error branch video segment corresponding to the actual error cause of each key point and difficulty, and integrate the error branch video segment corresponding to the identifier with the basic explanation video segment in the supplementary learning video to obtain a personalized supplementary learning explanation video.

[0035] Play the personalized tutoring video.

[0036] Thirdly, it provides a supplementary learning method, including:

[0037] In response to a tutoring request from a tutoring requester for a target test question, a tutoring video for the target test question is obtained. The tutoring video is generated using the tutoring video generation method based on the key points and difficulties of the test question in the first aspect, and the tutoring video contains adapted questions related to each key point and difficulty.

[0038] During the playback of the supplementary learning videos, whenever a key or difficult branch node is played, the video is paused and relevant questions related to the current key or difficult point are pushed.

[0039] Obtain the answer result of the tutoring request object for the adapted question;

[0040] If the answer is correct, the basic explanation video segment corresponding to the current key point will continue to play;

[0041] If the answer is incorrect, determine the actual cause of the error in the current key point, find the identifier of the error branch video segment corresponding to the actual cause of the error in the current key point according to the mapping relationship embedded in the tutoring video, and continue playing the error branch video segment corresponding to the identifier.

[0042] Fourthly, an electronic device is provided, comprising: a memory and a processor;

[0043] The memory is used to store programs;

[0044] The processor is configured to execute the program to implement the various steps of the supplementary learning video generation method based on the key points and reasons for errors in test questions as described in any one of the first aspects of this application, or to implement the various steps of the supplementary learning method of the second or third aspect of this application.

[0045] Fifthly, a readable storage medium is provided, on which a computer program is stored, wherein when the computer program is executed by a processor, it implements the various steps of the supplementary learning video generation method based on the key points and reasons for errors in test questions as described in any one of the first aspects of this application, or implements the various steps of the supplementary learning method of the second or third aspects of this application.

[0046] In a sixth aspect, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the steps of the supplementary learning video generation method based on the key points and errors of test questions as described in any of the first aspects of this application, or implements the steps of the supplementary learning method of the second or third aspects of this application.

[0047] Using the above technical solution, this application, for the target test questions for which supplementary learning videos are to be generated, first determines the key points and difficulties of the test questions and the list of errors corresponding to each key point and difficulty. Then, based on the target test questions, key points and difficulties, and the list of errors, a personalized video script is generated. This script includes a basic script for explaining each key point and difficulty of the target test questions, and an error-causing branch script inserted into the basic script related to each key point and difficulty. Each error-causing branch script explains one error corresponding to a key point and difficulty. Basic explanation video segments corresponding to the basic script and error-causing branch video segments corresponding to each error-causing branch script are generated respectively. The basic explanation video segments and each error-causing branch video segment are integrated to obtain a video file. The identification of the error-causing branch video segment and the mapping relationship between the error-causing branch video segment and the error-causing error are embedded in the video file, thereby obtaining a personalized supplementary learning video for the target test questions.

[0048] This case study analyzed students' individualized learning situations for the target test questions, identifying the key points and difficulties of the questions and the reasons for students' errors. The resulting personalized supplementary learning videos included basic explanations of the key points and difficulties, as well as segmented video clips corresponding to common student errors. By embedding a mapping relationship between error-causing segment identifiers and error causes, subsequent video playback can selectively play the corresponding segment based on the student's performance (whether an error occurred, and the specific cause). Therefore, this case study can generate personalized supplementary learning videos that are more closely matched to students' learning situations. These videos contain segmented video clips corresponding to different error causes related to key points and difficulties, providing targeted explanations for different error causes. This allows students to fill knowledge gaps and improve their learning efficiency after watching the videos. Attached Figure Description

[0049] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0050] Figure 1 A schematic diagram of an implementation system architecture for the auxiliary learning video generation and auxiliary learning method provided in this application embodiment;

[0051] Figure 2 A schematic flowchart of a method for generating supplementary learning videos based on the key points, difficulties, and error causes of test questions, provided in an embodiment of this application;

[0052] Figure 3 A tree structure diagram of a personalized video script provided in an embodiment of this application;

[0053] Figure 4 This is a schematic diagram of a learning assistance method provided in an embodiment of this application;

[0054] Figure 5 This is a schematic diagram of another learning assistance method provided in the embodiments of this application;

[0055] Figure 6 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0057] It is understood that before using the technical solutions disclosed in the various embodiments of this application, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this application in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0058] Existing tutoring video production technologies revolve around generating general knowledge point text scripts and then producing videos. Whether scripts are manually written by curriculum researchers or automatically generated by AI, they fail to incorporate personalized content design based on students' individual learning needs. First, script generation focuses solely on explaining knowledge points and general problem-solving steps, without linking student answer data or error analysis. For example, if student A answers a quadratic equation problem incorrectly due to a formula misremembering, and student B answers the same problem incorrectly due to a calculation oversight, the existing script will uniformly explain the correct solution, failing to provide customized explanations for the different reasons for each student's error. This leaves students unable to address their knowledge gaps even after watching the video. Second, the generated scripts have a fixed text structure, and subsequent video production relies entirely on this script. There are no branching scripts for error analysis; even if students encounter problems while watching, they cannot jump to the corresponding error correction section, lacking dynamic adjustment capabilities. Ultimately, this results in insufficient personalization of the generated tutoring videos, failing to meet the learning needs of different students.

[0059] This paper addresses the aforementioned issues by proposing a method for generating personalized tutoring videos based on the key points, difficulties, and error causes of test questions. This method can be applied to, for example... Figure 1 The system architecture shown may include a terminal 100 and a server 200. The server 200 may include one or more servers (…). Figure 1 (This example uses a server as an illustration).

[0060] Either terminal 100 or server 200 can be used independently to execute the generation of supplementary learning videos based on the key points, difficulties, and error causes of test questions provided in the embodiments of this application. Alternatively, terminal 100 and server 200 can also be used collaboratively to execute the generation of supplementary learning videos based on the key points, difficulties, and error causes of test questions provided in the embodiments of this application.

[0061] The following description Figure 1 The product form of the mid-terminal 100;

[0062] The terminal 100 in this application embodiment can be a mobile phone, tablet computer, teaching large screen, augmented reality (AR) / virtual reality (VR) device, laptop computer, ultra-mobile personal computer (UMPC), netbook, personal digital assistant (PDA), etc., and this application embodiment does not impose any restrictions on it.

[0063] Figure 2 A schematic diagram of an optional hardware structure for terminal 100 is shown.

[0064] This application provides a method for generating supplementary learning videos based on the key points, difficulties, and error causes of test questions. This method can generate supplementary learning videos for target test questions across multiple subjects. Taking the application of this method to a computer device as an example, the computer device can specifically be... Figure 1 The system consists of terminal 100 or a combination of terminal 100 and server 200. (Refer to...) Figure 2 The method for generating supplementary learning videos based on the key points, difficulties, and error causes of test questions specifically includes the following steps:

[0065] Step S100: For the target test questions to be used to generate supplementary learning videos, determine the key points and difficulties of the target test questions and the list of error reasons corresponding to each key point and difficulty.

[0066] This application can generate supplementary learning videos for target test questions in various subjects, such as Chinese, mathematics, physics, and chemistry.

[0067] To generate personalized supplementary learning videos for the target test questions, this step identifies the key points and difficulties of the target test questions, along with a list of common errors for each key point and difficulty. The key points and difficulties of the target test questions refer to the important and challenging knowledge points, which can be selected from the various knowledge points tested in the target test questions. Each key point and difficulty has a corresponding list of common errors, including the reasons why students are prone to making mistakes related to that key point and difficulty. The common errors for different key points and difficulties can be the same or different. A list of common errors for a single key point and difficulty can include one or more common errors, such as incorrect formula recall or calculation oversight.

[0068] In this step, the list of key points, difficulties, and reasons for errors in the target test questions can be determined automatically, for example, through model prediction or rule recognition, or the target test questions themselves can carry the list of key points, difficulties, and reasons for errors, for example, by manual annotation or other methods.

[0069] Step S110: Based on the target test questions, the list of key points, difficulties and error causes of the target test questions, generate personalized video scripts. The personalized video scripts include a basic script for explaining each key point and difficulty of the target test questions, and one or more error cause branch scripts inserted into the basic scripts related to each key point and difficulty. Each error cause branch script is used to explain one error cause corresponding to the key point and difficulty.

[0070] The personalized video script generated in this step includes a base script and error branch scripts inserted into the base script, which are related to each key and difficult point.

[0071] For the target test question, NLP large-scale modeling techniques can be used to transform the text of the key and difficult points corresponding to the target test question into a basic explanation script, which is used to explain the key and difficult points of the target test question. Furthermore, to meet the needs of different students, corresponding error-cause branch scripts are generated for each error-cause of a key or difficult point. Each error-cause branch script explains one error-cause of the key or difficult point. The number of error-cause branch scripts related to a key or difficult point is the same as the number of errors in the error-cause list of the key or difficult point. Taking key or difficult point A as an example, if its error-cause list includes s1 errors, then corresponding error-cause branch scripts are generated for each of the s1 errors, resulting in s1 error-cause branch scripts related to key or difficult point A.

[0072] Personalized video scripts can be represented using a tree-like branching structure, such as... Figure 3 As shown, the basic script serves as the main body. Within the basic script, at each point explaining a key or difficult point, branch scripts detailing the causes of errors related to that key or difficult point are inserted. For example... Figure 3 As shown, for each key and difficult point, multiple script branches are set up in parallel, including the basic script as the trunk, and several error-causing branch scripts parallel to the basic script.

[0073] Optionally, a unique identifier can be assigned to each error branch script.

[0074] Compared to existing video scripts with fixed text structures that only cover general knowledge points and explanation steps and cannot provide customized content for different error causes, this case proposes a structured framework of a basic script covering the key and difficult points plus error cause branch scripts for the key and difficult points. This ensures that personalized video scripts can include customized content such as error cause analysis and correction steps, adapting to students' individual needs.

[0075] Step S120: Based on the personalized video script, generate basic explanation video clips corresponding to the basic script, and error branch video clips corresponding to each error branch script.

[0076] Specifically, video generation technologies (such as manual recording, digital human-driven technology, animation production technology, etc.) can be used to generate basic explanation video clips corresponding to the basic script, as well as error branch video clips corresponding to each error branch script.

[0077] Furthermore, a unique identifier is assigned to each error-causing branch video segment. In one optional example, the identifier of the error-causing branch script can be used as the identifier of the corresponding generated error-causing branch video segment.

[0078] Step S130: Integrate the basic explanation video clips and the video clips of each error cause branch to obtain a video file. Embed the identifiers of the error cause branch video clips and the mapping relationship between the error causes in the video file to obtain the supplementary learning video for the target test question.

[0079] Specifically, video editing techniques can be used to integrate the basic explanation video clips and the video clips for each error branch to obtain a video file. Optionally, the basic explanation video clips and the video clips for each error branch can be integrated according to the tree-like branch structure of a personalized video script to obtain a video file.

[0080] Furthermore, the mapping relationship between the identifiers of the error-causing branch video segments and the specific error-causing factors is embedded into the video file to obtain personalized supplementary learning videos for the target test questions. This makes it easier to find and play the corresponding error-causing branch video segments based on the actual error-causing factors in the user's answers.

[0081] The method provided in this embodiment analyzes students' personalized learning situations for the target test questions, that is, it determines the key points and difficulties of the target test questions and the reasons for students' mistakes. The generated personalized tutoring videos include basic explanation video segments for the key points and difficulties, as well as branch video segments corresponding to the common mistakes made by students. By embedding the mapping relationship between the error branch video segment identifiers and the error reasons, the corresponding branch video can be selected and played in a targeted manner according to the student's answer (whether there is an error, the specific reason for the error). Therefore, this method can generate personalized tutoring videos that are more matched to students' learning situations. It contains error branch video segments corresponding to different reasons for students' mistakes on key points and difficulties, and can provide targeted explanations for different reasons for mistakes, so that students can fill their knowledge gaps and improve tutoring efficiency after watching.

[0082] Using the method described in this application, personalized supplementary learning videos for target videos can be generated in batches.

[0083] In some embodiments of this application, a model-based prediction scheme is provided for the process of determining the key points and difficulties of the target test questions and the list of error causes corresponding to each key point and difficulty in step S100.

[0084] Specifically, this application can use a configured key point and error prediction model to determine the key points and difficulties of the target test questions and the list of error causes corresponding to each key point and difficulty.

[0085] The prediction model for key points, difficulties, and error causes uses sample test questions as training samples and lists of key points, difficulties, and error causes corresponding to the sample test questions as sample labels for training.

[0086] In some embodiments of this application, a multi-dimensional error cause classification system can be constructed to ensure that error causes cover multiple dimensions, for example:

[0087] Knowledge level: errors in formula memorization, confusion of concepts, omissions of knowledge points (such as not mastering a certain derivation step), etc.

[0088] Ability level: Misunderstanding of question conditions, logical reasoning errors (such as errors caused by skipping steps);

[0089] Habitual aspects: Calculation oversights (such as incorrect symbols or numerical calculations), non-standard answer formats, etc.

[0090] This application can assign a unique label to each error cause (such as K001 - formula memory error, A001 - condition comprehension deviation), forming an error cause labeling system to ensure the consistency of subsequent model training.

[0091] The prediction model for key difficulties and error causes can be a pre-trained large model with strong text understanding and logical reasoning capabilities, such as GPT-4, Llama 3, Qwen3, etc. No detailed constraints are imposed in this case.

[0092] One possible training method for the key points, difficulties, and error cause prediction model includes the following two stages:

[0093] Pre-training phase: Use general educational text data (such as textbooks and supplementary teaching materials) to pre-train the basic large model so that it can master the subject knowledge system;

[0094] Fine-tuning phase: The pre-trained model is fine-tuned using the constructed training samples and sample labels. The loss function can be optimized using cross-entropy loss or other loss functions; no detailed constraints are specified in this case.

[0095] The prediction model for key difficulties and error causes trained using the above method can be evaluated using the F1 score. The required accuracy for predicting key difficulties is ≥ P1, and the accuracy for classifying error causes is ≥ P2. If these standards are not met, the amount of training data should be increased or the model parameters adjusted for further training. P1 can be 95% or other values, and P2 can be 80% or other values.

[0096] The training samples and sample labels used when training the prediction model for key points, difficulties, and error causes can be selected from the corpus data.

[0097] Specifically, the data is obtained from the corpus, which includes test question data and student response data.

[0098] The test question corpus includes the question stem, standard answer, key points and difficulties, difficulty level, and score. The student response corpus includes the response content, response result (correct / incorrect), error reasoning, and response time.

[0099] The marking of key and difficult points in the test question corpus can be done manually by the user. In addition, this application provides a method for analyzing and quantifying key and difficult points. For the obtained test question corpus, which may include the knowledge points being tested, the method provided in this application can be used to quantify and analyze the score of key and difficult points for each knowledge point being tested, and thus determine the key and difficult points from each knowledge point being tested.

[0100] This embodiment provides a quantitative analysis strategy for key and difficult points. The importance of a knowledge point is determined by its frequency of examination in test questions (e.g., the number of times it appears in the middle school entrance examinations in the past 3 years) and its percentage of marks (e.g., the percentage of the total marks for that knowledge point in the test paper). Difficulty is determined by the students' error rate (number of incorrect answers / total number of students who answered). Knowledge points with an error rate > 60% are defined as difficult knowledge points. The calculation method is as follows:

[0101] The quantitative index for key and difficult points is constructed as W = α × frequency percentage + β × score percentage + γ × error rate, where α, β, and γ are weight coefficients determined by the analytic hierarchy process, and α + β + γ = 1. The higher the W value, the higher the degree of key and difficult points of the knowledge point.

[0102] Using the above calculation method, the quantitative index W of the key and difficult points of each knowledge point tested in each question in the test question corpus is calculated, and the top n knowledge points (n is generally 1 or 2) with the highest W values ​​are selected as the key and difficult points of the test questions.

[0103] To address the problem that existing technologies do not deeply correlate the key points and difficulties of test questions with the reasons for students' mistakes, and lack the ability to extract personalized learning information from large-scale data, this embodiment constructs a correlation model that integrates quantitative analysis of the key points and difficulties of test questions with the classification of students' mistakes. This model enables accurate prediction of the key points and difficulties of test questions and the high-frequency reasons for students' mistakes, providing data support for the generation of personalized content.

[0104] After obtaining the corpus data, it can be preprocessed. Preprocessing methods include, but are not limited to:

[0105] 1. Data cleaning: Remove duplicate questions, invalid answers (such as invalid answers with a response time of less than 10 seconds), and label incorrect data (such as incorrect reasons that do not match the incorrect answers);

[0106] 2. Data Supplementation and Annotation: For answer data without annotated error reasons, supplementary annotations will be provided using a combination of manual annotation and rule matching. For example, if the incorrect answer is "Formula error" (such as writing the quadratic equation quadratic formula as...),... If the incorrect answer is "incorrect formula memory", it will be marked as "incorrect formula memory"; if the incorrect answer is "incorrect calculation result" (e.g., the steps are correct but the final result is wrong), it will be marked as "incorrect calculation result".

[0107] 3. Data format conversion: Convert the test question data and student answer data into a format that the large model can recognize (such as storing in JSON format, with fields including question_id, knowledge_point, student_answer, error_cause, etc.).

[0108] Based on the preprocessed corpus data, sample test questions are selected as training samples. The list of key points, difficulties and error causes corresponding to the sample test questions is used as sample labels to train the key points, difficulties and error causes prediction model.

[0109] In some embodiments of this application, the training samples for the training process of the key difficulty and error cause prediction model may include two types:

[0110] The first type of training samples only contains sample test questions, so the error list in the sample labels corresponding to the first type of training samples is the list of common error reasons for the key and difficult points.

[0111] The second type of training samples contains both sample questions and corresponding sample user answers. Therefore, the error list in the sample labels corresponding to the second type of training samples is the actual error list of the sample user answers.

[0112] In this embodiment, the user's answers are optional in the input of the prediction model for key points, difficulties, and error causes trained on the test. That is, user answers can be input or not. The process of determining the key points, difficulties, and error causes of the target test question and the corresponding error cause list for each key point and difficulty through the prediction model includes:

[0113] If there are user answers corresponding to the target test question, then input the target test question and the corresponding user answers into the key points, difficulties and error causes prediction model to obtain the key points and difficulties output by the model and the error causes corresponding to each key point and difficulty.

[0114] If no user answers the target question, the target question is input into the key points and error prediction model to obtain the key points and error list for each key point.

[0115] In some embodiments of this application, the personalized video script generated based on the target test question, the key points and difficulties of the target test question, and the list of reasons for error in the aforementioned step S110 is described.

[0116] In this embodiment, the personalized video script includes not only a basic script and one or more error branch scripts related to each key point and difficulty inserted into the basic script, but also: adaptation questions related to each key point and difficulty inserted into the basic script. The adaptation questions are used to examine whether the user has made the error corresponding to the key point and difficulty.

[0117] There are two ways to set up matching questions related to key and difficult points:

[0118] Firstly, corresponding suitable questions are set for each key and difficult point, and the suitable questions can cover all the reasons for errors in testing key and difficult points.

[0119] Secondly, for each error-causing branch script related to each key difficulty, corresponding adaptation questions are set. Each adaptation question for an error-causing branch script can examine the error cause of that branch script. For example, if key difficulty X has three error-causing branch scripts, corresponding to errors 1-3 respectively, then for each error-causing branch script, corresponding adaptation questions are set. Taking the error-causing branch script corresponding to error 1 as an example, the adaptation questions set can examine whether the user has encountered error 1.

[0120] In this embodiment, by adding relevant questions related to each key and difficult point to the personalized video script, the supplementary learning videos generated based on the personalized video script also include these relevant questions. During the playback of the supplementary learning videos, these relevant questions can be output for students to answer. Based on the students' answers, it can be determined whether the students have mastered the current key and difficult points, as well as the specific reasons for their mistakes. Then, video clips corresponding to the error branches can be played, thereby providing personalized video explanations for different students and improving the effectiveness of supplementary learning.

[0121] In some embodiments of this application, step S110 is further described as an optional implementation of generating a personalized video script based on the target test question, the key points and difficulties of the target test question, and the list of reasons for errors.

[0122] In this embodiment, a personalized video script can be generated based on the target test question, the key points and difficulties of the target test question, and the list of reasons for errors, using a configured personalized video script generation model.

[0123] The personalized video script generation model uses sample test questions, a list of key points and difficulties of the sample test questions, and a list of reasons for errors as training samples, and uses the personalized video scripts corresponding to the sample test questions as sample labels for training.

[0124] Among them, the personalized video script generation model can select a pre-trained large model with strong text understanding and logical reasoning capabilities, such as GPT-4, Llama 3, Qwen3, etc.

[0125] For the sample test questions in the training process of the personalized video script generation model, typical test questions can be selected from the corpus data. Typical test questions are those that can cover the core difficulties and high-frequency error causes. The selection criteria are as follows:

[0126] The key and difficult points of the test questions are the frequently tested points in the subject; the historical answer data of the test questions shows ≥3 types of error reasons (ensuring that multi-branch explanation content can be designed); the difficulty coefficient of the test questions is between 0.4 and 0.6 (medium difficulty, which can effectively distinguish students' knowledge mastery).

[0127] Furthermore, if the personalized video script generated by the personalized video script generation model needs to include matching questions related to key and difficult points, then when selecting typical questions, several matching questions can be selected for each key and difficult point of the typical questions.

[0128] Furthermore, the personalized video script tags corresponding to the sample questions in the personalized video script generation model training process can be obtained in the following way:

[0129] S1. Generate basic scripts for explaining the key and difficult points of the sample test questions.

[0130] S2. Insert key points and difficult points annotations before the explanation of each key and difficult point in the basic script.

[0131] For example, before explaining the key and difficult point of the quadratic formula, insert a key and difficult point label: the range of values ​​of the discriminant b-4ac under the square root, to provide a mark for subsequent branching.

[0132] S3. Locate the marked positions of each key and difficult point in the basic script, use the current position as the branch splitting node, take the explanation content of the key and difficult points related to the current position in the basic script as the main branch script, and create one or more error cause branch scripts parallel to the main branch script to obtain the personalized video script corresponding to the sample test questions.

[0133] Among them, one or more error cause branch scripts parallel to the main branch script correspond one-to-one with each error cause of the key and difficult point at the current position, and each error cause branch script is used to explain one error cause of the key and difficult point.

[0134] Each error-causing branch script includes, but is not limited to: error analysis, error case demonstration, correction methods, and comparison with similar questions. For example, for the error-causing factor of "misremembering the formula" in the question about "the formula for finding the roots of a quadratic equation," the branch script content would be:

[0135] i. Error Analysis: Many students remember the quadratic formula as... Forgetting that the denominator is 2a is because the derivation of the formula was not understood;

[0136] ii. Error Case Display: Show students' incorrect answers and mark the location of the errors;

[0137] iii. Correction method: Re-derive the quadratic formula, emphasizing the origin of the denominator 2a;

[0138] iv. Comparison of similar questions): Show one wrong question caused by "incorrect memory of denominator" to reinforce the correction effect.

[0139] When creating more than one error-causing branch script parallel to the main branch script, it can be written manually, generated by machine, or a combination of human and machine methods. For example, the error-causing branch script can be written by calling a large model and then manually verified and modified.

[0140] The personalized video scripts corresponding to the sample test questions can be represented using a tree-like branching structure, with the base script as the trunk and each error reason branch as a child node. The following is an example of a personalized video script in JSON format: [

[0142] {

[0143] "step": "Introduction section",

[0144] "content": "Includes part of the video script"

[0145] },

[0146] {

[0147] "step": "Core Explanation Section"

[0148] "knowleage": "Key and difficult point 1",

[0149] "content": "main branch script",

[0150] "branch": {

[0151] "branch_1-K001": "Error branch script"

[0152] },

[0153] "topic": [

[0154] "The first matching question"

[0155] The second adaptation question, ]

[0157] },

[0158] {

[0159] "step": "Core Explanation Section"

[0160] "knowleage": "Key and difficult points 2",

[0161] "content": "main branch script",

[0162] "branch": {

[0163] "branch_1-K002": "Error branch script"

[0164] },

[0165] "topic": [

[0166] "The first matching question"

[0167] "The second adaptation question" ]

[0169] },

[0170] {

[0171] "step": "Test Application Section",

[0172] "content": "Video script for the test question application section"

[0173] }

[0174] ].

[0175] The personalized video script in the example above consists of three main parts:

[0176] Introduction: Explains the importance of the key and difficult points;

[0177] Core Explanation Section: Step-by-step explanation of key and difficult points (such as formula derivation process, applicable conditions, etc.);

[0178] Application of test questions: Explain the application methods of knowledge points by combining typical test questions (such as "Let's take this question as an example and see how to solve it using the quadratic formula").

[0179] In the core explanation section, a main branch script and a error-causing branch script are designed for each key and difficult point. In the example above, two matching questions are designed for each error-causing branch script. Of course, the number of matching questions can also be set to other values.

[0180] In some embodiments of this application, an optional training method for the personalized video script generation model is further provided, as follows:

[0181] S1. Using sample test questions, a list of key points and difficulties of sample test questions and the reasons for errors as training samples, and personalized video scripts corresponding to sample test questions as sample labels, the configured base model is fine-tuned using SFT to obtain the SFT model.

[0182] The large model after SFT has the ability to generate main branch scripts, error cause branch scripts, and adaptation questions. Among them, adaptation questions are optional. If the personalized video scripts used as sample labels contain adaptation questions, the large model of SFT has the ability to generate adaptation questions. If the personalized video scripts used as sample labels do not contain adaptation questions, the large model of SFT may not have the ability to generate adaptation questions.

[0183] The following is an example of a prompt word that instructs a large model to generate a personalized video script:

[0184] You are a professional teacher with deep expertise in subject teaching, skilled at designing logically clear and highly practical video explanation scripts that combine key and difficult knowledge points with common student errors. Please write a video script based on the complete test questions, key and difficult knowledge points, and corresponding common error causes provided by the user, according to the following requirements:

[0185] Script core structure (must be strictly followed):

[0186] Introduction: Explain the importance of learning the knowledge points by combining their position in the subject system (such as their fundamental role, high-frequency test points, and practical application scenarios) to attract students' attention;

[0187] Core Explanation Section: Breaking down the content by knowledge point, each knowledge point must include three elements:

[0188] Main video script: Breaks down the core principles and problem-solving logic of knowledge points step by step, highlighting key points and difficulties;

[0189] Branch video script: Analyze the common causes of errors for this knowledge point, including "the essence of the common mistakes + avoidance methods + typical counterexamples";

[0190] Recommended questions: 2 basic practice questions that relate to this knowledge point (they should address key points, difficult points, and common causes of errors);

[0191] The test application section combines typical test questions provided by users to explain "knowledge point connections + problem-solving steps + warnings of common mistakes," demonstrating the practical application value of knowledge points.

[0192] User-provided information format:

[0193] Complete test questions: including the question stem, options (if it is a multiple-choice question), answer, and explanation (if there is no explanation, it can be omitted);

[0194] Key points and difficulties and their causes of error: Strictly follow the format {"Knowledge Point 1":["Common Error 1 (Briefly explain the reasons for common mistakes)"],"Knowledge Point 2":["Common Error 2 (Briefly explain the reasons for common mistakes)"]}.

[0195] Output requirements:

[0196] The output must be in JSON format, with the following structure (no fields can be added or removed; field content must be relevant to the actual teaching scenario, using conversational language suitable for video explanations): [{"step": "Introduction","content": "Video script highlighting the importance of the knowledge point (conversational and engaging)"},{"step": "Core Explanation","knowledge": "Knowledge Point 1","content": "Main video script explaining Knowledge Point 1 step-by-step (highlighting key points and difficulties)","branch": {"branch_1-K001": "Branch video script addressing common mistakes in Knowledge Point 1 (including error analysis and avoidance methods)"},"topic": ["First basic exercise for Knowledge Point 1","Second basic exercise for Knowledge Point 1"]},{"step": "Core Explanation","knowledge": "Knowledge Point 2","content": "Main video script explaining Knowledge Point 2 step-by-step (highlighting key points and difficulties)","branch": {"branch_1-K002": "Branch video script addressing Knowledge Point 2"]} Common error branch video scripts (including error analysis and avoidance methods)"},"topic": ["First basic practice question adapted to knowledge point 2","Second basic practice question adapted to knowledge point 2"]},{"step": "Question application section","content": "Video scripts explaining the application of knowledge points based on typical questions provided by users (including problem-solving steps and error warnings)"}).

[0197] S2. Using the SFT large model, perform multiple inferences on the configured typical test questions, the key points and difficulties of the typical test questions, and the list of reasons for errors to obtain multiple personalized video scripts for the typical test questions, and obtain human preference scores for multiple personalized video scripts.

[0198] Specifically, typical test questions can be selected from a pool of sample test questions. These typical test questions, along with a list of their key points, difficulties, and error causes, can be fed into the SFT (Simplified Chinese Logic Calculation) model for multiple inferences to generate multiple personalized video scripts for each typical test question. Each personalized video script is then manually scored based on preferences. One possible scoring method uses metrics including "content matching (relevance of error causes to the script / question), logical coherence (fluency of explanation steps / question expression), and relevance (whether it covers the core of error correction)." Each metric can be assigned a corresponding weight; for example, the weights for the three metrics mentioned above are 0.4, 0.3, and 0.3 respectively. The total score is the weighted sum of the scores for each metric.

[0199] S3. Using typical test questions, the key points and difficulties of typical test questions and the list of reasons for errors as training samples, and personalized video scripts labeled with human preference scores as sample labels, reinforcement learning strategies are used to perform reinforcement learning on the base large model or SFT large model to obtain the personalized video script generation model after reinforcement learning.

[0200] Reinforcement learning can employ algorithms such as PPO and GRPO.

[0201] The method provided in this embodiment, through SFT fine-tuning and reinforcement learning strategies, enables the trained personalized video script generation model to generate personalized video scripts that better match user preferences.

[0202] Based on the supplementary learning video generation method for target test questions based on the key points, difficulties, and error causes of test questions described in any of the foregoing embodiments, this application further provides a supplementary learning method. The supplementary learning method of this embodiment can be applied to intelligent supplementary learning devices such as learning machines, mobile phones, tablets, and smart blackboards. (Refer to...) Figure 4 Tutoring methods may include the following steps:

[0203] Step S200: In response to the tutoring request object's tutoring request for the target test question, obtain the tutoring video for the target test question and the actual answer data of the tutoring request object for the target test question.

[0204] The supplementary learning videos are generated using the supplementary learning video generation method based on the key points, difficulties, and error causes of the test questions, as described in any of the aforementioned embodiments.

[0205] In one optional supplementary learning scenario, a supplementary learning request object (which could be a student or other user) can initiate a supplementary learning request for a target test question. For example, clicking on a supplementary learning video related to the target test question is considered initiating a supplementary learning request. In this case, the supplementary learning video for the target entity and the actual answer data of the supplementary learning request object for the target test question can be obtained.

[0206] In this embodiment, the tutoring requester can upload actual answer data for the target test question, or submit actual answer data for the target test question online.

[0207] Step S210: Determine the key points and difficulties of the target test questions, and the actual reasons for errors in the actual answer data for each of the key points and difficulties.

[0208] In some possible implementations, the key points, difficulties, and error prediction model trained in the aforementioned embodiments can be used to determine the key points and difficulties of the target test question, as well as the actual error causes of the actual answer data for each key point and difficulty. Specifically, the target test question and the actual answer data can be fed into the key points, difficulties, and error prediction model to obtain the key points and difficulties of the target test question output by the model, and the actual error causes of the actual answer data for each key point and difficulty.

[0209] It should be noted that if the actual answer data does not contain any errors regarding a particular key or difficult point, then the corresponding actual reason for the error can be empty.

[0210] Step S220: Based on the mapping relationship embedded in the supplementary learning video, find the identifier of the error branch video segment corresponding to the actual error cause of each key point and difficulty, and integrate the error branch video segment corresponding to the identifier with the basic explanation video segment to obtain a personalized supplementary learning explanation video.

[0211] Specifically, the supplementary learning videos contain the identification of error-causing branch video segments and the mapping relationship between the error-causing factors. In this step, the mapping relationship can be queried to determine the identification of the error-causing branch video segment corresponding to the actual error-causing factor of each key point and difficulty. The error-causing branch video segment corresponding to the identification is then integrated with the basic explanation video segment to obtain a complete personalized supplementary learning explanation video.

[0212] Step S230: Play the personalized tutoring video.

[0213] The supplementary learning method provided in this embodiment requires the requesting learner to provide actual answer data for the target test question before watching the supplementary learning explanation video. By analyzing the actual answer data, the actual reasons for the errors can be determined. Then, the corresponding error branch video segments in the supplementary learning video for the target test question are selected and integrated with the basic explanation video segments to form a complete personalized supplementary learning explanation video. This video can explain the actual reasons for the errors of each key and difficult point that the requesting learner actually encounters, making it more targeted and improving the supplementary learning effect.

[0214] To address the lack of dynamic interactive mechanisms in existing supplementary learning videos, which prevent adjustments to the content based on student responses, this application establishes a dynamic mechanism of user response-error analysis-branch selection. This mechanism automatically matches the optimal explanation branch based on students' real-time responses, allowing for flexible adjustments to the video content.

[0215] Some embodiments of this application further provide another supplementary learning method. This supplementary learning method is based on the supplementary learning video generation method based on the key points and difficulties of test questions and their error causes, as described in the preceding embodiments. It requires that the generated supplementary learning video contains appropriate questions related to each key point and difficulty. Based on this, combined with... Figure 5 As shown, supplementary learning methods may include the following steps:

[0216] Step S300: In response to the tutoring request object's tutoring request for the target test question, obtain the tutoring video for the target test question.

[0217] The supplementary learning videos are generated using the supplementary learning video generation method described in the aforementioned embodiment, and each supplementary learning video contains appropriate questions related to each key and difficult point.

[0218] In the tutoring scenario of this embodiment, the tutoring requester does not need to provide actual answer data before initiating the tutoring request. For example, the tutoring requester can click on the tutoring video related to the target test question on the interactive interface, which is regarded as initiating the tutoring request.

[0219] Step S310: During the playback of the supplementary learning video, whenever a key or difficult branch node is played, pause the video and push the corresponding questions related to the current key or difficult point.

[0220] The supplementary learning videos explain key and difficult points one by one. Each key and difficult point includes a basic explanation video segment and a segment explaining common mistakes. When a branch node of a key or difficult point is played, the video pauses and displays relevant questions to guide the user in answering them. These questions assess the user's understanding of the current key or difficult point.

[0221] Step S320: Obtain the answer results of the tutoring request object for the adapted questions.

[0222] Step S330: If the answer is correct, continue playing the basic explanation video clip corresponding to the current key and difficult points.

[0223] Specifically, if the student who requested tutoring answers the questions correctly, the video clip explaining the basic points and difficulties can continue playing, without needing to play the video clip explaining the reasons for the errors.

[0224] Step S340: If the answer is incorrect, determine the actual cause of the current key point and find the identifier of the error branch video segment corresponding to the actual cause of the current key point and continue playing the error branch video segment corresponding to the identifier.

[0225] If the student requesting tutoring answers incorrectly on the relevant questions related to the current key points and difficulties, the actual cause of the error can be further determined. For example, the key point and error cause prediction model trained using the aforementioned embodiments can be used to determine the actual cause of the error in the answer related to the current key points and difficulties. Alternatively, when the relevant questions are multiple-choice, the actual cause of the error can be pre-set for each incorrect option. Then, when determining that the student has answered incorrectly, the actual cause of the error can be directly determined based on the selected option. This method is simpler and requires less computation.

[0226] After identifying the actual cause of the error, based on the mapping relationship embedded in the supplementary learning video, find the identifier of the error branch video segment corresponding to the actual cause of the current key point and difficulty, and jump to play the error branch video segment corresponding to that identifier.

[0227] The supplementary learning method provided in this embodiment, for scenarios where there are no user responses, can output appropriate questions for each key and difficult point during the playback of supplementary learning videos, interactively guiding the user to answer. Based on the actual answers, it can dynamically select to play basic explanation video segments (if the answers are correct) or video segments corresponding to the error branches (if the answers are incorrect). This allows for adjustment of the explanation content based on the real-time answers of the user, making the supplementary learning videos flexible and more suitable for the actual learning situation of the user, thus improving the supplementary learning effect.

[0228] This application also provides an electronic device in its embodiments. (See reference...) Figure 6 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, tablets, learning machines, large teaching screens, wearable devices, etc. Figure 6 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0229] like Figure 6 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 1, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 2 or a program loaded from a storage device 8 into a random access memory (RAM) 3, to implement the supplementary learning video generation method or supplementary learning method based on the key points and reasons for errors in test questions according to the foregoing embodiments of this application. When the electronic device is powered on, the RAM 3 also stores various programs and data required for the operation of the electronic device. The processing unit 1, ROM 2, and RAM 3 are interconnected via a bus 4. An input / output (I / O) interface 5 is also connected to the bus 4.

[0230] Typically, the following devices can be connected to I / O interface 5: input devices 6 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 7 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 8 including, for example, memory cards, hard drives, etc.; and communication devices 9. Communication device 9 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 6 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0231] This application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are run on an electronic device, the electronic device enables the electronic device to implement any of the supplementary learning video generation methods or supplementary learning methods based on the key points, difficulties, and error causes of test questions provided in this application.

[0232] This application also provides a computer-readable storage medium that carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the supplementary learning video generation methods or supplementary learning methods based on the key points, difficulties, and error causes of test questions provided in this application.

[0233] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0234] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0235] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0236] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0237] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

Claims

1. A method for generating a learning video based on the difficulty of a test question, characterized in that, The method comprises the following steps: determining the key points and the error cause list corresponding to each key point of the target test question; generating an individualized video script based on the target test question, the key points of the target test question, and the error cause list, wherein the individualized video script comprises a basic script for explaining each key point of the target test question, and one or more error cause branch scripts related to each key point inserted in the basic script, each error cause branch script being used to explain one error cause corresponding to the key point; generating a basic explanation video segment corresponding to the basic script and an error cause branch video segment corresponding to each error cause branch script based on the individualized video script; integrating the basic explanation video segment and each error cause branch video segment to obtain a video file, and embedding a mapping relationship between the identification of the error cause branch video segment and the error cause in the video file to obtain the auxiliary learning video of the target test question.

2. The method of claim 1, wherein, The individualized video script further comprises an adaptive question related to each key point inserted in the basic script, which is used to examine whether the user has the error cause corresponding to the key point.

3. The method according to claim 1 or 2, characterized in that, The process of generating an individualized video script based on the target test question, the key points of the target test question, and the error cause list comprises: generating an individualized video script based on the target test question, the key points of the target test question, and the error cause list by using a configured individualized video script generation model; The individualized video script generation model uses sample test questions, key points of the sample test questions, and error cause lists as training samples, and uses individualized video scripts corresponding to the sample test questions as sample labels to obtain.

4. The method of claim 3, wherein, The training process of the individualized video script generation model comprises: using the sample test questions, the key points of the sample test questions, and the error cause lists as training samples, and using the individualized video scripts corresponding to the sample test questions as sample labels to perform SFT fine-tuning on a configured base large model to obtain an SFT large model; using the SFT large model to perform multiple inferences on typical test questions, key points of the typical test questions, and error cause lists to obtain multiple individualized video scripts of the typical test questions, and obtaining artificial preference scores of the multiple individualized video scripts; using the typical test questions, the key points of the typical test questions, and the error cause lists as training samples, and using individualized video scripts with artificial preference scores as sample labels to perform reinforcement learning on the base large model or the SFT large model by using a reinforcement learning strategy to obtain an individualized video script generation model after reinforcement learning.

5. The method of claim 3, wherein, The individualized video script corresponding to the sample test question is obtained in the following way: generating a basic script for explaining each key point of the sample test question; inserting a key point label before the explanation content of each key point in the basic script; positioning each difficult point annotation position in the basic script, taking the current position as a branch splitting node, taking the difficult point explanation content about the current position in the basic script as a main branch script, and creating one or more error cause branch scripts parallel to the main branch script to obtain a personalized video script corresponding to the sample test; wherein the one or more error cause branch scripts correspond to each error cause of the difficult point at the current position one by one, and each error cause branch script is used to explain one error cause of the difficult point.

6. The method of claim 1, wherein, The process of determining the difficult point of the target test and the error cause list corresponding to each difficult point includes: determining the difficult point of the target test and the error cause list corresponding to each difficult point by a configured difficult point and error cause prediction model; The difficult point and error cause prediction model uses a sample test as a training sample and uses the difficult point and error cause list corresponding to the sample test as a sample label to obtain.

7. The method of claim 6, wherein, The training sample of the difficult point and error cause prediction model includes two types: The first type of training sample only contains a sample test, and the error cause list in the sample label corresponding to the first type of training sample is the common error cause list of the difficult point; The second type of training sample contains a sample test and a corresponding sample user answer, and the error cause list in the sample label corresponding to the second type of training sample is the actual error cause list existing in the sample user answer; The process of determining the difficult point of the target test and the error cause list corresponding to each difficult point by the difficult point and error cause prediction model includes: If there is a user answer corresponding to the target test, input the target test and the corresponding user answer into the difficult point and error cause prediction model to obtain the difficult point and the error cause list corresponding to each difficult point output by the model; If there is no user answer corresponding to the target test, input the target test into the difficult point and error cause prediction model to obtain the difficult point and the error cause list corresponding to each difficult point output by the model.

8. A method of assisting learning, characterized by, It includes: In response to the tutoring request of the tutoring request object for the target test, obtaining the tutoring video of the target test and the actual answer data of the tutoring request object for the target test, and the tutoring video is obtained by the tutoring video generation method based on the test difficult point error cause in any one of claims 1-7; determining the difficult point of the target test, and the actual error cause of the actual answer data for each difficult point; According to the mapping relationship embedded in the tutoring video, find the identifier of the error cause branch video segment corresponding to the actual error cause of each difficult point, integrate the error cause branch video segment corresponding to the identifier and the basic explanation video segment in the tutoring video to obtain a personalized tutoring explanation video; playing the personalized tutoring explanation video.

9. A method of assisting learning, characterized by, It includes: In response to the tutoring request of the tutoring request object for the target test, obtaining the tutoring video of the target test, and the tutoring video is obtained by the tutoring video generation method based on the test difficult point error cause in claim 2; During the process of playing the tutoring video, whenever a difficult point branch node is played, pause and push the adaptive test related to the current difficult point; Obtaining an answer result of the auxiliary learning request object to the adaptive question; If the answer result is correct, continue playing the basic explanation video segment corresponding to the current heavy and difficult point; If the answer result is incorrect, determine the actual mistake cause of the current heavy and difficult point, find the identifier of the mistake cause branch video segment corresponding to the actual mistake cause of the current heavy and difficult point according to the mapping relationship embedded in the auxiliary learning video, and continue playing the mistake cause branch video segment corresponding to the identifier.

10. An electronic device, comprising: Comprise: A memory and a processor; The memory is used for storing a program; The processor is used for executing the program to realize each step of the auxiliary learning video generation method based on the mistake cause of the heavy and difficult point of the test question according to any one of claims 1-7, or realize each step of the auxiliary learning method according to claim 8 or 9.

Images