Personalized knowledge graph learning system and method based on dynamic weight rule

Abstract

The invention discloses a personalized knowledge graph learning system and method based on a dynamic weight rule, and relates to the technical field of personalized learning, and the system comprises a dynamic question grouping module, a multi-dimensional evaluation analysis module, a knowledge graph lightening module, an Ebbinghaus review reminding module and an automatic question generation module. According to the method, a dynamic question grouping algorithm based on a knowledge label test frequency grade is constructed, scientific distribution of exercises is realized, a multi-dimensional exercise result analysis algorithm model is established, knowledge mastering grades are divided in combination with scores and key conditions, and the knowledge mastering degree is visually presented through a knowledge graph according to a set rule. According to the method, the user can intuitively feel the problem owned by the user at present, defects can be checked and repaired in time, a review reminding mechanism which combines behavior habits and data of a user group and is based on an optimized Ebbinghaus forgetting curve is designed, and the knowledge memory retention rate of the user is increased.

Description

Technical Field

[0001] This invention relates to the technical field of personalized learning, specifically to a personalized knowledge graph learning system and method based on dynamic weighting rules. Background Technology

[0002] Personalized knowledge graph learning systems are intelligent education platforms based on dynamic weighting algorithms and knowledge graph visualization technology. Through multi-module collaboration, these systems achieve precise analysis of user learning behavior, personalized learning path planning, and scientific review management, effectively solving some problems existing in traditional online education. However, existing learning systems have significant shortcomings in practice and knowledge management, mainly in the following three aspects: ① Lack of personalization and targeting in question selection: Most systems adopt random question selection or fixed question bank mode, without combining real-time syllabus knowledge point test frequency level and user learning behavior data for scientific allocation. As a result, the practice results cannot accurately reflect the user's mastery of high-frequency test points, making it difficult to achieve the core goal of a complete knowledge mastery process from "accurate assessment" and "personalized learning" to "dynamic review reminders".

[0003] ② Vague feedback on knowledge mastery: Existing educational scenarios or digital systems often present practice results with a single score, making it difficult to establish a correlation between scores and knowledge mastery levels for each learning group with limited teaching resources. Furthermore, it is difficult to intuitively display users' knowledge gaps through visualization methods (such as knowledge graphs). As a result, learning groups cannot obtain effective information from external feedback, and therefore cannot quickly and accurately locate their weaknesses, making it difficult to effectively master the knowledge content.

[0004] ③ The review mechanism lacks scientific rigor: Most educational and teaching scenarios and digital systems simply review and explain the content of the questions, such as the wrong questions, after the user completes the exercises. They do not set up personalized review nodes based on the forgetting curve that conforms to the user's behavioral habits, resulting in low retention rate of the user's learned knowledge and difficulty in improving learning efficiency.

[0005] Furthermore, most existing learning systems adopt a "one-to-many" teaching model, where a fixed question bank or a fixed learning path is offered to all users. This lacks the ability to deeply perceive and dynamically adjust to individual learning habits, cognitive differences, and knowledge gaps. This model makes it difficult to achieve truly "personalized learning" and also fails to build a "many-to-many" interactive learning relationship. Summary of the Invention

[0006] The purpose of this invention is to provide a personalized knowledge graph learning system and method with dynamic weight rules to overcome the above-mentioned defects in the prior art.

[0007] A personalized knowledge graph learning system with dynamic weight rules, the system comprising: The dynamic question generation module is used to generate personalized question sets based on the frequency level of knowledge tags and user learning data. The question types include basic questions, intermediate questions, and intensive questions, and are allocated according to a preset ratio. The multi-dimensional assessment and analysis module is used to analyze users' answers, classify knowledge mastery levels through a scoring calculation model, and map the results into visual feedback. The knowledge graph highlighting module is used to calculate a weighted score based on the assessment results and update the display status of knowledge points in the knowledge graph in combination with key conditions, using different colors to indicate the degree of mastery; The Ebbinghaus review reminder module is used to automatically activate and push personalized review reminders based on the optimized forgetting curve time nodes; The automatic question generation module calls the AI ​​big model interface and provides the AI ​​big model with structured parameters such as knowledge point description, test frequency level, and target difficulty. The AI ​​big model then generates new questions that meet the requirements in real time.

[0008] Preferably, the question-generating rules of the dynamic question-generating module include: dividing the practice into the smallest units based on knowledge tags, extracting 10 questions from each smallest practice unit, configuring the question types according to the ratio of 50% basic questions, 30% advanced questions, and 20% sprint questions, with corresponding weight scores of 1, 2, and 3 respectively, and the total score calculation formula is: total score = (number of correct basic questions × 1) + (number of correct advanced questions × 2) + (number of correct sprint questions × 3), with a theoretical maximum score of 17 points; The assessment questions are divided into initial assessment and assessment during the learning process. The initial assessment draws questions based on the frequency of the subject in five levels. The assessment during the learning process is limited to drawing questions from the knowledge content that the user has already learned. The frequency level allocation rules are based on the number of questions drawn to meet the requirements of 30, 40 or 50 questions.

[0009] Preferably, the multi-dimensional assessment and analysis module adopts a score calculation model and a mastery level classification rule, wherein the score calculation model is as follows:

[0010] in, It is a percentage score. Let M = 17 be the score for the i-th frequency level, and M = 17 be the full score for the question corresponding to a single frequency level. The mastery level is divided into four levels: weak, passable, good, and excellent, which correspond to an algorithm score of ≤5 points, 5 < score ≤8 points, 8 < score ≤13 points, and 13 < score ≤17 points, respectively. Each level has a clear description of the ability.

[0011] Preferably, the working mechanism of the knowledge graph lighting module includes: the after-class exercises are fixed at 20 questions, consisting of one or more minimum practice units, and are bound to the minimum lighting unit of the knowledge graph; the weighted score calculation formula is as follows:

[0012] Where m is the number of exercises. For the score of the k-th exercise, For the knowledge tag set of the k-th exercise, Let j be the frequency value of knowledge tag j, and n be the total number of knowledge tags; Knowledge graph nodes are presented in different colors according to the level of mastery. Dark green indicates mastery, with a score of ≥12 points and ≥4 correct answers on basic questions and ≥2 correct answers on advanced questions. Dark yellow indicates familiarity, with a score of 5 < ≤11 points and ≥3 correct answers on basic questions. Gray indicates understanding, with a score of ≤5 points. White indicates the initial state of not having started learning. If the score range is met but the key conditions are not met, the node will be automatically downgraded.

[0013] Preferably, the optimized Ebbinghaus forgetting curve review reminder module is automatically activated after the user generates a learning record. The review time nodes are set according to the time elapsed since the first learning: the first review reminder is after 2 days, the second after 6 days, the third after 15 days, and the fourth after 21 days, accurately matching the optimized Ebbinghaus forgetting curve.

[0014] Preferably, the workflow of the automatic question generation module is as follows: the system monitors the question bank status and user learning data in real time. When it detects that the coverage of questions for a specific knowledge point is insufficient or the practice demand is saturated, the question generation process is triggered. The system extracts structured parameters such as knowledge point description, test frequency level, and target difficulty, and calls the AI ​​large model interface to generate new questions in batches, including question stems, options, answers, and explanations. After the new questions are automatically screened by the system, they enter the manual review process. The questions that pass the review are marked and dynamically added to the corresponding question bank, realizing the transformation of the question bank from static pre-set to dynamic growth.

[0015] Preferably, when the number of questions to be selected for the assessment is 30, 10 questions are selected from frequency levels 1 and 2 combined, 5 questions are selected from frequency levels 3 and 4 each, and 10 questions are selected from frequency level 5. When the number of questions to be drawn is 40, 10 questions are drawn from frequency levels 1 and 2 combined, 10 questions are drawn from frequency levels 3 and 4 each, and 10 questions are drawn from frequency level 5. When the number of questions drawn is 50, 10 questions are drawn from each of the frequency levels 1 to 5; Furthermore, frequency levels 1 and 2, 3 and 4 can be combined to form the smallest practice unit when the number of questions drawn is 30, and a single frequency level can be used as the smallest practice unit when the number of questions is an integer multiple of 10.

[0016] Preferably, the system adopts a many-to-many intelligent matching mode, which realizes the accurate matching of practice questions with users' knowledge needs through the dynamic question grouping module, realizes the quantitative assessment of knowledge mastery through the multi-dimensional assessment and analysis module, realizes the visualization of knowledge gaps through the knowledge graph highlighting module, and strengthens knowledge memory retention through the Ebbinghaus review reminder module.

[0017] A personalized knowledge graph learning method with dynamic weight rules includes the following steps: (1) Users log in to the system and select the assessment or after-class practice function; (2) The dynamic question generation module generates corresponding questions based on the function type, knowledge tag frequency level, and the user's existing knowledge range; (3) After the user completes the test, the multi-dimensional assessment and analysis module counts the number of correct answers for each question type, calculates the frequency level score and percentage score. And classify the levels of knowledge mastery; (4) The knowledge graph lighting module is based on And determine the level of knowledge mastery based on key conditions, and update the corresponding node color accordingly; (5) The system stores and records user assessment results, practice scores, and learning behavior data; (6) The Ebbinghaus review reminder module activates review reminders based on the user's answer time and pushes review notifications at preset time nodes.

[0018] The beneficial effects achieved by this invention are as follows: 1. This application constructs a dynamic question-assignment algorithm based on the frequency level of "knowledge tags" to achieve scientific allocation of practice questions, establishes a multi-dimensional practice result analysis algorithm model, classifies knowledge mastery levels by combining scores and key conditions, and visualizes the degree of knowledge mastery through a knowledge graph through set rules, allowing users to intuitively feel their current problems and make timely corrections. It also designs a review reminder mechanism based on the user group's behavioral habits and data and an optimized Ebbinghaus forgetting curve to improve the user's knowledge retention rate.

[0019] 2. This application significantly improves the accuracy of assessment results and more accurately identifies user knowledge gaps by introducing a dynamic question-based and multi-dimensional analysis model. The system can learn user behavior data in real time, automatically generate and optimize the configuration of personalized learning paths, replacing the limitations of traditional teachers' ability to provide personalized group tutoring in a programmatic manner. This achieves efficient allocation and intelligent scheduling of teaching resources, thereby greatly improving user knowledge retention rates and forming an intelligent teaching network from "one-to-many" to "many-to-many". By accurately pushing learning content and review nodes, the system significantly reduces users' ineffective learning time, improves learning efficiency, and indirectly reduces individual learning costs. Simultaneously, the system can serve a massive number of users on a large scale, reducing the teaching resource investment per user, and has good economic applicability and promotional value. This system is applicable to various online education scenarios and can effectively alleviate the reality of uneven distribution of educational resources. It provides sustainable, high-quality personalized learning support, especially for learners in remote areas and those lacking teachers, promoting educational equity and accessibility, enabling every user to obtain a learning experience similar to "one-on-one digital tutoring," and promoting the overall improvement of educational intelligence and social education levels. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the intelligent batch and targeted generation of questions according to the present invention.

[0021] Figure 2 This is a flowchart of the personalized learning path generation system of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims and drawings of this application are intended to cover non-exclusive inclusion.

[0024] The term "embodiment" as used herein means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of the phrase "embodiment" in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0025] This invention provides a personalized knowledge graph learning system and method based on dynamic weighting rules. The invention achieves personalized intelligent learning assistance through four core modules: a dynamic question-setting module, a multi-dimensional assessment and analysis module, a knowledge graph illumination module, and an optimized Ebbinghaus review reminder module. The technical solutions and functions of each module are as follows: (1) Dynamic Question Generation Module ① Minimum Practice Unit Question Grouping Rules: For a specific "knowledge tag" content, the system will randomly select ten questions from the corresponding question bank. If a video course has multiple "knowledge tags" content, then ten questions will be randomly selected from all "knowledge tags". The specific details of the question type settings are shown in Table 1 below (where the score corresponding to each question does not represent the score obtained by the user during practice, but is only used for backend algorithm calculation): Table 1: Question Type Proportion and Weight

[0026] Total score = (Number of correct answers to basic questions × 1) + (Number of correct answers to advanced questions × 2) + (Number of correct answers to intensive review questions × 3) Theoretically highest score = 5 × 1 + 3 × 2 + 2 × 3 = 17 points ② Assessment Question Selection Rules: For the "Initial Assessment," a frequency distribution sampling method is used, extracting questions based on the five levels of "exam frequency" across the entire subject. In contrast, for the "Assessment During the Learning Process," question selection is limited to knowledge previously learned by the user. These two assessment formats differ in their content sources, thus differentiating them and allowing for personalized settings. The correspondence between user assessment results and scores, as well as the question selection rules and coverage for each level, conform to the requirements in Table 2 below (applicable to: Initial Assessment, Assessment During the Learning Process): Table 2: Exam Frequency and Question Composition Rules

[0027] When there are 30 questions, 5 questions are randomly selected from each of the frequency levels 1 and 2, and 3 and 4, and combined to form a "minimum practice unit" of 10 questions. A level with a number of questions that is an integer multiple of 10 can be used as a "minimum practice unit" on its own, and the corresponding judgment rules and criteria can be used to analyze the results. The same rule applies to the subsequent 40 and 50 questions.

[0028] (2) Multi-dimensional assessment and analysis module ① Score Calculation Model: Based on the "basic question-setting rules," the score corresponding to each test frequency can be obtained. The percentage between the overall average score and the total score can be mapped to a percentage score, which can be used as the score obtained by the user when answering questions. The specific formula is as follows: (1) simplify: (2) in, : The percentage score directly mapped from the average score of the test frequency to the total score; n: The number of test frequency levels (5 levels); : Score of the i-th frequency level; M: Full score of the question corresponding to a single frequency level ( M =17 points); ② Knowledge Mastery Level Classification Rules: After obtaining a total score, the user's current level of understanding of the content can be determined based on the classification levels, resulting in Table 3 below: Table 3: Evaluation Results and Analysis

[0029] The user's results are graded into four levels: "Weak", "Pass", "Good", and "Excellent". The result of the "score range" calculated in the algorithm is used as the basis for judgment, and it is directly converted into a percentage score, which is the user's actual score on the interactive interface. The results and the corresponding "ability descriptions" are analyzed to draw the assessment conclusions. Based on the strict division of the test questions, the user's current ability in terms of mastering the basic content, comprehensive application, and complex calculations in this subject can be determined.

[0030] (3) Knowledge Graph Illumination Module ① Rules for setting up practice questions: There are 20 fixed practice questions after class. The entire practice set, consisting of multiple or one "smallest practice unit", is bound to the "smallest lit unit" of the knowledge graph. Conversely, one "smallest lit unit" corresponds to one or more courses.

[0031] ② Weighted Score Calculation: If a minimum lighting unit corresponds to multiple video courses, i.e., multiple exercises, then the weighted score of each exercise constitutes the final lighting score. Below is the specific formula for the current situation:

[0032] Where m is the number of exercises. For the score of the k-th exercise, For the knowledge tag set of the k-th exercise, Let j be the frequency value of knowledge tag j, and n be the total number of knowledge tags; ③ The rules for illuminating the graph are shown in Table 4: Table 4: Rules for Judging the Level of Mastery

[0033] After each practice exercise, the results are analyzed and fed back into the "knowledge graph," with different colors corresponding to different levels of mastery. If the score meets the weighted score range but does not meet the corresponding "key conditions," then the system automatically determines whether the next level of mastery is met. For example, if the score is greater than 12 points but does not meet the subsequent "key conditions," then the system automatically determines whether the score meets the standard for the next level, 5-11 points, regardless of whether the subsequent "key conditions" are met. (4) Ebbinghaus Review Reminder Module After a user creates a study record, the system automatically activates a review reminder, as detailed in Table 5: Table 5: Optimized Review Timeline

[0034] It accurately matches the optimized Ebbinghaus forgetting curve time points to help users strengthen their memory.

[0035] (5) Automatic Question Generation (with Manual Review and Intervention): The system calls the AI ​​large model interface, providing the large model with structured parameters such as knowledge point descriptions, exam frequency levels, and target difficulty (basic / advanced / intensive). The large model then generates new questions that meet the requirements in real time (including question stems, options, answers, and explanations). After preliminary review and verification, the generated new questions can be dynamically added to the corresponding question bank. The innovation of this module lies in the deep integration of generative artificial intelligence with personalized learning systems, realizing a paradigm shift from "static pre-set" to "dynamic growth" of the question bank. It can not only greatly enrich the question bank resources and reduce the cost of manual question creation, but also build a self-evolving learning environment that becomes "smarter with use".

[0036] (6) Analysis of the effectiveness of the technical solution ① The dynamic question-setting module solves the problem of insufficient targeting of existing systems by strictly controlling the frequency level and difficulty ratio of test questions. Furthermore, subsequent assessments will be limited to the learning data that users have studied, ensuring that the assessments can accurately reflect the user's mastery of core knowledge points. ② The multi-dimensional assessment and analysis module combines scores and key conditions to classify levels, and is coupled with a knowledge graph visualization of mastery level, which solves the problem of vague feedback on knowledge mastery level, allowing users to quickly identify weak areas; ③ The optimized Ebbinghaus review reminder module, based on user behavior habits, sets nodes according to the scientific forgetting curve, solving the problem of unscientific review mechanisms and significantly improving the retention rate of user knowledge.

[0037] The following section uses a "Computer Fundamentals Course" as an example to explain in detail the implementation process of this invention: (1) Overview of the overall technical solution This embodiment targets the "Computer Fundamentals" course. It generates assessment and after-class practice questions through a dynamic question-generating module, calculates scores and classifies mastery levels through a multi-dimensional and multi-weighted calculation and analysis module, updates the graph colors through a knowledge graph highlighting module, and finally pushes review nodes through an Ebbinghaus review reminder module.

[0038] (2) Design concept and technical features ① Core program flow: User enters the system → Triggers the initial assessment / practice → Dynamic question generation module generates questions → User answers questions → Multi-dimensional assessment and analysis module calculates scores and levels → Knowledge graph highlighting module updates colors → System records learning behavior → Ebbinghaus review reminder module activates reminders.

[0039] ② Key nodes in the software flowchart: Node 1 (Question Trigger): After the user selects "Assessment" or "Practice", the system calls the dynamic question generation module to read the frequency data of knowledge tags related to "Computer Basics Course". If it is a subsequent "Assessment in the learning process", then the basis for question extraction is limited to the high-frequency "knowledge tags" of the content that the user has learned.

[0040] Node 2 (Question Generation): If it is an assessment (30 questions), questions will be drawn according to the rule of "10 questions of level 5 frequency, 5 questions of level 4 frequency, 5 questions of level 3 frequency, and 10 questions of level 1+2 frequency". The difficulty of the questions will be distributed according to "50% basic questions, 30% advanced questions, and 20% sprint questions", specifically according to the content of "Table 2 Examination Frequency and Question Grouping Rules"; Node 3 (Score Calculation): After the user completes the test, the system first counts the number of correct answers for each question type, calculates the frequency level score, and then calculates the percentage score using formula (1). ; Node 4 (Level and Map Illumination Update): If = 85 points (corresponding to an algorithm score of 14 points), according to the calculation formula (3), it can be judged as "excellent", and the corresponding node in the knowledge graph is lit up in green; if = 45 points (corresponding to an algorithm score of 7 points), then it is judged as "passing", and the corresponding node is lit up in yellow; Node 5 (Review Reminder Activation): The system records the user's answer time and pushes review reminders according to the "Review Time Nodes after Optimization in Table 5".

[0041] (3) Automatic generation of question bank content (with manual review and intervention): This invention relates to an artificial intelligence-based method for automatically generating and optimizing question bank content. This solution achieves intelligent construction of the question bank content from static preset to dynamic growth and continuous evolution through deep integration of the AI ​​large-scale model API with the system. The execution logic of this method is as follows: First, the system monitors the status of the question bank content and user learning behavior data in real time. When it identifies insufficient coverage of questions for a specific knowledge point or when user practice demand for that knowledge point is approaching saturation, it automatically triggers the question generation process. The system structurally analyzes the generation requirements, extracting multiple parameters including knowledge point description, test frequency level, and target difficulty. During the question generation phase, the system calls a third-party AI large-scale model interface and integrates multi-dimensional data sources for batch generation: first, existing questions in the question bank and their associated knowledge system data; second, a database of user question-answering behavior built based on historical user response records. This database, after analysis algorithms, covers deep learning features such as knowledge point mastery, typical error patterns, and high-frequency error tags for "knowledge tags." Based on the above data, the AI ​​model generates question content and explanations that meet the parameter requirements and extracts one or more "knowledge tags" for the corresponding questions. Figure 1 As shown: The generated questions undergo an automatic initial screening process by the system through verification of the questions and answers, before proceeding to the manual review stage. Questions that pass the review are automatically tagged with knowledge labels and added to the database, which is also updated synchronously to the dynamic question bank for use by the test paper generation module.

[0042] The entire process forms a closed-loop feedback mechanism of "monitoring-generation-review-database entry-application", which realizes continuous self-optimization of the question bank content in terms of structure, quality and adaptability, and reflects the innovative breakthrough of artificial intelligence in the dynamic generation and personalized adaptation of educational resources.

[0043] (4) Step description (in process order), such as Figure 2 : Step 1: The user logs into the system, clicks "Computer Basics Courses", selects the "Assessment" function, and the system front-end sends an "Assessment Question Grouping Request" to the back-end dynamic question grouping algorithm module; Step 2: After receiving the request, the dynamic question-setting module accesses the database to read the exam frequency data of the knowledge points of the "Computer Fundamentals Course" (e.g., exam frequency level 5: Computer System Composition, exam frequency level 4: Von Neumann Principle, exam frequency level 3: System Software, exam frequency level 2: Brief History of Computer Development, exam frequency level 1: Types of Computers). Step 3: This module combines questions according to the corresponding question rules: After the question extraction rules in "Table 2 Examination Frequency and Question Combination Rules" are completed by the backend algorithm, the question list is generated and returned to the frontend; Step 4: The user completes the questions and submits them. The system calculates the accuracy rate of each question type and uses formula (3) to calculate the grade score for this exercise. Step 5: The system calculates the percentage score. The scores will be assigned to the "passing" grade according to the rules in "Table 3 Assessment Results and Analysis"; Step 6: Knowledge Graph Illumination Module Reading Based on the key conditions, the "Computer System Composition" node is determined to meet the "Familiarity" level and is highlighted in yellow; Step 7: The system records the user's answers and evaluation analysis results, activates the Ebbinghaus review reminder module, and sets reminder nodes according to "Table 5 Optimized Review Time Nodes".

[0044] The embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A personalized knowledge graph learning system with dynamic weighting rules, characterized in that: The system includes: The dynamic question generation module is used to generate personalized question sets based on the frequency level of knowledge tags and user learning data. The question types include basic questions, intermediate questions, and intensive questions, and are allocated according to a preset ratio. The multi-dimensional assessment and analysis module is used to analyze users' answers, classify knowledge mastery levels through a scoring calculation model, and map the results into visual feedback. The knowledge graph highlighting module is used to calculate a weighted score based on the assessment results and update the display status of knowledge points in the knowledge graph in combination with key conditions, using different colors to indicate the degree of mastery; The Ebbinghaus review reminder module is used to automatically activate and push personalized review reminders based on the optimized forgetting curve time nodes; The automatic question generation module calls the AI ​​big model interface and provides the AI ​​big model with structured parameters such as knowledge point description, test frequency level, and target difficulty. The AI ​​big model then generates new questions that meet the requirements in real time.

2. The personalized knowledge graph learning system with dynamic weight rules according to claim 1, characterized in that: The dynamic question-generating module's question-generating rules include: dividing the practice into the smallest units based on knowledge tags, with 10 questions selected from each unit. The question types are configured in a ratio of 50% basic questions, 30% advanced questions, and 20% intensive questions, with corresponding weight scores of 1, 2, and 3 respectively. The total score is calculated as follows: Total Score = (Number of Correct Basic Questions × 1) + (Number of Correct Advanced Questions × 2) + (Number of Correct Intensive Questions × 3), with a theoretical maximum score of 17 points. The assessment questions are divided into initial assessment and assessment during the learning process. The initial assessment draws questions based on the frequency of the subject in five levels. The assessment during the learning process is limited to drawing questions from the knowledge content that the user has already learned. The frequency level allocation rules are met when the number of questions drawn meets the requirements of 30, 40 or 50 questions.

3. The personalized knowledge graph learning system with dynamic weight rules according to claim 1, characterized in that: The multi-dimensional assessment and analysis module uses a score calculation model and mastery level classification rules, wherein the score calculation model is as follows: in, It is a percentage score. Let M = 17 be the score for the i-th frequency level, and M = 17 be the full score for the question corresponding to a single frequency level. The mastery level is divided into four levels: weak, passable, good, and excellent, which correspond to an algorithm score of ≤5 points, 5 < score ≤8 points, 8 < score ≤13 points, and 13 < score ≤17 points, respectively. Each level has a clear description of the ability.

4. The personalized knowledge graph learning system with dynamic weight rules according to claim 1, characterized in that: The working mechanism of the knowledge graph lighting module includes: the after-class exercises are fixed at 20 questions, consisting of one or more minimum exercise units, and are bound to the minimum lighting unit of the knowledge graph; the weighted score calculation formula is as follows: Where m is the number of exercises. For the score of the k-th exercise, For the knowledge tag set of the k-th exercise, Let j be the frequency value of knowledge tag j, and n be the total number of knowledge tags; Knowledge graph nodes are presented in different colors according to the level of mastery. Dark green indicates mastery, with a score of ≥12 points and ≥4 correct answers on basic questions and ≥2 correct answers on advanced questions. Dark yellow indicates familiarity, with a score of 5 < ≤11 points and ≥3 correct answers on basic questions. Gray indicates understanding, with a score of ≤5 points. White indicates the initial state of not having started learning. If the score range is met but the key conditions are not met, the node will be automatically downgraded.

5. A personalized knowledge graph learning system with dynamic weighting rules according to claim 1, characterized in that: The optimized Ebbinghaus forgetting curve review reminder module is automatically activated after the user generates a learning record. The review time nodes are set according to the time since the first learning: the first review reminder is after 2 days, the second after 6 days, the third after 15 days, and the fourth after 21 days, which accurately matches the optimized Ebbinghaus forgetting curve.

6. The personalized knowledge graph learning system with dynamic weight rules according to claim 1, characterized in that: The workflow of the automatic question generation module is as follows: The system monitors the question bank status and user learning data in real time. When it detects that the coverage of questions for a specific knowledge point is insufficient or the practice demand is saturated, the question generation process is triggered. The system extracts structured parameters such as knowledge point description, test frequency level, and target difficulty, and calls the AI ​​large model interface to generate new questions in batches, including question stems, options, answers, and explanations. After the new questions are automatically screened by the system, they enter the manual review process. The questions that pass the review are marked and dynamically added to the corresponding question bank, realizing the transformation of the question bank from static pre-set to dynamic growth.

7. A personalized knowledge graph learning system with dynamic weighting rules according to claim 2, characterized in that: When the number of questions to be selected for the assessment is 30, 10 questions are selected from frequency levels 1 and 2 combined, 5 questions are selected from frequency levels 3 and 4 each, and 10 questions are selected from frequency level 5. When the number of questions drawn is 40, 10 questions are drawn from frequency levels 1 and 2 combined, 10 questions are drawn from frequency levels 3 and 4 each, and 10 questions are drawn from frequency level 5. When the number of questions drawn is 50, 10 questions are drawn from each of the frequency levels 1 to 5; Furthermore, frequency levels 1 and 2, 3 and 4 can be combined to form the smallest practice unit when the number of questions drawn is 30, and a single frequency level can be used as the smallest practice unit when the number of questions is an integer multiple of 10.

8. A personalized knowledge graph learning system with dynamic weighting rules according to claim 2, characterized in that: The system adopts a many-to-many intelligent matching mode, which realizes the precise matching of practice questions with users' knowledge needs through the dynamic question grouping module, realizes the quantitative assessment of knowledge mastery through the multi-dimensional assessment and analysis module, realizes the visualization of knowledge gaps through the knowledge graph highlighting module, and strengthens knowledge memory retention through the Ebbinghaus review reminder module.

9. A personalized knowledge graph learning method with dynamic weight rules, characterized in that: Includes the following steps: (1) Users log in to the system and select the assessment or after-class practice function; (2) The dynamic question generation module generates corresponding questions based on the function type, knowledge tag frequency level, and the user's existing knowledge range; (3) After the user completes the test, the multi-dimensional assessment and analysis module counts the number of correct answers for each question type, calculates the frequency level score and percentage score. And classify the levels of knowledge mastery; (4) The knowledge graph lighting module is based on And determine the level of knowledge mastery based on key conditions, and update the corresponding node color accordingly; (5) The system stores and records user assessment results, practice scores, and learning behavior data; (6) The Ebbinghaus review reminder module activates review reminders based on the user's answer time and pushes review notifications at preset time nodes.

Images