Personalized intelligent tutoring method and system

Abstract

The invention relates to the technical field of education, and discloses a personalized intelligent tutoring system and method. According to the system, personalized intelligent tutoring is realized by constructing a digital resource library and comprehensively analyzing answering conditions of students. According to the technical key points, the method comprises the steps of evaluating test question quality and student learning cognition quality through a student test question answering big data analysis model, and generating a comprehensive analysis result; and through a personalized tutoring decision analysis model, specific reasons of wrong questions of the students are determined based on an analysis result, and corresponding tutoring measures are provided. The system can accurately identify weak links in learning according to double dimensions of test question quality and student learning cognition, formulate a targeted review scheme, and effectively improve the learning effect and review efficiency of students.

Description

Technical Field

[0001] The present application relates to the field of intelligent education, and in particular to the field of personalized learning tutoring technology based on artificial intelligence and big data analysis. Background Art

[0002] With the rapid development of the new generation of information technology, various intelligent tools and systems have gradually been introduced into the field of education to assist teachers in their teaching work and improve students' learning outcomes. As an emerging educational technology, intelligent tutoring systems have been widely used in many schools and educational institutions. Such systems usually evaluate students' knowledge by analyzing their answers, especially the reasons for wrong answers, and develop personalized review plans for students accordingly. This method can help students review more specifically and improve their learning efficiency to a certain extent.

[0003] However, when analyzing the reasons why students get their questions wrong, existing intelligent tutoring systems usually only focus on students' knowledge errors (such as insufficient understanding of knowledge points) or non-knowledge errors (such as carelessness), while ignoring the quality of the test questions themselves. For example, whether the test questions are too difficult, whether they exceed the syllabus, or whether the test points are clearly stated. This single analysis method leads to incomplete analysis results of students' wrong questions, which in turn affects the accuracy and effectiveness of personalized review plans. In addition, when providing tutoring measures, existing systems often only recommend review content based on students' knowledge mastery, but fail to comprehensively consider the relationship between test question quality and student cognitive quality, which may lead to insufficient pertinence of tutoring measures.

[0004] Due to the above shortcomings, the existing intelligent tutoring system has certain limitations in helping students improve their academic performance. Especially when students encounter difficult test questions, the system often cannot correctly identify the quality of the test questions, causing students to repeatedly encounter the same confusion during the review process, affecting their learning efficiency. At the same time, due to the lack of comprehensive analysis of the quality of the test questions and the cognitive level of students, the tutoring measures provided by the system are also difficult to meet the actual needs of students. Therefore, it is urgent to introduce better tutoring systems and methods to further improve students' learning effects. Summary of the invention

[0005] The purpose of this application is to provide a personalized intelligent tutoring method and system to solve the problems raised in the above background technology.

[0006] The present application discloses a personalized intelligent tutoring method, characterized in that it comprises the following steps: Step A: constructing a digital resource library, the digital resource library comprising a knowledge point library, a knowledge point explanation courseware library, a case analysis library and a test question library;

[0007] Step B: constructing a big data analysis model for students' test questions and answers, the big data analysis model for students' test questions and answers includes: a) calculating a test question quality evaluation index A, the test question quality evaluation index A is based on the content rationality, difficulty and discrimination of the test questions in the test question bank; b) calculating a student learning cognition quality evaluation index B, the student learning cognition quality evaluation index B is based on the students' knowledge point cognition degree in answering test questions in previous exams and the degree of learning of knowledge points in daily learning; c) calculating a test question answering big data analysis result Y according to the test question quality evaluation index A and the student learning cognition quality evaluation index B;

[0008] Step C: constructing a personalized tutoring decision analysis model, wherein the personalized tutoring decision analysis model determines the cause of a student's individual wrong answer based on the test answer big data analysis result Y, the test question quality evaluation index A, and the student learning cognitive quality evaluation index B, and selects corresponding tutoring resources from the digital resource library according to the cause determination result;

[0009] Step D: According to the determination result of the personalized tutoring decision analysis model, provide students with personalized review plans: a) When it is determined that the error is caused by poor test question quality, select relevant classic case solutions from the case analysis library, and recommend high-quality test questions from the test question library; b) When it is determined that the error is caused by low student learning cognitive quality, select knowledge point courseware of students' weak knowledge points from the knowledge point explanation courseware library, and match test questions similar to the wrong questions from the test question library; c) When it is determined that the error is caused by poor test question quality and low student learning cognitive quality, simultaneously execute the tutoring measures in steps a) and b) to form multi-level tutoring resources;

[0010] Step E: According to the tutoring measures determined by the personalized tutoring decision analysis model, corresponding tutoring resources are selected from the digital resource library to generate a personalized review plan for the individual student.

[0011] In a preferred embodiment, the method for calculating the test question quality evaluation index A in step B includes:

[0012] A=α 1 a1+α 2 a2+α 3 a3 (1)

[0013] Among them, a1 is the influencing factor of the rationality of the test content, and the calculation formula is:

[0014]

[0015] a2 is the factor affecting the difficulty of the test question, and the calculation formula is:

[0016]

[0017] a3 is the influencing factor of the test question discrimination, and the calculation formula is:

[0018]

[0019] in,

[0020] n1, n2, and n3 are the numbers of knowledge points, secondary knowledge points, and general knowledge points in the test questions respectively;

[0021] X i , X j , X k are the assessment weights of the i, j, and k knowledge points in the test questions respectively;

[0022] Y i , Y j , Y k They are the mastery requirements of the i, j, and k knowledge points in the knowledge point domain in this teaching stage;

[0023] Z i , Z j , Z k are the cognitive levels of the teaching group for the i, j, and k knowledge points in the test questions respectively;

[0024] c i ,d j 、e k Respectively indicate whether the i-th, j-th, and k-th knowledge points are interference knowledge points;

[0025] L g is the score rate of the g-th student on this test question;

[0026] P s′ The average score of this question among the teaching groups with a medium to high comprehensive knowledge and cognition level;

[0027] Q s′ The average score of the test question in the teaching group with a lower-middle range of comprehensive knowledge and cognition level;

[0028] Q r Full marks for the test questions;

[0029] Z fi , Z fj , Z fk are the cognitive levels of the fth student on the i-th, j-th, and k-th knowledge points respectively;

[0030] G f Score the test paper of the fth student;

[0031] m is the number of people in the teaching group;

[0032] α 1 , α2 , α 3 , γ 1 , γ 2 ,δ 1 ,δ 2 ,δ 3 , ε 1 , ε 2 , ε 3 ,∈ 1 ,∈ 2 ,θ 1 ,θ 2 ,θ 3 ,θ 1 ,θ 2 , is the weight coefficient.

[0033] In a preferred embodiment, the calculation method of the student learning cognitive quality evaluation index B in step B includes:

[0034] B=β 1 b1+β 2 b2 (5)

[0035] Among them, b1 is the knowledge point cognition level influencing factor of answering questions in previous examinations, and the calculation formula is:

[0036]

[0037] b2 is the influence factor of knowledge point learning degree in daily learning situation analysis, and the calculation formula is:

[0038]

[0039] in,

[0040] n1, n2, and n3 are the numbers of knowledge points, secondary knowledge points, and general knowledge points in the test questions respectively;

[0041] p′ is the number of questions that the student answered that included the i-th assessment knowledge point;

[0042] S pi It refers to the student's score rate when answering the pth question containing the ith assessment knowledge point;

[0043] q′ is the number of questions that the student answers that contain the jth secondary assessment knowledge point;

[0044] S qj It refers to the student's score rate in answering the qth question containing the jth secondary assessment knowledge point;

[0045] r′ is the number of questions that the student answered that included the kth general knowledge point;

[0046] S rk It refers to the score rate of students answering the rth question containing the kth general knowledge point;

[0047] s′ is the total number of times the student browses / answers the materials / questions containing the i-th assessment knowledge point;

[0048] u si The value of whether the material / question containing the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the writing material / question;

[0049] v si The value of whether the material / question containing the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the listening and speaking material / question;

[0050] t si Refers to the time when the student browsed / answered the material / question containing the i-th assessment knowledge point for the sth time;

[0051] w′ is the total number of times the student browses / answers the materials / questions containing the jth secondary assessment knowledge point;

[0052] u wj The value of whether the material / question containing the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the writing material / question;

[0053] v wj The value of whether the material / question containing the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the listening and speaking material / question;

[0054] t wj Refers to the time when the student browsed / answered the material / question containing the jth secondary assessment knowledge point for the wth time;

[0055] z′ is the total number of times the student browses / answers the materials / questions containing the kth general knowledge point;

[0056] u zk The value of whether the material / question containing the kth general knowledge point browsed / answered by the student for the zth time belongs to the writing material / question;

[0057] v zk The value of whether the material / question containing the kth general knowledge point browsed / answered by the student for the zth time belongs to the listening and speaking material / question;

[0058] t zk Refers to the time when the student browsed / answered the material / question containing the kth general knowledge point for the zth time;

[0059] β 1 , β 2 , μ1 , μ 2 , μ 3 , 1 , 2 , π 1 , π 2 , σ 1 , σ 2 , τ 1 , τ 2 is the weight coefficient.

[0060] In a preferred example, the calculation method of the test question answer big data analysis result Y in step B includes:

[0061] Y=λ 1 A+λ 2 B (8)

[0062] Among them, Y is the result of big data analysis of test answers;

[0063] A is the test question quality assessment index, and its calculation method is as described in claim 2;

[0064] B is the student learning cognitive quality evaluation index, and its calculation method is as described in claim 3;

[0065] λ 1 , 2 is the weight coefficient, which is used to adjust the importance of the test question quality evaluation index and the student learning cognitive quality evaluation index in the final analysis results and meets the following requirements:

[0066] 0<λ 1 <1, 0<λ 2 <1,λ 1 +λ 2 =1;

[0067] The weight coefficient λ 1 , 2 It can be dynamically adjusted according to actual teaching needs to adapt to different teaching scenarios and assessment objectives.

[0068] In a preferred embodiment, the specific method for determining the cause of the error in step C includes:

[0069] (1) Three thresholds are set: Y′ is the threshold of the big data analysis results of the test questions; A′ is the threshold of the test question quality evaluation index; B′ is the threshold of the student learning cognitive quality evaluation index;

[0070] (2) Determine the cause of the error based on the comparison results of Y, A, B and the corresponding thresholds:

[0071] a) When Y>Y′, it is determined that the student made mistakes due to carelessness and no tutoring measures may be taken;

[0072] b) When Y ≤ Y′, further compare A and B with their corresponding thresholds: i) When A < A′ and B ≥ B′, it is determined that the poor quality of the test questions causes the student to answer the questions wrongly; ii) When A ≥ A′ and B < B′, it is determined that the low learning and cognitive quality of the student causes the student to answer the questions wrongly; iii) When A < A′ and B < B′, it is determined that the poor quality of the test questions and the low learning and cognitive quality of the student jointly cause the student to answer the questions wrongly; iv) When A ≥ A′ and B ≥ B′, it is determined that other factors cause the student to answer the questions wrongly, and further analysis or manual intervention may be required;

[0073] Among them, the thresholds Y′, A′, and B′ can be dynamically adjusted according to the actual teaching situation and statistical data to ensure the accuracy and adaptability of the wrong cause determination. Moreover, the wrong cause determination result is used as the basis for subsequent personalized tutoring decisions, for selecting corresponding tutoring resources and formulating tutoring strategies.

[0074] In a preferred example, the method for recommending high-quality test questions in step D includes:

[0075] (1) Calculate the recommendation degree F(x) of high-quality test questions, and its calculation formula is:

[0076]

[0077] Among them, F(x) is the result of the recommendation degree of high-quality test questions;

[0078] F(x) (1) represents the degree to which the high-quality test question contains the assessment knowledge points of this wrong question;

[0079] F(x) (2) represents the content rationality of the high-quality test question;

[0080] F(x) (3) represents the difficulty of the high-quality test question;

[0081] F(x) (4) represents the discrimination degree of the high-quality test question;

[0082] are weight coefficients respectively, and satisfy:

[0083] And

[0084] (2) Calculate the recommendation degree F(x) for all test questions in the test question bank, and sort them in descending order according to the recommendation degree;

[0085] (3) Select the top N test questions with the highest recommendation degree as the recommended high-quality test questions, where N is a configurable parameter, and the default value can be set to 3 or 5;

[0086] (4) When recommending high-quality test questions, the following factors should be considered: a) the knowledge point coverage of the test questions, ensuring that the recommended test questions can cover the main knowledge points tested in the wrong questions; b) the difficulty gradient of the test questions, ensuring that the difficulty of the recommended test questions has a certain progressive relationship; c) the diversity of the test question types, ensuring that the recommended test questions include different types, such as multiple-choice questions, fill-in-the-blank questions, and essay questions;

[0087] (5) Calculate the degree to which the high-quality test questions contain the knowledge points tested in the wrong questions F(x) (1) Method: F(x) (1) = (number of identical assessment knowledge points) / max (number of assessment knowledge points for wrong questions, number of assessment knowledge points for high-quality questions);

[0088] (6) Calculate the content rationality F(x) of high-quality test questions (2) Method: The specific implementation method is the same as formula (2) in this paper;

[0089] (7) Calculate the difficulty of high-quality test questions F(x) (3) Method: The specific implementation method is the same as formula (3) in this paper;

[0090] (8) Calculate the discrimination F(x) of high-quality test questions (4) Method: The specific implementation method is the same as formula (4) in this paper;

[0091] (9) Regularly update the high-quality test question bank and eliminate poor-quality test questions to ensure that the test question bank always contains high-quality test questions;

[0092] (10) Dynamically adjust the weight coefficient based on students’ feedback and learning progress To optimize the effect of the recommendation algorithm.

[0093] Through this method, the most suitable high-quality test questions can be recommended to students, helping them to better master the knowledge points and improve their learning effects.

[0094] In a preferred embodiment, the method for matching similar test questions in step D includes:

[0095] (1) Calculate the matching degree F(y) of similar test questions. The calculation formula is:

[0096] F(y)=ω1*(F(y) (1) ) +ω2*(F(y) (2) )+ω3*(F(y) (3) )+ω4*(F(y) (4) ) (10)

[0097] Among them, F(y) is the matching result of similar test questions;

[0098] F(y)(1) Indicates the matching degree between similar test questions and the wrong question;

[0099] F(y) (2) Indicates the matching degree of the number of knowledge points that the similar test questions and the wrong question contain the same knowledge points;

[0100] F(y) (3) Indicates the matching degree of knowledge point assessment weights for the same knowledge point;

[0101] F(y) (4) Indicates the knowledge point cognitive index matching degree of the same knowledge point;

[0102] ω1, ω2, ω3, ω4 are weight coefficients respectively, and they satisfy:

[0103] 0<ωi<1(i=1,2,3,4), and ω1+ω2+ω3+ω4=1

[0104] (2) Calculate the matching degree F(y) for all test questions in the test question bank and sort them from high to low according to the matching degree;

[0105] (3) Select the first M test questions with the highest matching degree as recommended similar test questions, where M is a configurable parameter and the default value can be set to 5;

[0106] (4) When matching similar test questions, the following factors should be considered: a) Question type matching: give priority to questions of the same or similar type as the wrong questions; b) Knowledge point coverage: ensure that the matching test questions can cover the main knowledge points tested in the wrong questions; c) Difficulty gradient: select test questions that are slightly lower, equivalent to, or slightly higher in difficulty than the wrong questions to form a difficulty gradient; d) Cognitive level: select test questions that are suitable for the students' current learning stage based on their knowledge point cognitive level to ensure that the matching test questions fully cover the students' weak knowledge points;

[0107] (5) Calculate the question type matching degree F(y) (1) Method:

[0108] If the question type is exactly the same, F(y) (1) =1;

[0109] If the question types are similar (such as multiple choice questions and true / false questions), F(y) (1) =0.8;

[0110] If the question types are different but belong to the same category (e.g., all are objective questions), F(y) (1) =0.5;

[0111] If the question type is completely different, F(y) (1) =0.2;

[0112] (6) Calculate the matching degree of the number of knowledge points F(y) (2) Method: F(y) (2) = (number of identical knowledge points) / max (number of wrong knowledge points, number of similar knowledge points)

[0113] (7) Calculate the knowledge point assessment weight matching degree F(y) (3) Method: F(y) (3) =1-|Average assessment weight of wrong questions and knowledge points-Average assessment weight of similar questions and knowledge points|

[0114] (8) Calculate the knowledge point cognitive index matching degree F(y) (4) Method: F(y) (4) =1-|Average cognitive index of wrong questions-Average cognitive index of similar questions|

[0115] (9) According to students’ feedback and learning progress, the weight coefficients ω1, ω2, ω3, and ω4 are dynamically adjusted to optimize the effect of the matching algorithm.

[0116] The present application also discloses a personalized intelligent tutoring system, which is characterized by comprising:

[0117] A digital resource library construction module is used to construct a digital resource library, wherein the digital resource library includes a knowledge point library, a knowledge point explanation courseware library, a case analysis library, and a test question library;

[0118] A student test answer big data analysis model construction module is used to construct a student test answer big data analysis model, and the student test answer big data analysis model includes: a test question quality evaluation index calculation unit, which is used to calculate the test question quality evaluation index A, and the test question quality evaluation index A is based on the content rationality, difficulty and discrimination of the test questions in the test question bank; a student learning cognition quality evaluation index calculation unit is used to calculate the student learning cognition quality evaluation index B, and the student learning cognition quality evaluation index B is based on the knowledge point cognition degree of the students' test answers in previous exams and the learning degree of knowledge points in daily learning; a test answer big data analysis result calculation unit is used to calculate the test answer big data analysis result Y according to the test question quality evaluation index A and the student learning cognition quality evaluation index B;

[0119] A personalized tutoring decision analysis model construction module is used to construct a personalized tutoring decision analysis model. The personalized tutoring decision analysis model determines the cause of a student's personal wrong question based on the test answer big data analysis result Y, the test question quality evaluation index A and the student learning cognitive quality evaluation index B, and selects corresponding tutoring resources from the digital resource library according to the cause determination result;

[0120] A personalized review plan generation module is used to provide students with personalized review plans based on the judgment results of the personalized tutoring decision analysis model, including: when it is determined that the wrong reason is poor test question quality, selecting relevant classic case solutions from the case analysis library, and recommending high-quality test questions from the test question library; when it is determined that the wrong reason is low student learning cognitive quality, selecting knowledge point courseware of the student's weak knowledge points from the knowledge point explanation courseware library, and matching test questions similar to the wrong questions from the test question library; when it is determined that the wrong reason is poor test question quality and the student's learning cognitive quality is low, executing the above two tutoring measures at the same time to form a multi-level tutoring resource;

[0121] The tutoring resource selection module is used to select corresponding tutoring resources from the digital resource library according to the tutoring measures determined by the personalized tutoring decision analysis model, and generate a personalized review plan for individual students.

[0122] The personalized intelligent tutoring system of this application significantly improves the accuracy and effectiveness of test question quality assessment and learning cognitive quality assessment through in-depth analysis of students' wrong questions and reasons. In traditional intelligent tutoring systems, wrong question analysis often only focuses on students' mastery of knowledge points, while ignoring the quality of the test questions themselves. This application innovatively combines the dual analysis methods of test question quality and learning cognitive quality to provide students with more accurate tutoring.

[0123] In terms of question quality analysis, this system can comprehensively identify quality problems in test questions by evaluating multiple dimensions such as the rationality of the test content, difficulty, and discrimination. For example, through analysis, the system can find that some test questions are too difficult, or their assessment points are not clearly expressed, and cannot effectively test students' actual mastery. By identifying these problems, the system can help educators eliminate unqualified test questions and optimize the overall quality of the test question bank, thereby providing students with a more reasonable and scientific learning and testing environment. This precise analysis of the quality of test questions not only improves the effectiveness of the test, but also makes the students' review process more targeted, helping them better master the knowledge they have learned.

[0124] In terms of learning cognitive quality analysis, this application can accurately evaluate students' mastery of each knowledge point through in-depth analysis of students' previous exams and daily learning data. This analysis method can help teachers and students identify weak links in learning, so as to strengthen the review and consolidation of relevant knowledge points in a targeted manner. This precise cognitive analysis avoids students from blindly learning and repeating ineffective exercises during the review process, which significantly improves learning efficiency.

[0125] In addition, this application takes into account both the quality of test questions and the quality of students' learning cognition when formulating tutoring plans for students. The result of this comprehensive analysis enables the system to provide students with more comprehensive and personalized tutoring measures. For example, when the system identifies problems with the quality of test questions, the system will automatically recommend relevant classic case analysis and high-quality test questions to help students better understand and master the knowledge points. On the other hand, when the system finds that students lack cognition on certain knowledge points, it will recommend targeted knowledge point explanation courseware and similar test questions to help students gradually improve their understanding and application of these knowledge points.

[0126] Through this dual-consideration tutoring strategy, this application not only effectively makes up for students' shortcomings in learning, but also ensures that they can fully master various knowledge points during the review process. This comprehensive, multi-level tutoring program can ultimately help students improve their learning outcomes and achieve significant improvements in grades. The application of this system not only optimizes students' learning experience, but also provides teachers with more scientific teaching aids, greatly improving the overall efficiency and quality of education.

[0127] A large number of technical features are recorded in the specification of this application, which are distributed in various technical solutions. If all possible combinations of technical features of this application (i.e., technical solutions) are to be listed, the specification will be too long. In order to avoid this problem, the various technical features disclosed in the above-mentioned invention content of this application, the various technical features disclosed in the various embodiments and examples below, and the various technical features disclosed in the accompanying drawings can be freely combined with each other to form various new technical solutions (these technical solutions are all deemed to have been recorded in this specification), unless the combination of such technical features is technically infeasible. For example, in one example, feature A+B+C is disclosed, and in another example, feature A+B+D+E is disclosed, and features C and D are equivalent technical means that play the same role. Technically, only one can be used, and it is impossible to use them at the same time. Feature E can be combined with feature C technically. Then, the solution of A+B+C+D should not be deemed to have been recorded because it is technically infeasible, and the solution of A+B+C+E should be deemed to have been recorded. BRIEF DESCRIPTION OF THE DRAWINGS

[0128] Figure 1 It is a schematic diagram of constructing a digital resource library in the personalized intelligent tutoring method according to the first embodiment of the present application.

[0129] Figure 2 It is a flowchart of a personalized intelligent tutoring method according to the first embodiment of the present application.

[0130] Figure 3 It is a structural diagram of a personalized intelligent tutoring system according to the second embodiment of the present application. DETAILED DESCRIPTION

[0131] In the following description, many technical details are provided to help readers better understand the present application. However, those skilled in the art can understand that the technical solution claimed in the present application can be implemented even without these technical details and various changes and modifications based on the following embodiments.

[0132] Description of some concepts:

[0133] Personalized intelligent tutoring system: refers to an educational tutoring system based on big data analysis and artificial intelligence technology, which can provide targeted review plans based on students' learning situation and wrong question analysis results, aiming to improve students' learning outcomes.

[0134] Digital resource library: including knowledge point library, knowledge point explanation courseware library, case analysis library and test question library, etc., used to store and manage digital resources related to student learning to support the implementation of personalized tutoring.

[0135] Knowledge point library: a collection of knowledge points extracted and organized according to the subject syllabus. Each knowledge point corresponds to a unique identifier and is associated with attributes such as teaching stage and mastery requirements.

[0136] Knowledge point explanation courseware library: contains multimedia explanation materials for each knowledge point, in the form of text descriptions, pictures, audio and video, etc., to help students better understand and master the knowledge points.

[0137] Case Analysis Library: Collects and categorizes the problem-solving ideas and detailed analysis instructions for various typical test questions, including problem-solving steps and key point explanations, to assist students in understanding and reviewing relevant knowledge points.

[0138] Question bank: Stores and manages various types of test questions, and annotates their attributes (such as question type, test knowledge points, difficulty, discrimination, etc.), supporting the recommendation and generation of test questions in personalized review plans.

[0139] Big data analysis model of students’ test answers: used to comprehensively evaluate the quality of students’ test answers and the quality of their learning cognition. By analyzing the rationality, difficulty, and discrimination of test questions as well as students’ understanding of knowledge points, data analysis results for personalized tutoring decisions are generated.

[0140] Question quality assessment index (A): An indicator to measure whether a question is suitable for teaching testing, obtained through a comprehensive evaluation of factors such as the rationality, difficulty and discrimination of the question content.

[0141] Student learning cognitive quality assessment index (B): An indicator to measure whether students can correctly answer test questions, obtained by analyzing students' knowledge point cognition in previous exams and their knowledge point learning in daily study.

[0142] Personalized tutoring decision analysis model: Based on the big data analysis results of students' test answers, the test quality assessment index and the student learning cognitive quality assessment index, the reasons for students' individual wrong answers are determined, and the corresponding tutoring resources are selected according to the determination results to provide students with personalized review plans.

[0143] High-quality test questions: Select test questions with moderate difficulty, reasonable content and good discrimination from the test question bank to enhance students' review effect.

[0144] Similar test questions: Test questions that are similar to students’ wrong questions in terms of question type, knowledge points and assessment weight, are used to help students further consolidate their weak knowledge points.

[0145] Determination of causes of errors: Based on the big data analysis results of test answers, the test question quality assessment index and the student learning cognitive quality assessment index, the reasons why students made mistakes are analyzed, including poor test question quality, low student learning cognitive quality or both.

[0146] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below in conjunction with the accompanying drawings.

[0147] The first embodiment of the present application relates to a personalized intelligent tutoring method, see Figure 1 and Figure 2 , the method comprises the following steps:

[0148] Step 100: Construct a digital resource library, which includes a knowledge point library, a knowledge point explanation courseware library, a case analysis library, and a test question library.

[0149] Specifically, in step 100 of the present application, the construction method of the digital resource library is the basis of the entire personalized intelligent tutoring system.

[0150] First, the knowledge point library is constructed based on the subject syllabus and examination syllabus, systematically extracting and organizing various knowledge points, and assigning a unique identifier to each knowledge point. The relationship between these knowledge points is also clearly established, including pre-knowledge and subsequent knowledge, and the attributes of each knowledge point are also marked, such as the teaching stage and mastery requirements. In order to ensure that the knowledge point library is always consistent with the latest teaching requirements, the library will be updated regularly.

[0151] Secondly, the construction of the knowledge point explanation courseware library includes the collection of multimedia explanation materials for each knowledge point, such as text descriptions, pictures, audio and video, etc. These explanation courseware are standardized to ensure uniform format and complete content, and corresponding metadata are added, including information such as knowledge points, difficulty level, and applicable learning stage. In order to ensure the quality of the courseware, a quality assessment mechanism has been established, and the courseware content will be regularly reviewed and updated.

[0152] In addition, the construction of the case analysis library collects various typical cases and problem-solving ideas, and classifies them according to knowledge points, question types, difficulty, etc. Detailed instructions are written for each case analysis, including problem-solving steps and key point explanations, and tags are added for quick retrieval and matching. The case library will be updated regularly to add new typical questions and solutions to ensure its timeliness and practicality.

[0153] The construction of the question bank is to collect question bank materials through manual screening, review and batch import, and mark each question, including question type, test knowledge points, difficulty, discrimination and other attributes. In order to ensure the quality of the questions, the system has also established a question quality assessment mechanism to regularly calculate and update the question quality assessment index. At the same time, the question bank sets up question usage frequency statistics to prevent the same question from being used too frequently, and records the modification history of the questions through the version control mechanism.

[0154] The establishment of resource association mechanism is to establish association relationships between different resource libraries, such as associating knowledge points with related courseware, case analysis and test questions. In addition, a resource index is established to support rapid resource location by knowledge points, difficulty, question type and other dimensions, and realize the function of automatically combining different types of resources according to students' learning needs.

[0155] In terms of data security and privacy protection, the system implements strict data encryption and access control measures, and establishes a user rights management system to ensure that the access rights of different roles such as teachers, students, and administrators are reasonably set. In order to ensure the security and reliability of the resource library, the system regularly conducts data backup and recovery drills.

[0156] Finally, the expansion and optimization of the resource library supports the access of third-party educational resources by designing open interfaces. In addition, a resource evaluation mechanism has been established to collect user feedback and continuously optimize resource quality. The system uses machine learning technology to analyze the effect of resource use, automatically adjust resource recommendation strategies, and regularly clean up and optimize the resource library to delete outdated or low-quality resources. In this way, this application can build a comprehensive, high-quality, easy-to-manage and use digital resource library, providing a solid resource foundation for personalized intelligent tutoring systems.

[0157] Step 200: Construct a big data analysis model for students' test answering, the big data analysis model for students' test answering includes: a) calculating a test question quality evaluation index A, the test question quality evaluation index A is based on the content rationality, difficulty and discrimination of the test questions in the test question bank; b) calculating a student learning cognition quality evaluation index B, the student learning cognition quality evaluation index B is based on the students' knowledge point cognition level in answering test questions in previous exams and the learning level of knowledge points in daily learning; c) calculating a test question answering big data analysis result Y based on the test question quality evaluation index A and the student learning cognition quality evaluation index B.

[0158] Optionally, the method for calculating the test question quality assessment index A in step 200 includes:

[0159] A=α 1 a1+α 2 a2+α 3 a3 (1)

[0160] Among them, a1 is the influencing factor of the rationality of the test content, and the calculation formula is:

[0161]

[0162] a2 is the factor affecting the difficulty of the test question, and the calculation formula is:

[0163]

[0164] a3 is the influencing factor of the test question discrimination, and the calculation formula is:

[0165]

[0166] in,

[0167] n1, n2, and n3 are the numbers of knowledge points, secondary knowledge points, and general knowledge points in the test questions respectively;

[0168] X i , X j , X k are the assessment weights of the i, j, and k knowledge points in the test questions respectively;

[0169] Y i , Y j , Y k They are the mastery requirements of the i, j, and k knowledge points in the knowledge point domain in this teaching stage;

[0170] Z i , Z j , Z k are the cognitive levels of the teaching group for the i, j, and k knowledge points in the test questions respectively;

[0171] c i ,d j 、e k Respectively indicate whether the i-th, j-th, and k-th knowledge points are interference knowledge points;

[0172] L g is the score rate of the g-th student on this test question;

[0173] P s′ The average score of this question among the teaching groups with a medium to high comprehensive knowledge and cognition level;

[0174] Q s′ The average score of the test question in the teaching group with a lower-middle range of comprehensive knowledge and cognition level;

[0175] Q r Full marks for the test questions;

[0176] Z fi , Z fj , Z fk are the cognitive levels of the fth student on the i-th, j-th, and k-th knowledge points respectively;

[0177] G f Score the test paper of the fth student;

[0178] m is the number of people in the teaching group;

[0179] α 1 , α 2 , α 3 , γ 1 , γ 2 , δ 1 , δ 2 , δ 3 , ε 1 , ε 2 , ε 3 ,∈ 1 ,∈ 2 ,θ 1 ,θ 2 ,θ 3 ,θ 1 ,θ 2 , is the weight coefficient.

[0180] In this embodiment, P s′ and Q s′ These are two key evaluation indicators used to analyze the discrimination of test questions. The definition and application of these two functions are based on the comprehensive level of students' knowledge cognition, providing an important basis for teaching evaluation and personalized tutoring.

[0181] Specifically, P s′It represents the average score of a particular test question in a teaching group with a medium to high comprehensive knowledge level. For example, this group usually includes students who rank between the top 25% and 50% of the class. Correspondingly, Q s′ It represents the average score on the same test question for a teaching group with a lower-average comprehensive level of knowledge and cognition (usually students ranked between 50% and 75%).

[0182] The calculation process for these two functions first requires ranking the students in the entire class or teaching group based on their overall performance or their performance on a specific test. Then, we take the upper and lower middle groups and calculate their average scores on specific test questions. This approach allows us to focus on the "average" level of students, because this group usually best reflects the appropriate difficulty and discrimination ability of the test questions.

[0183] P s′ and Q s′ The application value of is mainly reflected in three aspects. First, by comparing these two values, the discrimination of the test questions can be evaluated. s′ Significantly higher than Q s′ , indicating that the test question can effectively distinguish students of different levels and is a high-quality test question. Secondly, these two indicators can be used to evaluate the difficulty of the test questions. If both values ​​are low, it may mean that the test question is too difficult for most students; conversely, if both values ​​are high, the test question may be too simple. Finally, in personalized teaching, teachers can use the students' scores and P values ​​to determine the difficulty of the test questions. s′ or Q s′ Provide students with learning materials and exercises of corresponding difficulty according to the degree of proximity.

[0184] By using P s′ and Q s′ These two functions can more accurately evaluate the quality of test questions and more effectively identify students' learning needs, thereby providing more personalized and targeted learning guidance. This not only helps to optimize test question design, but also provides scientific data support for truly teaching students in accordance with their aptitude.

[0185] Optionally, the calculation method of the student learning cognitive quality evaluation index B in step 200 includes:

[0186] B=β 1 b1+β 2 b2 (5)

[0187] Among them, b1 is the knowledge point cognition level influencing factor of answering questions in previous examinations, and the calculation formula is:

[0188]

[0189] b2 is the influence factor of knowledge point learning degree in daily learning situation analysis, and the calculation formula is:

[0190]

[0191] in,

[0192] n1, n2, and n3 are the numbers of knowledge points, secondary knowledge points, and general knowledge points in the test questions respectively;

[0193] p′ is the number of questions that the student answered that included the i-th assessment knowledge point;

[0194] S pi It refers to the student's score rate when answering the pth question containing the ith assessment knowledge point;

[0195] q′ is the number of questions that the student answers that contain the jth secondary assessment knowledge point;

[0196] S qj It refers to the student's score rate in answering the qth question containing the jth secondary assessment knowledge point;

[0197] r′ is the number of questions that the student answered that included the kth general knowledge point;

[0198] S rk It refers to the score rate of students answering the rth question containing the kth general knowledge point;

[0199] s′ is the total number of times the student browses / answers the materials / questions containing the i-th assessment knowledge point;

[0200] u si The value of whether the material / question containing the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the writing material / question;

[0201] v si The value of whether the material / question containing the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the listening and speaking material / question;

[0202] t si Refers to the time when the student browsed / answered the material / question containing the i-th assessment knowledge point for the sth time;

[0203] w′ is the total number of times the student browses / answers the materials / questions containing the jth secondary assessment knowledge point;

[0204] u wj The value of whether the material / question containing the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the writing material / question;

[0205] vwj The value of whether the material / question containing the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the listening and speaking material / question;

[0206] t wj Refers to the time when the student browsed / answered the material / question containing the jth secondary assessment knowledge point for the wth time;

[0207] z′ is the total number of times the student browses / answers the materials / questions containing the kth general knowledge point;

[0208] u zk The value of whether the material / question containing the kth general knowledge point browsed / answered by the student for the zth time belongs to the writing material / question;

[0209] v zk The value of whether the material / question containing the kth general knowledge point browsed / answered by the student for the zth time belongs to the listening and speaking material / question;

[0210] t zk Refers to the time when the student browsed / answered the material / question containing the kth general knowledge point for the zth time;

[0211] β 1 , β 2 , μ 1 , μ 2 , μ 3 , 1 , 2 , π 1 , π 2 , σ 1 , σ 2 , τ 1 , τ 2 is the weight coefficient.

[0212] Optionally, the calculation method of the test question answer big data analysis result Y in step 200 includes:

[0213] Y=λ 1 A+λ 2 B (8)

[0214] Among them, Y is the result of big data analysis of test answers;

[0215] A is the test question quality assessment index, and its calculation method is as described in claim 2;

[0216] B is the student learning cognitive quality evaluation index, and its calculation method is as described in claim 3;

[0217] λ 1 , 2is the weight coefficient, which is used to adjust the importance of the test question quality evaluation index and the student learning cognitive quality evaluation index in the final analysis results and meets the following requirements:

[0218] 0<λ 1 <1, 0<λ 2 <1,λ 1 +λ 2 =1;

[0219] The weight coefficient λ 1 , 2 It can be dynamically adjusted according to actual teaching needs to adapt to different teaching scenarios and assessment objectives.

[0220] In step 200, the system constructs a big data analysis model for students' test answering. The core of the model is to comprehensively evaluate the quality of test questions and the quality of students' learning cognition to generate big data analysis results of test answering. Specifically, the model achieves this goal through three main calculation steps:

[0221] 1) Calculation of the test question quality assessment index A: This step measures the quality of the test questions by analyzing the content rationality, difficulty and discrimination of the test questions. Formula (2), Formula (3) and Formula (4) respectively describe the calculation methods of the content rationality influencing factor, the difficulty influencing factor and the discrimination influencing factor. Among them, each influencing factor takes into account multiple attributes of the test questions, such as the assessment weight of the knowledge points, the mastery requirements of the students and the cognitive level of the teaching group. Through these calculations, the system can comprehensively evaluate all aspects of the test questions to ensure that the quality of the test questions meets the teaching requirements.

[0222] 2) Calculation of student learning cognitive quality evaluation index B: This step evaluates the student's cognitive level of knowledge points. Formula (6) and Formula (7) can quantify the student's understanding of specific knowledge points by analyzing the student's performance in previous exams and daily learning, thereby identifying the student's learning weaknesses. This evaluation is based on the student's answer history data and knowledge point mastery, and determines the student's cognitive level through a complex matching algorithm.

[0223] 3) Calculation of the big data analysis result Y of the test questions and answers: Finally, the system combines the test question quality evaluation index A with the student learning cognitive quality evaluation index B to calculate a comprehensive analysis result Y. This result can reflect the overall quality of the student's answer and be used for subsequent personalized tutoring decisions.

[0224] Through these technical means, the system can accurately identify quality problems in test questions and students’ learning weaknesses. This not only improves the accuracy of test question quality assessment, but also provides solid data support for personalized tutoring, so that tutoring programs can better match students’ actual needs, thereby significantly improving learning outcomes.

[0225] Step 300: Construct a personalized tutoring decision - making analysis model. Based on the big - data analysis result Y of test - question answering, the test - question quality assessment index A, and the student's learning and cognitive quality assessment index B, the model determines the reasons for the student's personal wrong questions and selects corresponding tutoring resources from the digital resource library according to the determination result of the reasons for the wrong questions.

[0226] Optionally, the specific method for determining the reasons for wrong questions in step 300 includes:

[0227] (1) Set three thresholds:

[0228] A′ is the threshold of the test - question quality assessment index; B′ is the threshold of the student's learning and cognitive quality assessment index;

[0229] (2) Determine the reasons for wrong questions according to the comparison results of Y, A, B with the corresponding thresholds:

[0230] a) When Y > Y′, it is determined that the student made a mistake due to carelessness, and no tutoring measures need to be taken;

[0231] b) When Y ≤ Y′, further compare A and B with their corresponding thresholds: i) When A < A′ and B ≥ B′, it is determined that the poor quality of the test - question leads to the student making a mistake; ii) When A ≥ A′ and B < B′, it is determined that the low learning and cognitive quality of the student leads to the student making a mistake; iii) When A < A′ and B < B′, it is determined that the poor quality of the test - question and the low learning and cognitive quality of the student jointly lead to the student making a mistake; iv) When A ≥ A′ and B ≥ B′, it is determined that other factors lead to the student making a mistake, and further analysis or manual intervention may be required;

[0232] Among them, the thresholds Y′, A′, B′ can be dynamically adjusted according to the actual teaching situation and statistical data to ensure the accuracy and adaptability of the determination of the reasons for wrong questions. And the determination result of the reasons for wrong questions is used as the basis for subsequent personalized tutoring decisions, for selecting corresponding tutoring resources and formulating tutoring strategies.

[0233] In step 300, in this embodiment, a personalized tutoring decision - making analysis model is constructed. The core function of this model is to determine the reasons for the student's wrong questions based on the big - data analysis result Y of test - question answering, the test - question quality assessment index A, and the student's learning and cognitive quality assessment index B, and select appropriate tutoring resources and strategies accordingly, as follows:

[0234] 1) Threshold setting and comparison: First, this embodiment sets three thresholds for the big - data analysis result Y of test - question answering, the test - question quality assessment index A, and the student's learning and cognitive quality assessment index B. Then, this embodiment determines the main reasons for the wrong questions by comparing these actual calculation results with the set thresholds.

[0235] When A is less than A' and B is greater than B', this embodiment determines that the error is mainly due to the quality of the test questions, which means that the test questions themselves may be too difficult or unreasonable in design, causing students to make mistakes.

[0236] When A is greater than A′ and B is less than B′, this embodiment determines that the error is caused by the low quality of students' learning cognition, indicating that the students do not have a sufficient grasp of the knowledge points.

[0237] When A is less than A′ and B is less than B′, it indicates that there may be problems with both the quality of the test questions and the quality of students’ cognition.

[0238] When A and B are both within their respective threshold ranges but wrong questions still occur, this embodiment may further analyze whether there are other potential influencing factors, such as the student's status, the examination environment, etc.

[0239] 2) Dynamic adjustment and feedback: This embodiment not only determines the cause of error based on current data, but also has the ability to make dynamic adjustments. Through real-time analysis and statistical data, this embodiment can dynamically adjust the threshold and analysis method according to the student's performance and test question characteristics to ensure the accuracy and adaptability of the tutoring decision.

[0240] Through this personalized tutoring decision analysis model, this embodiment can accurately determine the reasons why students make mistakes and provide students with more effective tutoring solutions. This technical method ensures the effective use of tutoring resources and can dynamically adjust tutoring strategies according to students' actual needs, thereby greatly improving students' learning effects and review efficiency.

[0241] Step 400: Providing a personalized review plan for students based on the determination result of the personalized tutoring decision analysis model:

[0242] a) When the error is determined to be due to poor question quality, relevant classic case solutions are selected from the case analysis library, and high-quality questions are recommended from the question library;

[0243] b) when it is determined that the error is caused by the low quality of the student's learning cognition, selecting knowledge point courseware of the student's weak knowledge point from the knowledge point explanation courseware library, and matching test questions similar to the wrong question from the test question library;

[0244] c) When the error is determined to be due to poor test quality and low cognitive quality of student learning, the tutoring measures in steps a) and b) are performed simultaneously to form a multi-level tutoring resource.

[0245] Optionally, the method for recommending high-quality test questions in step 400 includes:

[0246] (1) Calculate the recommendation degree F(x) of high-quality test questions. The calculation formula is:

[0247]

[0248] Among them, F(x) is the recommendation result of high-quality test questions;

[0249] F(x) (1) Indicates the degree to which the high-quality test question contains the knowledge points tested in the wrong question;

[0250] F(x) (2) Indicates the rationality of the content of high-quality test questions;

[0251] F(x) (3) Indicates the difficulty of high-quality test questions;

[0252] F(x) (4) The table shows the discrimination of high-quality test questions;

[0253] are weight coefficients respectively, and satisfy:

[0254]

[0255] (2) Calculate the recommendation degree F(x) for all test questions in the test question bank and sort them from high to low according to the recommendation degree;

[0256] (3) Select the top N test questions with the highest recommendation degree as recommended high-quality test questions, where N is a configurable parameter and the default value can be set to 3 or 5;

[0257] (4) When recommending high-quality test questions, the following factors may be considered:

[0258] a) The knowledge point coverage of the test questions, ensuring that the recommended test questions can cover the main test knowledge points in the wrong questions;

[0259] b) The difficulty gradient of the test questions, ensuring that the difficulty of the recommended test questions has a certain progressive relationship;

[0260] c) Diversity of test questions, ensuring that the recommended test questions include different types, such as multiple-choice questions, fill-in-the-blank questions, and essay questions;

[0261] (5) Calculate the degree to which the high-quality test questions contain the knowledge points tested in the wrong questions F(x) (1) Method: F(x) (1) = (number of identical assessment knowledge points) / max (number of assessment knowledge points for wrong questions, number of assessment knowledge points for high-quality questions);

[0262] (6) Calculate the content rationality F(x) of high-quality test questions (2) Method: The specific implementation method is the same as formula (2) in claim 1;

[0263] (7) Calculate the difficulty of high-quality test questions F(x) (3) Method: The specific implementation method is the same as formula (3) in claim 1;

[0264] (8) Calculate the discrimination F(x) of high-quality test questions (4) Method: The specific implementation method is the same as formula (4) in claim 1;

[0265] (9) Regularly update the high-quality test question bank and eliminate poor-quality test questions to ensure that the test question bank always contains high-quality test questions;

[0266] (10) Dynamically adjust the weight coefficient based on students’ feedback and learning progress To optimize the effect of the recommendation algorithm.

[0267] Through this method, the most suitable high-quality test questions can be recommended to students, helping them to better master the knowledge points and improve their learning effects.

[0268] Optionally, the method for matching similar test questions in step 400 includes:

[0269] (1) Calculate the matching degree F(y) of similar test questions. The calculation formula is:

[0270] F(y)=ω1*(F(y) (1) ) +ω2*(F(y) (2) )+ω3*(F(y) (3) )+ω4*(F(y) (4) ) (10)

[0271] Among them, F(y) is the matching result of similar test questions;

[0272] F(y) (1) Indicates the matching degree between similar test questions and the wrong question;

[0273] F(y) (2) Indicates the matching degree of the number of knowledge points that the similar test questions and the wrong question contain the same knowledge points;

[0274] F(y) (3) Indicates the matching degree of knowledge point assessment weights for the same knowledge point;

[0275] F(y) (4) Indicates the knowledge point cognitive index matching degree of the same knowledge point;

[0276] ω1, ω2, ω3, ω4 are weight coefficients respectively, and they satisfy:

[0277] 0<ωi<1(i=1,2,3,4), ω1+ω2+ω3+ω4=1

[0278] (2) Calculate the matching degree F(y) for all test questions in the test question bank and sort them from high to low according to the matching degree;

[0279] (3) Select the first M test questions with the highest matching degree as recommended similar test questions, where M is a configurable parameter and the default value can be set to 5;

[0280] (4) When matching similar test questions, consider the following factors: a) Question type matching: give priority to questions of the same or similar type as the wrong questions; b) Knowledge point coverage: ensure that the matching test questions cover the main knowledge points in the wrong questions; c) Difficulty gradient: select test questions that are slightly lower, equivalent to, or slightly higher in difficulty than the wrong questions to form a difficulty gradient; d) Cognitive level: select test questions that are suitable for the students' current learning stage based on their knowledge point cognitive level, and ensure that the matching test questions fully cover the students' weak knowledge points;

[0281] (5) Calculate the question type matching degree F(y) (1) Method:

[0282] If the question types are exactly the same, F(y) (1) =1;

[0283] If the question types are similar (such as multiple choice questions and true / false questions), F(y) (1) =0.8;

[0284] If the question types are different but belong to the same category (e.g., all are objective questions), F(y) (1) =0.5;

[0285] If the question type is completely different, F(y) (1) =0.2;

[0286] (6) Calculate the matching degree of the number of knowledge points F(y) (2) Method: F(y) (2) = (number of identical knowledge points) / max (number of wrong questions, number of similar questions)

[0287] (7) Calculate the knowledge point assessment weight matching degree F(y) (3) Method: F(y) (3) =1-|Average assessment weight of wrong questions and knowledge points-Average assessment weight of similar questions and knowledge points|

[0288] (8) Calculate the knowledge point cognitive index matching degree F(y) (4) Method: F(y) (4) =1-|Average cognitive index of wrong questions knowledge points-Average cognitive index of similar questions knowledge points|

[0289] (9) According to students’ feedback and learning progress, the weight coefficients ω1, ω2, ω3, and ω4 are dynamically adjusted to optimize the effect of the matching algorithm.

[0290] Through this method, the most similar and targeted test questions can be matched to students, helping them to better understand and master relevant knowledge points and improve their learning efficiency.

[0291] In step 400, the system generates a personalized review plan through the personalized tutoring decision analysis model to provide targeted learning support for students. The core of this step is to select appropriate tutoring resources and strategies based on the judgment results of the aforementioned analysis model, and optimize the review plan generation process to ensure that students can obtain the most appropriate review materials and exercises, as follows:

[0292] 1) Calculation of recommended high-quality test questions: The system first calculates the recommendation of each test question based on the knowledge points covered, content rationality, difficulty, and discrimination. Through formula (9), the system assigns a weight coefficient to each dimension to calculate the comprehensive recommendation F(x). The system then sorts all test questions and selects the test questions with the highest recommendation as high-quality test questions. This process ensures that students can practice high-quality test questions that cover weak knowledge points and maximize the review effect.

[0293] 2) Calculation of matching of similar test questions: For students who need to further consolidate their knowledge, the system will calculate the matching degree of similar test questions based on factors such as question type matching degree, knowledge point matching degree, and difficulty matching degree. Through formula (10) and the corresponding weight coefficient, the system can accurately match the questions closest to the wrong questions and give priority to recommending test questions of moderate difficulty to help students gradually master weak knowledge points.

[0294] 3) Dynamic adjustment and optimization: After each learning cycle, the system will dynamically adjust the weight coefficient and recommendation algorithm based on students’ feedback and learning progress. This process ensures that the generation of tutoring plans can continuously adapt to students’ learning needs and improve the recommendation accuracy of the algorithm.

[0295] Through these technical means, the system can provide students with accurate and personalized review plans, covering all levels from consolidating basic knowledge points to practicing moderately difficult simulation questions. This method not only improves the pertinence of learning, but also significantly improves students' review efficiency and learning effects. Students can gradually improve their overall learning level by repeatedly practicing and consolidating weak knowledge points.

[0296] Step 500: According to the tutoring measures determined by the personalized tutoring decision analysis model, corresponding tutoring resources are selected from the digital resource library to generate a personalized review plan for the student.

[0297] Optionally, the method for generating a personalized review plan in step 500 includes:

[0298] First, the system selects different combinations of tutoring resources based on the error cause determination result in step 300 to formulate a targeted review plan. When the error is caused by poor quality of test questions, the system will select 1 to 2 classic case solutions related to the wrong question knowledge points from the case analysis library, and select 3 to 5 high-quality test questions recommended according to the above method from the test question library to help students consolidate relevant knowledge points. If the error is caused by the low quality of students' learning cognition, the system will select 2 to 3 courseware related to students' weak knowledge points from the knowledge point explanation courseware library, and select 5 to 8 similar test questions matched according to the above method from the test question library to help students gradually improve their mastery of relevant knowledge points.

[0299] In some cases, the cause of the error may be related to both the poor quality of the test questions and the low quality of students' learning cognition. At this time, the system will comprehensively consider these two factors, select 1 classic case solution related to the knowledge point of the wrong question from the case analysis library, and select 2 to 3 courseware related to the students' weak knowledge points from the knowledge point explanation courseware library. At the same time, the system will also select 2 to 3 high-quality test questions and 3 to 5 similar test questions from the test question library to help students strengthen their knowledge mastery from multiple dimensions.

[0300] In addition, the system also includes a regular evaluation and adjustment mechanism. After each learning cycle, the system will evaluate the effectiveness of the review plan and dynamically adjust the combination of tutoring resources and learning strategies based on students' feedback and actual learning performance. By continuously optimizing the generation algorithm of personalized review plans, the system can continuously improve the tutoring effect and ensure that students' learning efficiency and learning effects are maximized. In this way, the system can tailor a comprehensive and effective personalized review plan for each student, make full use of various tutoring resources, and significantly improve students' learning outcomes.

[0301] In step 500, this embodiment generates a personalized review plan for each student based on the personalized tutoring decision analysis model. The core of this step is to accurately select and combine tutoring resources suitable for students through intelligent algorithms and models, so as to provide customized learning support for each student, as follows:

[0302] 1) Selection and combination of tutoring resources: This embodiment first decides which tutoring resources to select based on the result of the error cause determination in step 300. In the case of poor test question quality, this embodiment will extract relevant classic case solutions from the case analysis library and recommend high-quality test questions from the test question library. In the case of low student learning cognitive quality, this embodiment will select knowledge point explanation courseware and test questions similar to wrong questions for targeted tutoring. In some complex situations, such as the existence of both poor test question quality and low student cognitive quality, this embodiment can comprehensively consider multiple resources to consolidate and expand knowledge points from multiple dimensions.

[0303] 2) Dynamic optimization of personalized programs: After each learning cycle, this embodiment will evaluate and adjust the selection and combination of tutoring resources based on the students' actual learning progress and feedback. This process ensures that the review plan can be continuously optimized with the students' learning needs, and always maintains efficient and accurate tutoring results.

[0304] 3) Application of intelligent algorithms: When generating a review plan, this embodiment calculates the recommendation and matching degree of different test questions to select the test questions that best suit the student's current learning stage and knowledge level. This embodiment not only considers the rationality, difficulty and discrimination of the test questions, but also dynamically adjusts the recommendation algorithm through big data analysis and machine learning technology to ensure the scientificity and effectiveness of the review plan.

[0305] Through these technical means, step 500 realizes the intelligent generation of personalized review plans, ensuring that students can access the most suitable tutoring resources during the review process. This method not only improves the pertinence of learning, but also improves learning efficiency and learning effects through dynamic adjustment and optimization. Ultimately, this embodiment helps students better master knowledge points and significantly improve learning outcomes through multi-level and multi-dimensional personalized tutoring.

[0306] It should be noted that the above embodiment has achieved at least the following technical innovations:

[0307] First, the above embodiment constructs a complete set of digital resource libraries, including knowledge point libraries, knowledge point explanation courseware libraries, case analysis libraries, and test question libraries. Through systematic knowledge point extraction and management, standardized processing of multimedia courseware, refined compilation of case analysis, and intelligent management of test questions, a digital resource system with comprehensive content, clear structure, and controllable quality is formed, providing a solid data foundation for personalized intelligent tutoring.

[0308] Secondly, the above embodiment innovatively constructs a big data analysis model for students' test answering. The model comprehensively considers the two dimensions of test question quality and student learning cognitive quality, and establishes a set of scientific evaluation index systems. By calculating indicators such as the rationality of test question content, test question difficulty, and test question discrimination, the quality of test questions is evaluated in multiple dimensions; by analyzing the knowledge point cognition degree and daily learning behavior of students' answers to previous test questions, the quality of students' learning cognition is accurately evaluated. Based on the massive test answering data, the model can comprehensively and accurately analyze the students' test results, providing strong support for personalized tutoring decisions.

[0309] Thirdly, the above embodiment also constructs a personalized tutoring decision analysis model. The model is based on the big data analysis results of students' test answers, integrates the test question quality and students' cognitive level, determines the reasons for students' individual wrong answers, and forms a personalized tutoring strategy that teaches students in accordance with their aptitude. In response to different situations such as poor test question quality and low student learning cognitive quality, the model can intelligently match and recommend corresponding tutoring resources from the digital resource library, such as classic case analysis, knowledge point explanation courseware, high-quality test questions, etc., to form a progressive personalized review plan. Through the systematic combing of knowledge points, checking for deficiencies and filling in gaps, scientific control of difficulty, step-by-step progress, and accurate push of tutoring resources, the above embodiment can maximize the learning efficiency and effectiveness of students.

[0310] In addition, the above embodiments make full use of artificial intelligence technologies such as big data analysis and machine learning in the construction of digital resource libraries, accurate analysis of student portraits, and intelligent generation of personalized tutoring decisions, greatly improving the intelligence level of the system and the tutoring effect. The digital resource library supports functions such as intelligent retrieval, resource association, and combination recommendation, which facilitates the rapid matching and optimal combination of tutoring resources. The big data analysis model of student test answering can process massive amounts of student portrait data, accurately analyze and update students' cognitive levels in real time. The personalized tutoring decision analysis model can dynamically adjust tutoring strategies and resource push based on students' feedback, and continuously improve the fit between intelligent tutoring and students' actual needs.

[0311] In summary, the above embodiments have achieved systematic technical breakthroughs in consolidating the foundation of digital resources, deepening big data analysis, and intelligent personalized tutoring decision-making, forming key technologies and overall solutions to support personalized intelligent tutoring. The resource library, analysis model, strategy model and its coordination mechanism constructed by the above embodiments have better solved the problems of insufficient resources, extensive analysis, and "one size fits all" tutoring decision-making in traditional intelligent tutoring systems, and have significant technical advantages and application value in improving students' learning outcomes and promoting teaching students in accordance with their aptitude.

[0312] In order to better understand the technical solution of the present application, a specific example is provided below for illustration. The details listed in the example are mainly for ease of understanding and are not intended to limit the scope of protection of the present application.

[0313] This example proposes a personalized intelligent tutoring method. The method first provides resources for analyzing test results and tutoring decisions by constructing a digital resource library including a knowledge point library, a knowledge point explanation courseware library, a case analysis library, and a test question library. Secondly, a big data analysis model for student test answering is constructed. On the one hand, the model evaluates the test question quality through indicators such as the rationality of the test question content, the difficulty of the test question, and the discrimination of the test question. On the other hand, it evaluates the students' learning cognitive quality through the knowledge point cognition degree of the students' answers to previous exams and the knowledge point learning degree of the daily learning situation analysis, thereby achieving a comprehensive and accurate analysis of the test results. Further, a personalized tutoring decision analysis model is constructed. According to the big data analysis results of the test question answering, different reasons for the students' individual wrong questions are determined, and different tutoring measures are taken in a targeted manner to facilitate students to strengthen and consolidate the review of knowledge points.

[0314] Specific personalized intelligent tutoring systems such as Figure 1 shown.

[0315] The specific implementation method includes the following three parts:

[0316] 1. Building a digital resource library and database

[0317] Construct a digital resource library including a knowledge point library, a knowledge point explanation courseware library, and a test question library to provide resources for analyzing test results and coaching decision-making. The specific contents of the construction are as follows:

[0318] ① Knowledge point database: Based on the knowledge point information in the teaching syllabus and examination syllabus of each teaching stage of the subject, a knowledge point database is constructed, and the attributes of each knowledge point (teaching stage, mastery requirements) are marked.

[0319] ②Knowledge point explanation courseware library: collect the explanation courseware library of each knowledge point, including the phonetic symbols, word meanings, usage, examples and other analytical content of each knowledge point.

[0320] ③Case analysis library: collects analysis materials of classic case test questions, including analysis contents such as solution ideas and solution steps.

[0321] ④ Question bank: Collect question bank materials through manual screening, review and batch import, and mark the question attributes (question type, test knowledge points, difficulty, discrimination, etc.).

[0322] (II) Constructing a big data analysis model for students’ test answers

[0323] Construct a big data analysis model for students' test answers. This model comprehensively evaluates the quality of test questions and students' learning cognitive quality, obtains big data analysis results of test answers, and can be used for personalized tutoring decision analysis. The corresponding expression is as follows:

[0324] Y=λ 1 A+λ 2 B (1)

[0325] In formula (1), Y is the result of big data analysis of test questions, A is the test question quality evaluation index, B is the student learning cognitive quality evaluation index, and the weight values ​​λ are set according to the different importance of the evaluation index. 1 , 2 The specific calculation methods of each evaluation index are as follows:

[0326] ①Test Question Quality Assessment Index A

[0327] The test question quality assessment index is an indicator that measures whether the test questions are suitable for teaching groups to conduct testing, review and other teaching activities. The test question quality assessment index is obtained through a comprehensive assessment of factors such as the rationality of the test question content, the difficulty of the test question, and the discrimination of the test question.

[0328] Specifically, the calculation method of the test question quality assessment index is:

[0329] A=α 1 a1+α 2 a2+α 3 a3 (2)

[0330]

[0331] In formula (2), A is the test question quality evaluation index; a1 is the test question content rationality influencing factor, a2 is the test question difficulty influencing factor, a3 is the test question discrimination influencing factor; α 1 , α 2 , α 3 are the weight coefficients of each dimension respectively.

[0332] In formula (3), a1 is the rationality of the test content; n1 is the number of knowledge points in the test, n2 is the number of secondary knowledge points in the test, and n3 is the number of general knowledge points in the test; X i , X j , X k are the assessment weights of the i, j, and k knowledge points in the test questions, that is, the weight value of the assessment knowledge point is 2, the weight value of the secondary assessment knowledge point is 1, and the weight value of the general knowledge point is 0.5; i , Y j , Y kare the mastery requirements of the i, j, and k knowledge points in the knowledge point domain in this teaching stage, respectively. The values ​​are divided into y′, y″, and y′′′ according to the three levels of mastery requirements, and y′>y″>y′′′>0; Match(X i ,Y i )、Match(X j ,Y j )、Match(X k ,Y k ) are the matching indexes of the assessment weights and mastery requirements of the i, j, and k knowledge points in the test questions respectively; c i Is the i-th test knowledge point an interference knowledge point? If it is an interference knowledge point, it takes 1, otherwise it takes 0; d j Whether the jth secondary assessment knowledge point is an interference knowledge point, if it is an interference knowledge point, it takes 1, otherwise it takes 0; e k Is the kth general knowledge point an interference knowledge point? If it is an interference knowledge point, it takes 1, otherwise it takes 0; γ 1 , γ 2 , δ 1 , δ 2 , δ 3 , ε 1 , ε 2 , ε 3 are weight coefficients respectively.

[0333] In formula (4), a2 is the difficulty of the test questions; m is the number of teaching groups, L g is the score rate of the g-th student on this test question; n1 is the number of knowledge points in the test question, n2 is the number of secondary knowledge points in the test question, and n3 is the number of general knowledge points in the test question; X i , X j , X k are the assessment weights of the i, j, and k knowledge points in the test questions, respectively. i , Z j , Z k are the cognitive levels of the teaching group for the i, j, and k knowledge points in the test questions (the cognitive level can be obtained by the score rate of the teaching group in answering questions containing this knowledge point); Match(X i ,Y i )、Match(X j ,Y j )、Match(X k ,Y k ) are the matching indexes of the assessment weights of the i-th, j-th, and k-th knowledge points in the test questions and the cognitive level of the teaching group; ∈ 1 ,∈ 2 ,θ 1 ,θ 2 ,θ 3 are weight coefficients respectively.

[0334] In formula (5), a3 is the discrimination of the test question; P s′ is the average score of the test question in the teaching group with a medium to high comprehensive knowledge and cognition level, Q s′ is the average score of the teaching group with a lower-middle range of comprehensive knowledge and cognition level. r is the full score of the test question; m is the number of teaching groups, n1 is the number of test knowledge points in the test question, n2 is the number of secondary test knowledge points in the test question, n3 is the number of general knowledge points in the test question, X i , X j , X k are the assessment weights of the i, j, and k knowledge points in the test questions, respectively. fi , Z fj , Z fk are the cognitive levels of the fth student on the i-th, j-th, and k-th knowledge points, respectively, and G f is the test score of the fth student; θ 1 ,θ 2 , are weight coefficients respectively.

[0335] ②Student learning cognitive quality evaluation index B

[0336] The student learning cognitive quality evaluation index is an indicator that measures whether students' learning effect and knowledge cognition level can correctly answer test questions. The student learning cognitive quality evaluation index is obtained through a comprehensive analysis of factors such as the degree of students' knowledge point cognition in answering previous test questions and the degree of knowledge point learning in daily learning situation analysis.

[0337] Specifically, the calculation method of the student learning cognitive quality evaluation index is:

[0338] B=β 1 b1+β 2 b2 (6)

[0339]

[0340] In formula (6), B is the student learning cognitive quality evaluation index; b1 is the knowledge point cognition degree influencing factor of answering previous examination questions; b2 is the knowledge point learning degree influencing factor of daily learning situation analysis; β 1 , β 2 are the weight coefficients of each dimension respectively.

[0341] In formula (7), b1 is the degree of knowledge point recognition; n1 is the number of test points in the test questions, n2 is the number of secondary test points in the test questions, and n3 is the number of general knowledge points in the test questions; p′ is the number of questions that contain the i-th test point in the test questions answered by the student, SPi It refers to the student's score rate in answering the pth test question containing the i-th assessment knowledge point; q′ is the number of questions answered by the student containing the j-th secondary assessment knowledge point, S qj It refers to the score rate of the student in answering the qth test question containing the jth secondary assessment knowledge point; r′ is the number of questions that the student answered that contained the kth general knowledge point, S rk It refers to the score rate of students in answering the rth question containing the kth general knowledge point; μ 1 , μ 2 , μ 3 are weight coefficients respectively.

[0342] In formula (8), b2 is the degree of knowledge point learning; n1 is the number of test knowledge points in the test questions, n2 is the number of secondary test knowledge points in the test questions, and n3 is the number of general knowledge points in the test questions; s′ is the total number of times the student browses / answers the materials / test questions containing the i-th test knowledge point, and u si The value of whether the material / question containing the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the writing material / question. If it is a writing material / question, it is 1, otherwise it is 0; v si Whether the material / question of the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the listening and speaking material / question. If it is the listening and speaking material / question, it takes 1, otherwise it takes 0; t si refers to the time when the student browsed / answered the material / question of the i-th assessment knowledge point for the sth time; w′ is the total number of times the student browsed / answered the material / question containing the j-th secondary assessment knowledge point, u wj The value of whether the material / question containing the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the writing material / question. If it is a writing material / question, the value is 1, otherwise it is 0; v wj Whether the material / question of the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the listening and speaking material / question. If it is the listening and speaking material / question, it is 1, otherwise it is 0; t wj refers to the time when the student browsed / answered the material / question of the jth secondary knowledge point for the wth time; z′ is the total number of times the student browsed / answered the material / question containing the kth general knowledge point, u zk The value of whether the material / question containing the kth general knowledge point browsed / answered by the student for the zth time belongs to the writing material / question. If it is a writing material / question, it takes 1, otherwise it takes 0; v zk Whether the material / question of the kth general knowledge point browsed / answered by the student for the zth time belongs to the listening and speaking material / question. If it is the listening and speaking material / question, it takes 1, otherwise it takes 0; t zk Refers to the time when the student browsed / answered the material / question of the kth general knowledge point for the zth time; ρ 1 ,2 , π 1 , π 2 , σ 1 , σ 2 , τ 1 , τ 2 are weight coefficients respectively.

[0343] (3) Constructing a personalized tutoring decision analysis model

[0344] Construct a personalized tutoring decision analysis model. First, through the analysis results of big data on students' test question answers, the evaluation results of test question quality, and the evaluation results of students' learning cognitive quality, this model determines the different reasons for students' personal wrong questions. Secondly, according to the different determination results of the reasons for wrong questions, different tutoring measures are taken to provide a personalized review plan for individual students.

[0345] The specific implementation is as follows:

[0346] According to the analysis result Y of big data on students' test question answers, if it is greater than the threshold Y', it is determined that the student made a mistake due to carelessness, and no tutoring measures need to be taken; if it is less than the threshold Y', then the test question quality evaluation index A and the student learning cognitive quality evaluation index B are further determined. If A is less than the threshold A', it is determined that the poor quality of the test question caused the student to make a mistake. If B is less than the threshold B', it is determined that the low learning cognitive quality of the student caused the student to make a mistake. If A is less than the threshold A' and B is less than the threshold B', it is determined that the poor quality of the test question and the low learning cognitive quality of the student together caused the student to make a mistake. Then, according to the determination results of different reasons for wrong questions, corresponding tutoring measures are taken:

[0347] ① Poor quality of test questions: When A < A', it is determined that the reason for the wrong question of this test question is the poor quality of the test question, and tutoring measure 1 is taken: First, generate the "classic case solution" related to this test question from the case analysis library and recommend it to the student for consolidation learning; then recommend and generate "high-quality test questions" from the test question library and recommend them to the student for intensive consolidation review. The recommendation and generation of high-quality test questions are as follows:

[0348]

[0349] In formula (9), F(x) is the recommendation degree result of high-quality test questions; F(x) (1) indicates that this high-quality test question contains the assessment knowledge point of this wrong question; F(x) (2) indicates the content rationality of the high-quality test question; F(x) (3) indicates the difficulty of the high-quality test question; F(x) (4) indicates the discrimination degree of the high-quality test question; are weight coefficients respectively.

[0350] Note: a. For test questions with poor quality, they should be removed from the question bank to ensure that only questions that meet the quality standards are retained in the question bank. This is beneficial for students to answer questions better and effectively improve their academic performance. b. The number of high-quality questions generated can be defined. For example, the top three questions with the highest recommendation scores can be selected as recommended questions.

[0351] In other words, first, test questions with poor quality should be removed from the question bank: This is to ensure quality control of the question bank. The question bank is the basic resource for students to review and take tests. If it contains test questions with poor quality (such as overly difficult questions, unclear assessment points, questions beyond the syllabus, etc.), these questions may cause confusion among students during the learning process and affect their learning effect. Therefore, the system needs to regularly check and remove these unqualified questions to ensure that the questions in the question bank can effectively evaluate and improve students' learning levels. This approach can help students better understand and master knowledge points, thus effectively improving their academic performance. Second, the number of high-quality questions generated can be defined: When recommending high-quality questions, the system can calculate the recommendation score of each question according to an algorithm and select high-quality questions based on the score. This statement suggests that the number of recommended high-quality questions can be defined and adjusted according to actual needs. For example, the system can select the top three questions with the highest recommendation scores as the review practice questions for students. This can ensure that students only encounter carefully selected and qualified questions during review, helping them strengthen practice on key knowledge points and improve learning efficiency.

[0352] ② Low learning and cognitive quality of students: When B < B′, it is determined that the reason for the wrong question of this student is low learning and cognitive quality. Then, tutoring measure 2 is taken: First, generate the "knowledge point courseware" with a low cognitive index for this student from the knowledge point explanation courseware library and recommend it to the student for consolidation learning. Then, match "similar questions" from the question bank and recommend them to the student for enhanced consolidation review. The matching generation of similar questions is as follows:

[0353] F(y) = ω1 * (F(y) (1) ) + ω2 * (F(y) (2) ) + ω3 * (F(y) (3) ) + ω4 * (F(y) (4) ) (10)

[0354] In formula (10), F(y) is the matching result of similar questions; F(y) (1) represents the matching degree of the question type of similar questions and this wrong question; F(y) (2) represents the matching degree of the number of knowledge points of similar questions and this wrong question that contain the same knowledge points; F(y) (3) represents the matching degree of the assessment weight of the same knowledge points; F(y) (4)The matching degree of knowledge point recognition indexes representing the same knowledge point; ω1, ω2, ω3, and ω4 are weight coefficients respectively.

[0355] Note: The number of generated similar test questions can be defined. For example, the top five test questions with the highest matching degree can be selected as recommended test questions.

[0356] ③ Poor test question quality and low learning and cognitive quality of students: When A < A′ and B < B′, it is determined that the wrong reason for this test question is poor test question quality and low learning and cognitive quality of this student. Then, tutoring measure 3 is taken: First, generate the "knowledge point courseware" with a low cognitive index for this student from the knowledge point explanation courseware library and recommend it to this student for consolidation learning; then, generate the "classic case solutions" related to this test question from the case analysis library and recommend it to this student for consolidation learning; further, recommend and generate "high-quality test questions" from the test question library (the generation method of high-quality test questions is the same as ①) and recommend them to this student for intensive consolidation review; finally, match the "similar test questions" of the "high-quality test questions" from the test question library (the generation method of similar test questions is the same as ②) and recommend them to this student for intensive consolidation review.

[0357] The above embodiments have the following technical effects:

[0358] The personalized intelligent tutoring system of the above embodiments significantly improves the accuracy and effectiveness of test question quality assessment and learning and cognitive quality assessment through in-depth analysis of the wrong reasons of students' wrong questions. In traditional intelligent tutoring systems, wrong question analysis often only focuses on students' mastery of knowledge points and ignores the quality problems of the test questions themselves. However, the above embodiments innovatively combine the dual analysis methods of test question quality and learning and cognitive quality, thereby providing more accurate tutoring for students.

[0359] In terms of test question quality analysis, through the assessment of multiple dimensions such as the rationality, difficulty, and discrimination degree of test question content, the quality problems in the test questions can be comprehensively identified. For example, through analysis, the system can find that some test questions are too difficult, or the assessment points are not clearly expressed and cannot effectively test students' actual mastery. By identifying these problems, the system can help educators eliminate unqualified test questions and optimize the overall quality of the test question library, thereby providing a more reasonable and scientific learning and testing environment for students. This accurate analysis of test question quality not only improves the effectiveness of the test but also makes the review process of students more targeted and helps them better master the knowledge they have learned.

[0360] In terms of learning cognitive quality analysis, the above embodiment can accurately evaluate students' mastery of each knowledge point through in-depth analysis of students' previous exams and daily learning data. This analysis method can help teachers and students identify weak links in learning, so as to strengthen the review and consolidation of relevant knowledge points in a targeted manner. This precise cognitive analysis avoids students from blindly learning and repeating ineffective exercises during the review process, which significantly improves learning efficiency.

[0361] In addition, the above-mentioned embodiment comprehensively considers two factors, namely, the quality of test questions and the quality of students' learning cognition, when formulating tutoring plans for students. As a result of this comprehensive analysis, the system can provide students with more comprehensive and personalized tutoring measures. For example, when the system identifies problems with the quality of test questions, the system will automatically recommend relevant classic case analysis and high-quality test questions to help students better understand and master the knowledge points. On the other hand, when the system finds that students lack cognition on certain knowledge points, it will recommend targeted knowledge point explanation courseware and similar test questions to help students gradually improve their understanding and application of these knowledge points.

[0362] Through this dual-consideration tutoring strategy, the above embodiment not only effectively makes up for the students' shortcomings in learning, but also ensures that they can fully master various knowledge points during the review process. This comprehensive, multi-level tutoring program can ultimately help students improve their learning outcomes and achieve significant improvements in grades. The application of this system not only optimizes students' learning experience, but also provides teachers with more scientific teaching aids, greatly improving the overall efficiency and quality of education.

[0363] The second embodiment of the present application relates to a personalized intelligent tutoring system, the structure of which is as follows: Figure 3 As shown, the personalized intelligent tutoring system includes:

[0364] A digital resource library construction module is used to construct a digital resource library, wherein the digital resource library includes a knowledge point library, a knowledge point explanation courseware library, a case analysis library, and a test question library;

[0365] The student test answer big data analysis model construction module is used to construct a student test answer big data analysis model, and the student test answer big data analysis model includes:

[0366] A test question quality evaluation index calculation unit, used to calculate a test question quality evaluation index A, wherein the test question quality evaluation index A is based on the content rationality, difficulty and discrimination of the test questions in the test question library;

[0367] A student learning cognitive quality evaluation index calculation unit, used to calculate a student learning cognitive quality evaluation index B, wherein the student learning cognitive quality evaluation index B is based on the student's knowledge point cognition degree in answering questions in previous exams and the degree of learning of knowledge points in daily learning;

[0368] A test question answering big data analysis result calculation unit, used to calculate the test question answering big data analysis result Y according to the test question quality evaluation index A and the student learning cognition quality evaluation index B;

[0369] A personalized tutoring decision analysis model construction module is used to construct a personalized tutoring decision analysis model. The personalized tutoring decision analysis model determines the cause of a student's personal wrong question based on the test answer big data analysis result Y, the test question quality evaluation index A and the student learning cognitive quality evaluation index B, and selects corresponding tutoring resources from the digital resource library according to the cause determination result;

[0370] The personalized review plan generation module is used to provide students with a personalized review plan based on the determination result of the personalized tutoring decision analysis model, including:

[0371] When it is determined that the error is due to poor test question quality, relevant classic case solutions are selected from the case analysis library, and high-quality test questions are recommended from the test question library;

[0372] When it is determined that the error is caused by the low quality of the student's learning cognition, the knowledge point courseware of the student's weak knowledge point is selected from the knowledge point explanation courseware library, and the test questions similar to the wrong question are matched from the test question library;

[0373] When the error is judged to be due to poor test quality and low student learning cognitive quality, the above two tutoring measures are implemented simultaneously, forming a multi-level tutoring resource;

[0374] The tutoring resource selection module is used to select corresponding tutoring resources from the digital resource library according to the tutoring measures determined by the personalized tutoring decision analysis model, and generate a personalized review plan for individual students.

[0375] The first implementation manner is a method implementation manner corresponding to the present implementation manner. The technical details in the first implementation manner can be applied to the present implementation manner, and the technical details in the present implementation manner can also be applied to the first implementation manner.

[0376] It should be noted that those skilled in the art should understand that the implementation functions of each module shown in the above-mentioned embodiment of the personalized intelligent tutoring system can be understood by referring to the relevant description of the above-mentioned personalized intelligent tutoring method. The functions of each module shown in the above-mentioned embodiment of the personalized intelligent tutoring system can be realized by a program (executable instruction) running on a processor, or by a specific logic circuit. If the above-mentioned personalized intelligent tutoring system of the embodiment of the present application is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiment of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM, Read Only Memory), a disk or an optical disk. In this way, the embodiment of the present application is not limited to any specific combination of hardware and software.

[0377] Accordingly, an embodiment of the present application further provides a computer storage medium, in which computer executable instructions are stored. When the computer executable instructions are executed by a processor, the various method embodiments of the present application are implemented.

[0378] In addition, the embodiments of the present application also provide a personalized intelligent tutoring system, which includes a memory for storing computer executable instructions, and a processor; the processor is used to implement the steps in the above-mentioned method implementations when executing the computer executable instructions in the memory. Among them, the processor can be a central processing unit (Central Processing Unit, referred to as "CPU"), or other general-purpose processors, digital signal processors (Digital Signal Processor, referred to as "DSP"), application specific integrated circuits (Application Specific Integrated Circuit, referred to as "ASIC"), etc. The aforementioned memory can be a read-only memory (read-only memory, referred to as "ROM"), a random access memory (random access memory, referred to as "RAM"), a flash memory (Flash), a hard disk or a solid-state hard disk, etc. The steps of the methods disclosed in the embodiments of the present application can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0379] It should be noted that in the application documents of this patent, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the sentence "including one" do not exclude the existence of other identical elements in the process, method, article or device including the elements. In the application documents of this patent, if it is mentioned that an action is performed according to an element, it means that the action is performed at least according to the element, which includes two situations: performing the action only according to the element, and performing the action according to the element and other elements. Expressions such as multiple, multiple, and multiple include 2, 2 times, 2 kinds, and more than 2, more than 2 times, and more than 2 kinds.

[0380] All documents mentioned in this application are considered to be included in the disclosure of this application as a whole, so that they can be used as the basis for modification when necessary. In addition, it should be understood that after reading the above disclosure of this application, those skilled in the art can make various changes or modifications to this application, and these equivalent forms also fall within the scope of protection claimed in this application.

Claims

1. A personalized intelligent tutoring method, characterized in that , including the following steps: Step A: constructing a digital resource library, which includes a knowledge point library, a knowledge point explanation courseware library, a case analysis library and a test question library; Step B: constructing a big data analysis model for students' test questions and answers, the big data analysis model for students' test questions and answers includes: a) calculating a test question quality evaluation index A, the test question quality evaluation index A is based on the content rationality, difficulty and discrimination of the test questions in the test question bank; b) calculating a student learning cognition quality evaluation index B, the student learning cognition quality evaluation index B is based on the students' knowledge point cognition degree in answering test questions in previous exams and the degree of learning of knowledge points in daily learning; c) calculating a test question answering big data analysis result Y according to the test question quality evaluation index A and the student learning cognition quality evaluation index B; Step C: constructing a personalized tutoring decision analysis model, wherein the personalized tutoring decision analysis model determines the cause of a student's individual wrong answer based on the test answer big data analysis result Y, the test question quality evaluation index A, and the student learning cognitive quality evaluation index B, and selects corresponding tutoring resources from the digital resource library according to the cause determination result; Step D: According to the determination result of the personalized tutoring decision analysis model, provide students with personalized review plans: a) When it is determined that the error is caused by poor test question quality, select relevant classic case solutions from the case analysis library, and recommend high-quality test questions from the test question library; b) When it is determined that the error is caused by low student learning cognitive quality, select knowledge point courseware of students' weak knowledge points from the knowledge point explanation courseware library, and match test questions similar to the wrong questions from the test question library; c) When it is determined that the error is caused by poor test question quality and low student learning cognitive quality, simultaneously execute the tutoring measures in steps a) and b) to form multi-level tutoring resources; Step E: According to the tutoring measures determined by the personalized tutoring decision analysis model, corresponding tutoring resources are selected from the digital resource library to generate a personalized review plan for the individual student.

2. The method according to claim 1, characterized in that The calculation method of the test question quality evaluation index A in step B includes: A=α1a1+α2a2+α3a3 (1) Among them, a1 is the influencing factor of the rationality of the test content, and the calculation formula is: a2 is the factor affecting the difficulty of the test question, and the calculation formula is: a3 is the influencing factor of the test question discrimination, and the calculation formula is: in, n1, n2, and n3 are the numbers of knowledge points, secondary knowledge points, and general knowledge points in the test questions respectively; X i , X i , X k are the assessment weights of the i, j, and k knowledge points in the test questions respectively; Y i , Y j , Y k They are the mastery requirements of the i, j, and k knowledge points in the knowledge point domain in this teaching stage; Z i , Z i , Z k are the cognitive levels of the teaching group for the i, j, and k knowledge points in the test questions respectively; c i ,d i 、e k Respectively indicate whether the i-th, j-th, and k-th knowledge points are interference knowledge points; L g is the score rate of the g-th student on this test question; P s′ The average score of this question among the teaching groups with a medium to high comprehensive knowledge and cognition level; Q s′ The average score of the test question in the teaching group with a lower-middle range of comprehensive knowledge and cognition level; Q r Full marks for the test questions; Z fi , Z fj 、z fk are the cognitive levels of the fth student on the i-th, j-th, and k-th knowledge points respectively; G f Score the test paper of the fth student; m is the number of people in the teaching group; α1, α2, α3, γ1, γ2, δ1, δ2, δ3, ε1, ε2, ε3, ∈1, ∈2, θ1, θ2, θ3, are weight coefficients.

3. The method according to claim 1, characterized in that The calculation method of the student learning cognitive quality evaluation index B in step B includes: B=β1b1+β2b2 (5) Among them, b1 is the knowledge point cognition level influencing factor of answering questions in previous examinations, and the calculation formula is: b2 is the influence factor of knowledge point learning degree in daily learning situation analysis, and the calculation formula is: in, n1, n2, and n3 are the numbers of knowledge points, secondary knowledge points, and general knowledge points in the test questions respectively; p′ is the number of questions that the student answered that included the i-th assessment knowledge point; Spi refers to the student's score rate when answering the pth question containing the ith assessment knowledge point; $q'$ is the number of questions containing the $j$-th secondary assessment knowledge point among the questions answered by the student; S qj It refers to the student's score rate in answering the qth question containing the jth secondary assessment knowledge point; $r'$ is the number of questions containing the $k$-th general knowledge point among the questions answered by the student; S rk It refers to the score rate of students answering the rth question containing the kth general knowledge point; $s'$ is the total number of times the student views / answers the materials / questions containing the $i$-th assessment knowledge point; u si The value of whether the material / question containing the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the writing material / question; v si The value of whether the material / question containing the i-th assessment knowledge point browsed / answered by the student for the sth time belongs to the listening and speaking material / question; t si Refers to the time when the student browsed / answered the material / question containing the i-th assessment knowledge point for the sth time; $w'$ is the total number of times the student views / answers the materials / questions containing the $j$-th secondary assessment knowledge point; u wj The value of whether the material / question containing the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the writing material / question; v wj The value of whether the material / question containing the jth secondary assessment knowledge point browsed / answered by the student for the wth time belongs to the listening and speaking material / question; t wj Refers to the time when the student browsed / answered the material / question containing the jth secondary assessment knowledge point for the wth time; $z'$ is the total number of times the student views / answers the materials / questions containing the $k$-th general knowledge point; u zk The value of whether the material / question containing the kth general knowledge point browsed / answered by the student for the zth time belongs to the writing material / question; v zk The value of whether the material / question containing the kth general knowledge point browsed / answered by the student for the zth time belongs to the listening and speaking material / question; t zk Refers to the time when the student browsed / answered the material / question containing the kth general knowledge point for the zth time; $\beta_1, \beta_2, \mu_1, \mu_2, \mu_3, \rho_1, p_2, \pi_1, \pi_2, \sigma_1, \sigma_2, \tau_1, \tau_2$ are weight coefficients.

4. The method according to claim 1, characterized in that: The calculation method of the big data analysis result $Y$ of question answering in step B includes: $Y = \lambda_1A+\lambda_2B\ (8)$ where $Y$ is the big data analysis result of question answering; $A$ is the question quality assessment index, and its calculation method is as described in claim 2; $B$ is the student learning cognitive quality assessment index, and its calculation method is as described in claim 3; $\lambda_1, \lambda_2$ are weight coefficients used to adjust the importance of the question quality assessment index and the student learning cognitive quality assessment index in the final analysis result, and satisfy: $0\lt\lambda_1\lt1$, $0\lt\lambda_2\lt1$, $\lambda_1+\lambda_2 = 1$; The weight coefficients $\lambda_1, \lambda_2$ can be dynamically adjusted according to actual teaching needs to adapt to different teaching scenarios and assessment goals.

5. The method according to claim 1, characterized in that The specific method of error cause determination in step C includes: (1) Set three thresholds: $Y'$ is the threshold of the big data analysis result of question answering; $A'$ is the threshold of the question quality assessment index; $B'$ is the threshold of the student learning cognitive quality assessment index; (2) Make error cause determination according to the comparison results of $Y$, $A$, $B$ and their corresponding thresholds: a) When $Y\gt Y'$, it is determined that the student made mistakes due to carelessness, and no tutoring measures need to be taken; b) When $Y\leq Y'$, further compare $A$ and $B$ with their corresponding thresholds: i) When $A\lt A'$ and $B\geq B'$, it is determined that the poor question quality causes the student to make mistakes in the questions; ii) When $A\geq A'$ and $B\lt B'$, it is determined that the low learning cognitive quality of the student causes the student to make mistakes in the questions; iii) When $A\lt A'$ and $B\lt B'$, it is determined that the poor question quality and the low learning cognitive quality of the student jointly cause the student to make mistakes in the questions; iv) When $A\geq A'$ and $B\geq B'$, it is determined that other factors cause the student to make mistakes in the questions, and further analysis or manual intervention may be required; Among them, the thresholds $Y'$, $A'$, $B'$ can be dynamically adjusted according to actual teaching situations and statistical data to ensure the accuracy and adaptability of error cause determination. And the error cause determination result is used as the basis for subsequent personalized tutoring decisions to select corresponding tutoring resources and formulate tutoring strategies.

6. The method according to claim 1, characterized in that The recommended method of high-quality questions in step D includes: (1) Calculate the recommendation degree $F(x)$ of high-quality questions, and its calculation formula is: where $F(x)$ is the result of the recommendation degree of high-quality questions; F(x) (1) Indicates the degree to which the high-quality test question contains the knowledge points tested in the wrong question; F(x) (2) Indicates the rationality of the content of high-quality test questions; F(x) (3) Indicates the difficulty of high-quality test questions; F(x) (4) Indicates the discrimination of high-quality test questions; are weight coefficients respectively, and satisfy: and (2) Calculate the recommendation degree $F(x)$ for all questions in the question bank and sort them from high to low according to the recommendation degree; (3) Select the top $N$ questions with the highest recommendation degree as the recommended high-quality questions, where $N$ is a configurable parameter, and the default value can be set to 3 or 5; (4) When recommending high-quality test questions, the following factors should be considered: a) the knowledge point coverage of the test questions, ensuring that the recommended test questions can cover the main knowledge points tested in the wrong questions; b) the difficulty gradient of the test questions, ensuring that the difficulty of the recommended test questions has a certain progressive relationship; c) the diversity of the test question types, ensuring that the recommended test questions include different types, such as multiple-choice questions, fill-in-the-blank questions, and essay questions; (5) Calculate the degree to which the high-quality test questions contain the knowledge points tested in the wrong questions F(x) (1) Method: F(x) (1) = (number of identical assessment knowledge points) / max (number of assessment knowledge points for wrong questions, number of assessment knowledge points for high-quality questions); (6) Calculate the content rationality F(x) of high-quality test questions (2) Method: The specific implementation method is the same as formula (2) in claim 1; (7) Calculate the difficulty of high-quality test questions F(x) (3) Method: The specific implementation method is the same as formula (3) in claim 1; (8) Calculate the discrimination F(x) of high-quality test questions (4) Method: The specific implementation method is the same as formula (4) in claim 1; (9) Regularly update the high-quality test question bank and eliminate poor-quality test questions to ensure that the test question bank always contains high-quality test questions; (10) Dynamically adjust the weight coefficient based on students’ feedback and learning progress To optimize the effect of the recommendation algorithm. Through this method, the most suitable high-quality test questions can be recommended to students, helping them to better master the knowledge points and improve their learning effects.

7. The method according to claim 1, characterized in that The matching method of similar test questions in step D includes: (1) Calculate the matching degree F(y) of similar test questions. The calculation formula is: F(y)=ω1*(F(y)(1))+ω2*(F(y)(2))+ω3*(F(y)(3))+ω4*(F(y)(4)) (10) Among them, F(y) is the matching result of similar test questions; F(y) (1) Indicates the matching degree between similar test questions and the wrong question; F(y) (2) Indicates the matching degree of the number of knowledge points that the similar test questions and the wrong question contain the same knowledge points; F(y) (3) Indicates the matching degree of knowledge point assessment weights for the same knowledge point; F(y) (4) Indicates the knowledge point cognitive index matching degree of the same knowledge point; ω1, ω2, ω3, ω4 are weight coefficients respectively, and they satisfy: 0<ωi<1 (i=1, 2, 3, 4), and ω1+ω2+ω3+ω4=1 (2) Calculate the matching degree F(y) for all test questions in the test question bank and sort them from high to low according to the matching degree; (3) Select the first M test questions with the highest matching degree as recommended similar test questions, where M is a configurable parameter and the default value can be set to 5; (4) When matching similar test questions, the following factors should be considered: a) Question type matching: give priority to questions of the same or similar type as the wrong questions; b) Knowledge point coverage: ensure that the matching test questions can cover the main knowledge points tested in the wrong questions; c) Difficulty gradient: select test questions that are slightly lower, equivalent to, or slightly higher in difficulty than the wrong questions to form a difficulty gradient; d) Cognitive level: select test questions that are suitable for the students' current learning stage based on their knowledge point cognitive level to ensure that the matching test questions fully cover the students' weak knowledge points; (5) Calculate the question type matching degree F(y) (1) Method: If the question type is exactly the same, F(y) (1) =1; If the question types are similar (such as multiple choice questions and true / false questions), F(y) (1) =0.8; If the question types are different but belong to the same category (e.g., all are objective questions), F(y) (1) =0.5; If the question type is completely different, F(y) (1) =0.2; (6) Calculate the matching degree of the number of knowledge points F(y) (2) Method: F(y) (2) = (number of identical knowledge points) / max (number of wrong questions and similar questions) (7) Calculate the knowledge point assessment weight matching degree F(y) (3) Method: F(y) (3) =1-|Average assessment weight of wrong questions and knowledge points-Average assessment weight of similar questions and knowledge points| (8) Calculate the knowledge point cognitive index matching degree F(y) (4) Method: F(y) (4) =1-|Average cognitive index of wrong questions knowledge points-Average cognitive index of similar questions knowledge points| (9) According to students’ feedback and learning progress, the weight coefficients ω1, ω2, ω3, and ω4 are dynamically adjusted to optimize the effect of the matching algorithm.

8. A personalized intelligent tutoring system, characterized in that: include: A digital resource library construction module is used to construct a digital resource library, wherein the digital resource library includes a knowledge point library, a knowledge point explanation courseware library, a case analysis library, and a test question library; A student test answering big data analysis model construction module is used to construct a student test answering big data analysis model, and the student test answering big data analysis model includes: a test question quality evaluation index calculation unit, which is used to calculate the test question quality evaluation index A, and the test question quality evaluation index A is based on the content rationality, difficulty and discrimination of the test questions in the test question bank; a student learning cognition quality evaluation index calculation unit is used to calculate the student learning cognition quality evaluation index B, and the student learning cognition quality evaluation index B is based on the knowledge point cognition degree of the students' test answering in previous exams and the learning degree of knowledge points in daily learning; a test question answering big data analysis result calculation unit is used to calculate the test question answering big data analysis result Y according to the test question quality evaluation index A and the student learning cognition quality evaluation index B; A personalized tutoring decision analysis model construction module is used to construct a personalized tutoring decision analysis model. The personalized tutoring decision analysis model determines the cause of a student's personal wrong question based on the test answer big data analysis result Y, the test question quality evaluation index A and the student learning cognitive quality evaluation index B, and selects corresponding tutoring resources from the digital resource library according to the cause determination result; A personalized review plan generation module is used to provide students with personalized review plans based on the judgment results of the personalized tutoring decision analysis model, including: when it is determined that the wrong reason is poor test question quality, selecting relevant classic case solutions from the case analysis library, and recommending high-quality test questions from the test question library; when it is determined that the wrong reason is low student learning cognitive quality, selecting knowledge point courseware of the student's weak knowledge points from the knowledge point explanation courseware library, and matching test questions similar to the wrong questions from the test question library; when it is determined that the wrong reason is poor test question quality and the student's learning cognitive quality is low, executing the above two tutoring measures at the same time to form a multi-level tutoring resource; The tutoring resource selection module is used to select corresponding tutoring resources from the digital resource library according to the tutoring measures determined by the personalized tutoring decision analysis model, and generate a personalized review plan for individual students.

Images