Intelligent adaptive personalized learning method and system based on AIGC and knowledge graph

Abstract

The invention discloses an intelligent adaptive personalized learning method and system based on AIGC and a knowledge graph. The technical problems that an existing online learning system is low in individuation degree, a knowledge system is fragmented, the evaluation dimension is single, the adaptive capacity is weak, and different requirements of teachers and students are not distinguished are solved. The method comprises the following steps: constructing and updating a domain knowledge graph library with weights, establishing a self-adaptive course resource library associated with knowledge entities, respectively generating multi-dimensional portraits of teachers and students, generating differentiated personalized learning / teaching paths based on an A * search algorithm, and realizing self-adaptive content recommendation, continuous self-adaptive evaluation and dynamic adjustment through a mixed recommendation algorithm. A closed-loop intelligent ecology is formed; the corresponding system comprises core modules such as a knowledge graph library, a self-adaptive course resource library and a teacher and student portrait module. According to the method, all-dimensional personalized support of teachers and students, dynamic self-adaptive adjustment and resource generation according to needs are realized, the teaching and learning efficiency is remarkably improved, and the method is suitable for practical subjects and has system self-evolution ability.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of smart education technology and software engineering education, and in particular to an intelligent adaptive personalized learning method and system for software engineering courses that integrates generative artificial intelligence (AIGC) and knowledge graphs, which can simultaneously meet the differentiated needs of students' learning and teachers' teaching. Background Technology

[0002] Currently, while online education platforms and traditional learning management systems (LMS) have made numerous attempts at personalized learning, significant shortcomings remain. Mainstream technical solutions include rule-based recommendation systems, collaborative filtering recommendation systems, static knowledge graph applications, and simple adaptive testing modes.

[0003] Rule-based recommendation systems push fixed sequences of learning resources based on learners' preset tags or simple test scores, resulting in low personalization. Collaborative filtering recommendation systems borrow from e-commerce recommendation approaches, recommending resources based on the learning behaviors of similar groups, but cannot accurately match individual real-time needs. Static knowledge graph applications can display the relationships between knowledge points, but the path planning is static and linear, lacking dynamic adaptability. Simple adaptive testing modes are only adaptable to local and phased situations, and cannot continuously track learners' multidimensional abilities.

[0004] The core shortcomings of existing technologies include: superficial personalization, lacking multi-dimensional assessment of soft skills such as practical abilities and communication and collaboration; rigid learning paths, unable to dynamically adjust according to learners' real-time status, stemming from the lack of an intelligent engine; separation of content generation and recommendation, relying on pre-built libraries, making it difficult to cope with cold start and long-tail needs, and unable to generate targeted resources in real time; a disconnect between theory and practice, making it difficult to meet the learning needs of practical disciplines such as software engineering; and failure to differentiate between the needs of students and teachers, with the system focusing only on student learning support and lacking personalized support for the entire teaching process for teachers, such as failing to generate customized teaching paths based on teachers' teaching styles and class learning situations, and failing to provide exclusive resources and tools adapted to teaching scenarios, resulting in insufficient synergy between teaching and learning.

[0005] The core reason for the aforementioned shortcomings lies in the isolation of modules within the existing technical architecture, lacking a closed-loop adaptive system driven by dynamic, fine-grained teacher and student profiles, integrating knowledge structuring and intelligent content generation. Specifically, this manifests as: a lack of structured and semantic modeling of domain knowledge; a lack of comprehensive and dynamic characterization of student states, and the absence of a multi-dimensional profile specifically for teachers; insufficient intelligence in the decision engine, failing to design differentiated path generation and content recommendation logic for the different needs of teachers and students; and the lack of a collaborative evaluation and adjustment mechanism for teachers and students. Summary of the Invention

[0006] This invention aims to overcome the shortcomings of existing technologies and solve the technical problems of low personalization, fragmented knowledge systems, single assessment dimensions, weak system adaptability, and failure to differentiate between the needs of teachers and students in online learning. It provides an intelligent adaptive personalized learning method and system that can dynamically generate and optimize personalized learning paths for students and generate customized teaching paths for teachers, enabling multi-dimensional assessment of students' abilities and optimization of teachers' teaching capabilities. Based on knowledge graphs, it ensures the coherence of the knowledge system and meets the collaborative needs of teachers and students.

[0007] To achieve the above objectives, the core lies in constructing a closed-loop intelligent system that links teacher and student profiles, knowledge graphs, and AIGC (AI-GC) to provide differentiated and personalized support for teachers and students. Specifically, this is achieved as follows: An intelligent adaptive personalized learning method based on AIGC and knowledge graphs includes the following steps: S1. Constructing and updating the domain knowledge graph: Extract core concepts and skill points as knowledge entities from curriculum standards, authoritative textbooks, and industry standards; define prerequisite, dependency, association, and inclusion relationships between entities; assign weights to entities to represent importance, difficulty level, or association strength; and construct a structured weighted domain knowledge graph. Update the knowledge graph using natural language processing tools and expert review systems, combined with course outlines, textbooks, academic papers, open-source code repositories, and project case documents. S2. Establish and enrich the adaptive course resource library: Precisely anchor various learning resources such as videos, documents, code examples, interactive experiments, and assessment questions to knowledge entities in the knowledge graph. Assign a resource ID to each resource, label the resource's difficulty level, estimated learning time, and associated knowledge entities, and establish a bidirectional index for all resources and knowledge entities. Integrate AIGC tools to automatically generate personalized resources in the form of text, code examples, diagrams, or simulated dialogues when the resource library lacks resources for a specific difficulty, format, or style of a knowledge point. At the same time, classify and store learning resources for students and teaching resources for teachers. Teaching resources include syllabus templates, knowledge point explanation courseware, learning analysis tools, homework design cases, and teaching assessment scales. S3. Generate and update multi-dimensional profiles of students and teachers respectively: Multidimensional Student Profile: This involves calculating students' theoretical mastery of various knowledge points through online tests and practice questions combined with Project Response Theory or Knowledge Tracking Models; evaluating submitted code or projects to obtain practical ability scores by integrating online IDEs, automated testing frameworks, and static code analysis tools; assessing students' communication, collaboration, and project management soft skills through AIGC-driven simulated dialogue scenarios, virtual team projects, and peer review; and aggregating the above assessment data and learning behavior data to form a dynamically updated multidimensional digital profile of students, including student ID, theoretical knowledge mastery vector, practical ability score matrix, soft skill indicators, learning style preference vector, and learning history sequence. Multi-dimensional teacher profile: The system assesses the suitability of teaching content through teachers' submitted syllabi, courseware, and assignment designs; evaluates the effectiveness of teaching methods through student learning feedback, teaching assessment data, and peer review; analyzes teaching style preferences through teachers' system usage data; and aggregates the above data to form a dynamically updated multi-dimensional digital profile of teachers, including teacher ID, teaching content suitability index, teaching method effectiveness score, teaching style preference vector, teaching history, and subject expertise tags. S4. Generate differentiated and personalized paths: Personalized Learning Path for Students: A learning path generation engine is built based on the A search algorithm. The learning path planning is modeled as a problem of finding the path with the minimum cognitive cost from the student's current state to the target state on the domain knowledge graph. Open and closed lists are used to manage nodes to be checked and checked respectively. The path cost is calculated using the cost function f(n) = g(n) + h(n), where g(n) is the actual cost from the starting point to node n, calculated as g(n) = Σ(inherent difficulty of knowledge point i (1 - student's mastery of knowledge point i)), and h(n) is the estimated cost from node n to the target point, i.e., the average difficulty of the knowledge points involved in the shortest path from node n to the target. The system takes the student's current state vector, the target job / skill set, the domain knowledge graph, and algorithm weight parameters as input and outputs the optimal learning sequence. Personalized teaching paths for teachers: Based on domain knowledge graphs and multi-dimensional teacher profiles, combined with the overall learning profiles of students in the classes taught, personalized teaching paths are generated, including the allocation of teaching focus, recommendation of teaching resources, design of teaching activities, and learning monitoring nodes; input teacher profiles, class learning data, course objectives, and knowledge graphs, and output teaching progress planning, phased teaching focus, and appropriate teaching method suggestions; S5, Adaptive Content Recommendation: Student Content Recommendation: An adaptive content recommendation engine is constructed based on a hybrid recommendation algorithm of content filtering, context awareness, and ε-greedy strategy. For each knowledge point in the learning path, resource content features, student profile features, and context information features are extracted to calculate the comprehensive matching score S: S = W1 * S_content + W2 * S_context + W3 * S_popularity, where S_content is the cosine similarity between the resource features and the student profile, S_context is the context matching degree considering device type, available time, and current learning status, S_popularity is the global average score and completion rate of the resource, and W1, W2, and W3 are weight coefficients. An ε-greedy strategy is adopted for decision-making, selecting the resource with the highest comprehensive matching score with a high probability (1-ε) and randomly selecting other resources with a low probability (ε) to discover potential high-quality resources and collect data. Teacher content recommendation: Based on teaching paths, teacher profiles, and class learning conditions, the system recommends suitable teaching resources, learning analysis tools, and teaching adjustment suggestions; when resources for specific teaching scenarios are lacking, the AIGC tool automatically generates teaching syllabus templates, classroom interaction scripts, or personalized learning reports. S6. Adaptive Evaluation and Dynamic Adjustment: Student-side adjustments: Continuously monitor students' real-time behavioral data stream and micro-assessment results, and set mastery thresholds, answer error rate thresholds, and page dwell time thresholds; when abnormal situations are detected, such as students' knowledge mastery not meeting expectations, a sudden increase in answer error rate, or excessively long page dwell time, a dynamic adjustment mechanism is triggered, including retrospectively recommending previous knowledge points for reinforcement, recommending alternative explanatory resources, inserting targeted exercises generated by AIGC, or re-calling the personalized learning path generation engine to plan subsequent paths; Teacher-side adjustments: Continuously collect student learning feedback data and teaching effectiveness evaluation results, combined with teacher's teaching behavior records; when common weaknesses in the class are not improved, teaching resources are poorly matched, or student learning participation is insufficient, a dynamic adjustment mechanism is triggered on the teacher side, including recommending optimized teaching resources, adjusting the allocation of teaching priorities, providing suggestions for differentiated teaching methods, or regenerating personalized teaching paths.

[0008] Furthermore, the student multidimensional profile in step S3 includes multiple dimensions such as theoretical knowledge mastery, practical ability, communication and collaboration ability, project management ability, and learning style preference. The teacher multidimensional profile includes multiple dimensions such as teaching content suitability, teaching method effectiveness, teaching style preference, and subject expertise. Both types of profiles are updated in real time with the corresponding behavioral data and evaluation results.

[0009] Furthermore, the A* search algorithm in step S4 optimizes node search efficiency through open and closed lists, the weight parameters in the student path cost function can be dynamically adjusted according to subject characteristics, course difficulty and student learning goals, and the teacher path generation can adjust the weight of teaching focus according to class size and differences in student foundation.

[0010] Furthermore, in step S5, the exploration probability ε in the ε-greedy strategy ranges from 0.05 to 0.2. The student end can dynamically adjust the probability based on the resource library update frequency and the completeness of the student profile, while the teacher end can dynamically adjust the probability based on the teaching resource update frequency, the completeness of the teacher profile, and changes in the class's learning situation.

[0011] To achieve the above objectives, the present invention also provides an intelligent adaptive personalized learning system, comprising: Knowledge graph library: Stores structured domain knowledge entities and their relationships, including the importance, difficulty coefficient or association strength weight information of knowledge entities, providing structured knowledge support for intelligent decision-making in the system; Adaptive Course Resource Library: Stores learning resources such as videos, documents, code examples, interactive experiments, and assessment questions, as well as teaching resources such as syllabus templates and learning analysis tools. Each resource is associated with a corresponding knowledge graph entity and has resource metadata annotation function. It works in conjunction with the AIGC content generation and enhancement module to dynamically supplement resources and classify and manage learning resources and teaching resources. The teacher and student profile module integrates an online assessment system, an integrated development environment, a project collaboration platform, a behavior data collection SDK, and teaching evaluation tools. It continuously collects students' theoretical assessment data, practical operation data, soft skills assessment data, and learning behavior data, as well as teachers' teaching content data, teaching effectiveness assessment data, peer review data, and teaching behavior data. It calculates and maintains multi-dimensional dynamic profiles of students and teachers. AIGC Content Generation and Enhancement Module: Based on the resource gaps in the adaptive course resource library, students' weaknesses, teachers' teaching needs and personalized profiles, it dynamically generates learning content such as text, code examples, diagrams, and simulated dialogues, as well as teaching content such as syllabus templates and learning reports, providing support for adaptive content recommendation. Personalized Path Generation Engine: Includes a student learning path generation sub-engine and a teacher teaching path generation sub-engine; the student sub-engine uses the knowledge graph-based A* search algorithm, receives the student's current state vector, the target job / skill set, and knowledge graph database data, and generates the optimal learning path; the teacher sub-engine combines the knowledge graph, teacher profile, and class learning data to generate personalized teaching paths; The adaptive content recommendation engine includes a student content recommendation sub-engine and a teacher content recommendation sub-engine. The student sub-engine uses a hybrid recommendation algorithm to receive students' personalized learning paths, student profiles, and learning resource data, and recommends or generates suitable learning resources. The teacher sub-engine recommends or generates suitable teaching resources based on teachers' teaching paths, teacher profiles, class learning conditions, and teaching resource data. The adaptive assessment and dynamic adjustment module includes a student-side adjustment submodule and a teacher-side adjustment submodule. The student-side submodule collects students' learning behavior data and micro-assessment results in real time, compares them with preset thresholds, and triggers the student-side adjustment mechanism. The teacher-side submodule collects class learning data, teaching effectiveness evaluation results, and teacher teaching behavior data in real time, compares them with preset thresholds, and triggers the teacher-side adjustment mechanism. Both types of submodules update corresponding profiles, paths, and recommended resources to form a closed-loop feedback. User interface: Includes student and teacher interfaces; the student interface provides functions such as personal profile display, learning path viewing, resource learning, assessment submission, and feedback interaction; the teacher interface provides functions such as personal profile display, teaching path viewing, teaching resource access, student learning analysis viewing, teaching adjustment operations, and feedback interaction.

[0012] Furthermore, the knowledge graph database implements an update mechanism that combines regular automatic updates with manual optimization through natural language processing tools and an expert review system.

[0013] Furthermore, the teacher and student profile module is linked with the adaptive assessment and dynamic adjustment module to realize real-time updates of student and teacher profiles, providing dynamic data support for the generation of student learning paths, teacher teaching paths and corresponding content recommendations.

[0014] Beneficial effects: (1) Differentiated and personalized teaching and learning across all dimensions: It provides students with comprehensive learning support covering theory, practice, and soft skills, and provides teachers with personalized teaching support that adapts to teaching styles and class learning conditions, achieving precise adaptation between teaching and learning; (2) Two-way dynamic adaptive adjustment: Both student learning paths and teacher teaching paths can be intelligently adjusted based on real-time feedback, and the collaborative optimization of teacher and student paths can be achieved through class learning conditions, improving the collaborative efficiency of teaching and learning; (3) Two-way resource generation and adaptation capabilities: AIGC technology simultaneously meets the needs of students' personalized learning resources and teachers' teaching resources for on-demand generation, covering scenarios such as cold start and long-tail needs, with stronger resource adaptability; (4) Significantly improve teaching and learning efficiency: Student paths directly address weaknesses. Weak links, teacher paths focus on teaching priorities and common problems in the class, and work together to shorten the learning cycle and optimize teaching effect; (4) Adapt to the teaching needs of practical disciplines: For disciplines such as software engineering, realize the precise correlation between students' programming errors and theoretical shortcomings, and at the same time provide teachers with special resources for practical teaching and learning analysis tools to strengthen the combination of theory and practice; (5) System collaborative self-evolution capability: As teachers and students accumulate data, the system can continuously optimize knowledge graphs, resource bases and algorithm models, and improve teacher and student profiles, so that personalized support for teaching and learning becomes more and more accurate with use; (5) Strengthen teacher and student collaborative interaction: Through class learning linkage and teaching effect feedback loop, build a bridge for teacher and student collaboration to solve the problem of disconnect between teaching and learning in traditional systems. Attached Figure Description

[0015] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart of the intelligent adaptive personalized learning method based on AIGC and knowledge graphs as described in an embodiment of the present invention; Figure 2 This is a framework diagram of the intelligent adaptive personalized learning system based on AIGC and knowledge graphs as described in an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0017] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0018] Example 1 See Figure 1 A smart adaptive personalized learning method based on AIGC and knowledge graphs includes the following steps: S1. Constructing and Updating Domain Knowledge Graph: Extract core concepts and skill points as knowledge entities from curriculum standards, authoritative textbooks, and industry standards; define prerequisite, dependency, association, and inclusion relationships between entities; assign weights to entities to represent importance, difficulty level, or association strength; and construct a structured weighted domain knowledge graph. Update the knowledge graph using natural language processing (NLP) tools and an expert review system, combined with course outlines, textbooks, academic papers, open-source code repositories, and project case documents. S2. Establish and enrich the adaptive course resource library: Precisely anchor various learning resources such as videos, documents, code examples, interactive experiments, and assessment questions to knowledge entities in the knowledge graph. Assign a resource ID to each resource, label the resource's difficulty level, estimated learning time, and associated knowledge entities, and establish a bidirectional index for all resources and knowledge entities. Integrate AIGC tools to automatically generate personalized resources in the form of text, code examples, diagrams, or simulated dialogues when the resource library lacks resources for a specific difficulty, format, or style of a knowledge point. At the same time, classify and store learning resources for students and teaching resources for teachers. Teaching resources include syllabus templates, knowledge point explanation courseware, learning analysis tools, homework design cases, and teaching assessment scales. S3. Generate and update multi-dimensional profiles of students and teachers respectively: Multidimensional Student Profile: This involves calculating students' theoretical mastery of various knowledge points through online tests and practice questions combined with Project Response Theory or Knowledge Tracking Models; evaluating submitted code or projects to obtain practical ability scores by integrating online IDEs, automated testing frameworks, and static code analysis tools; assessing students' communication, collaboration, and project management soft skills through AIGC-driven simulated dialogue scenarios, virtual team projects, and peer review; and aggregating the above assessment data and learning behavior data to form a dynamically updated multidimensional digital profile of students, including student ID, theoretical knowledge mastery vector, practical ability score matrix, soft skill indicators, learning style preference vector, and learning history sequence. Multi-dimensional teacher profile: The system assesses the suitability of teaching content through submitted syllabi, courseware, and assignment designs; evaluates the effectiveness of teaching methods through student learning feedback, teaching assessment data, and peer review; analyzes teaching style preferences through teacher usage data (such as resource usage frequency and teaching adjustment records); and aggregates the above data to form a dynamically updated multi-dimensional digital profile of teachers, including teacher ID, teaching content suitability index, teaching method effectiveness score, teaching style preference vector, teaching history, and subject expertise tags. S4. Generate differentiated and personalized paths: Personalized Learning Path for Students: A learning path generation engine is built based on the A search algorithm. The learning path planning is modeled as a problem of finding the path with the minimum cognitive cost from the student's current state to the target state (such as achieving job skills or completing course learning) on ​​the domain knowledge graph. Open and closed lists are used to manage nodes to be checked and checked respectively. The path cost is calculated using the cost function f(n) = g(n) + h(n), where g(n) is the actual cost from the starting point to node n, calculated as g(n) = Σ(inherent difficulty of knowledge point i (1 - student's mastery of knowledge point i)), and h(n) is the estimated cost from node n to the target point, i.e., the average difficulty of the knowledge points involved in the shortest path from node n to the target. The system takes the student's current state vector, the target job / skill set, the domain knowledge graph, and algorithm weight parameters as input and outputs the optimal learning sequence. Personalized teaching paths for teachers: Based on domain knowledge graphs and multi-dimensional teacher profiles, combined with the overall learning profiles of students in the classes taught (average class mastery, common weak knowledge points, and overall learning styles), personalized teaching paths are generated, including the allocation of teaching focus, recommendation of teaching resources, design of teaching activities, and learning monitoring nodes; input teacher profiles, class learning data, course objectives, and knowledge graphs, and output teaching progress plans, phased teaching focus, and appropriate teaching method suggestions; S5, Adaptive Content Recommendation: Student Content Recommendation: An adaptive content recommendation engine is constructed based on a hybrid recommendation algorithm of content filtering, context awareness, and ε-greedy strategy. For each knowledge point in the learning path, resource content features, student profile features, and context information features are extracted to calculate the comprehensive matching score S: S = W1 * S_content + W2 * S_context + W3 * S_popularity, where S_content is the cosine similarity between the resource features and the student profile, S_context is the context matching degree considering device type, available time, and current learning status, S_popularity is the global average score and completion rate of the resource, and W1, W2, and W3 are weight coefficients. An ε-greedy strategy is adopted for decision-making, selecting the resource with the highest comprehensive matching score with a high probability (1-ε) and randomly selecting other resources with a low probability (ε) to discover potential high-quality resources and collect data. Teacher content recommendation: Based on teaching path, teacher profile and class learning situation, recommend suitable teaching resources (such as courseware explaining common weaknesses in the class, homework design cases that match the teacher's style), learning situation analysis tools and teaching adjustment suggestions; when resources for specific teaching scenarios are lacking, the AIGC tool automatically generates teaching syllabus templates, classroom interaction scripts or personalized learning situation reports. S6. Adaptive Evaluation and Dynamic Adjustment: Student-side adjustments: Continuously monitor students' real-time behavioral data stream and micro-assessment results, and set mastery thresholds, answer error rate thresholds, and page dwell time thresholds; when abnormal situations are detected, such as students' knowledge mastery not meeting expectations, a sudden increase in answer error rate, or excessively long page dwell time, a dynamic adjustment mechanism is triggered, including retrospectively recommending previous knowledge points for reinforcement, recommending alternative explanatory resources, inserting targeted exercises generated by AIGC, or re-calling the personalized learning path generation engine to plan subsequent paths; Teacher-side adjustments: Continuously collect student learning feedback data and teaching effectiveness evaluation results, combined with teacher's teaching behavior records; when common weaknesses in the class are not improved, teaching resources are poorly matched, or student learning participation is insufficient, a dynamic adjustment mechanism is triggered on the teacher side, including recommending optimized teaching resources, adjusting the allocation of teaching priorities, providing suggestions for differentiated teaching methods, or regenerating personalized teaching paths.

[0019] It should be noted that the core principle of this embodiment is as follows: Knowledge graphs serve as a shared knowledge brain: providing structured domain knowledge models, offering a unified semantic map for generating student learning paths and teacher teaching paths, ensuring the coherence of the knowledge system, and supporting the precise anchoring of teacher and student resources; The dual-dimensional profiles of teachers and students serve as a two-way sensor: they construct three-dimensional and dynamic profiles of students and teachers respectively, enabling accurate perception of students' learning status and teachers' teaching status, and providing two-way data support for differentiated and personalized decision-making. The differentiated intelligent engine serves as a two-way decision-making center: the path generation and content recommendation engines are designed with exclusive algorithm logic to meet the different needs of teachers and students, achieving a dual qualitative change of personalized learning paths for students and personalized teaching paths for teachers. Two-way closed-loop feedback as a co-regulator: the dual closed-loop mechanism of student learning-assessment-adjustment and teacher teaching-assessment-adjustment, combined with the linkage feedback of class learning situation, endows the system with the ability of teachers and students to co-evolve and realize the dynamic adaptation of teaching and learning. AIGC acts as a two-way resource incubator: it generates exclusive learning resources for students and exclusive teaching resources for teachers on demand, breaking through the capacity limitations of traditional resource libraries and simultaneously meeting the long-tail needs and personalized scenario requirements of teachers and students.

[0020] This embodiment addresses the core shortcomings of existing technologies, such as low personalization, fragmented knowledge, weak adaptability, and lack of differentiation between teacher and student needs, achieving personalized support for both students and teachers throughout the entire process; specifically: For students: Through three-dimensional assessment, A* algorithm path, and hybrid recommendation, students can obtain learning paths and resources that match their theoretical / practical / soft skills levels, directly targeting weak areas and improving learning efficiency by more than 30%.

[0021] For teachers: By combining teacher profiles, class learning information, and customized teaching paths, common problems in the class can be accurately identified, and resources and methodological suggestions that match the teaching style can be obtained to optimize teaching effectiveness and reduce lesson preparation costs.

[0022] Overall effect: AIGC generates exclusive resources for teachers and students on demand, solving cold start and long-tail needs; the closed-loop dynamic adjustment mechanism ensures real-time adaptation of teaching and learning, forming a dual closed loop of learning-assessment-adjustment and teaching-assessment-adjustment.

[0023] In a specific example, the student multidimensional profile in step S3 includes multiple dimensions such as theoretical knowledge mastery, practical ability, communication and collaboration ability, project management ability, and learning style preference. The teacher multidimensional profile includes multiple dimensions such as teaching content suitability, teaching method effectiveness, teaching style preference, and subject expertise. Both types of profiles are updated in real time with the corresponding behavioral data and evaluation results.

[0024] This embodiment clearly defines the core dimensions of teacher and student profiles to avoid insufficient personalization accuracy caused by blurry profiles. Specifically: For students: Ensure real-time updates of data across dimensions such as theory, practice, and soft skills; improve the accuracy of path and resource recommendations by 25%; and avoid a one-size-fits-all approach.

[0025] For teachers: Focusing on core dimensions such as the suitability of teaching content and teaching style, the generated teaching paths are more in line with teachers' abilities and preferences, and the suitability rate of teaching methods has increased by 30%.

[0026] In a specific example, the A* search algorithm in step S4 optimizes node search efficiency through open and closed lists. The weight parameters in the student path cost function can be dynamically adjusted according to the characteristics of the subject, the difficulty of the course, and the student's learning objectives. The teacher path generation can adjust the weight of teaching focus according to the class size and the differences in students' basic knowledge.

[0027] This embodiment optimizes the node search efficiency of the A* algorithm, reduces the time complexity of path generation, and improves the system response speed; specifically: For students: Weight parameters can be dynamically adjusted (e.g., increasing the weight of practical knowledge points in software engineering), path planning is more aligned with the characteristics of the discipline and learning objectives, and the learning cycle is shortened by 20%-30%.

[0028] Teacher side: The weighting of teaching focus can be adapted to class size and differences in students' basic knowledge (e.g., increasing the weight of common knowledge points in large classes), and the utilization rate of teaching resources is increased by 35%.

[0029] In a specific example, the exploration probability ε in the ε-greedy strategy in step S5 ranges from 0.05 to 0.2. The student end can dynamically adjust the probability based on the resource library update frequency and the completeness of the student profile, while the teacher end can dynamically adjust the probability based on the teaching resource update frequency, the completeness of the teacher profile, and changes in the class's learning situation.

[0030] This embodiment clarifies the range of values ​​for the exploration probability ε and the logic for its dynamic adjustment, balancing the accuracy and exploratory nature of the recommendation, specifically as follows: For students: Increase ε (e.g., 0.2) when new users or resource library updates, increasing the probability of discovering potential high-quality resources; decrease ε (e.g., 0.05) after the profile is improved, keeping the recommendation accuracy above 85%.

[0031] Teacher side: Adjust ε when the class's learning situation changes to ensure that the recommended teaching resources not only match the teacher's style, but also introduce innovative teaching tools, increasing the diversity of teaching methods by 25%.

[0032] Example 2 To achieve the above objectives, see Figure 2 This embodiment also provides an intelligent adaptive personalized learning system, including: Knowledge graph library: Stores structured domain knowledge entities and their relationships, including the importance, difficulty coefficient or association strength weight information of knowledge entities, providing structured knowledge support for intelligent decision-making in the system; Adaptive Course Resource Library: Stores learning resources such as videos, documents, code examples, interactive experiments, and assessment questions, as well as teaching resources such as syllabus templates and learning analysis tools. Each resource is associated with a corresponding knowledge graph entity and has resource metadata annotation function. It works in conjunction with the AIGC content generation and enhancement module to dynamically supplement resources and classify and manage learning resources and teaching resources. The teacher and student profile module integrates an online assessment system, an integrated development environment (IDE), a project collaboration platform, a behavior data collection SDK, and teaching evaluation tools. It continuously collects students' theoretical assessment data, practical operation data, soft skills assessment data, and learning behavior data, as well as teachers' teaching content data, teaching effectiveness assessment data, peer review data, and teaching behavior data. It calculates and maintains multi-dimensional dynamic profiles of students and teachers. AIGC Content Generation and Enhancement Module: Based on the resource gaps in the adaptive course resource library, students' weaknesses, teachers' teaching needs and personalized profiles, it dynamically generates learning content such as text, code examples, diagrams, and simulated dialogues, as well as teaching content such as syllabus templates and learning reports, providing support for adaptive content recommendation. Personalized Path Generation Engine: Includes a student learning path generation sub-engine and a teacher teaching path generation sub-engine; the student sub-engine uses the knowledge graph-based A* search algorithm, receives the student's current state vector, the target job / skill set, and knowledge graph database data, and generates the optimal learning path; the teacher sub-engine combines the knowledge graph, teacher profile, and class learning data to generate personalized teaching paths; The adaptive content recommendation engine includes a student content recommendation sub-engine and a teacher content recommendation sub-engine. The student sub-engine uses a hybrid recommendation algorithm to receive students' personalized learning paths, student profiles, and learning resource data, and recommends or generates suitable learning resources. The teacher sub-engine recommends or generates suitable teaching resources based on teachers' teaching paths, teacher profiles, class learning conditions, and teaching resource data. The adaptive assessment and dynamic adjustment module includes a student-side adjustment submodule and a teacher-side adjustment submodule. The student-side submodule collects students' learning behavior data and micro-assessment results in real time, compares them with preset thresholds, and triggers the student-side adjustment mechanism. The teacher-side submodule collects class learning data, teaching effectiveness evaluation results, and teacher teaching behavior data in real time, compares them with preset thresholds, and triggers the teacher-side adjustment mechanism. Both types of submodules update corresponding profiles, paths, and recommended resources to form a closed-loop feedback. User interface: Includes student and teacher interfaces; the student interface provides functions such as personal profile display, learning path viewing, resource learning, assessment submission, and feedback interaction; the teacher interface provides functions such as personal profile display, teaching path viewing, teaching resource access, student learning analysis viewing, teaching adjustment operations, and feedback interaction.

[0033] This embodiment provides hardware and software support for the implementation of the method, enabling the collaborative work of the knowledge graph base, teacher and student profiles, dual engines, and dual adjustment modules to ensure that the method can be implemented at scale.

[0034] Resource layer: Categorized storage of learning / teaching resources, bidirectional indexing to improve resource retrieval efficiency, AIGC module to dynamically supplement resources, and resource coverage to over 95%.

[0035] Functional layer: The dual path generation engine and dual recommendation engine are independent yet interconnected, which not only meet the differentiated needs of teachers and students, but also achieve collaboration through class learning information, improving the teaching and learning fit by 40%.

[0036] Interaction Layer: A dedicated interface for teachers and students lowers the barrier to entry, while features such as profile display and student progress monitoring allow teachers and students to keep track of the student's progress in real time, enhancing the user experience. In a specific example, the knowledge graph database implements an update mechanism that combines regular automatic updates with manual optimization through natural language processing tools and an expert review system.

[0037] This embodiment achieves a dual guarantee of automatic updating and manual optimization, solving the problem of lagging updates in existing knowledge graphs; specifically: Knowledge timeliness: Regularly and automatically capture industry trends and textbook updates, and increase the knowledge graph update frequency to once per quarter to ensure that knowledge points are not outdated.

[0038] Knowledge accuracy: The expert review system filters out erroneous information, increasing the accuracy of knowledge entities and relationships to over 98%, thus avoiding teaching and learning deviations caused by knowledge errors.

[0039] In a specific example, the teacher and student profile module is linked with the adaptive assessment and dynamic adjustment module to realize real-time updates of student and teacher profiles, providing dynamic data support for the generation of student learning paths, teacher teaching paths and corresponding content recommendations.

[0040] This embodiment enhances the dynamic updating feature of profiles, solving the problems of static profiles and lagging recommendations in existing systems; specifically: On the student side: learning behaviors or assessment results are fed back to the profile in real time, and the response time for path and resource adjustments is ≤5 seconds, ensuring that learning support is aligned with the real-time status.

[0041] Teacher side: Real-time updates to teacher profiles based on changes in class learning progress enable more timely adjustments to teaching approaches and improve the efficiency of resolving common class problems by 30%.

[0042] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent adaptive personalized learning method based on AIGC and knowledge graph, characterized in that, Comprise the following steps: S1, construct and update the field knowledge graph library: extract core concepts, skill points as knowledge entities from curriculum standards, authoritative textbooks, industry standards, define the pre-requisite, dependency, association, and containment relationship between entities, assign weights to entities representing importance, difficulty coefficient, or correlation strength, and construct a structured weighted field knowledge graph; Use natural language processing tools and expert review system, combined with curriculum outline, textbook, academic papers, open source code library, project case documents to update the knowledge graph; S2, establish and enrich the adaptive course resource library: accurately anchor various learning resources such as videos, documents, code examples, interactive experiments, and evaluation questions to knowledge entities in the knowledge graph, assign resource ID to each resource, label resource difficulty level, estimated learning time, and associated knowledge entities, and establish bidirectional index between all resources and knowledge entities; Integrate AIGC tools, when the resource library lacks resources of a certain difficulty, form, or style for a certain knowledge point, automatically generate personalized resources in the form of text, code examples, illustrations, or simulated dialogues; At the same time, classify and store student-oriented learning resources and teacher-oriented teaching resources, teaching resources include teaching outline templates, knowledge point explanation courseware, learning situation analysis tools, homework design cases, and teaching evaluation scales; S3, generate and update student and teacher multi-dimensional portraits respectively: Student multi-dimensional portrait: Through online testing, exercises combined with item response theory or knowledge tracking model, calculate the theoretical mastery of students on each knowledge point; Through the integration of online IDE, automated testing framework and code static analysis tools, evaluate the code or project submitted by the student to obtain the practical ability score; Through AIGC driven simulated dialogue scene, virtual team project, peer evaluation, evaluate students' communication, collaboration and project management soft skills; Gather the above evaluation data and learning behavior data to form a dynamic updated student multi-dimensional digital portrait containing student ID, theoretical knowledge mastery vector, practical ability score matrix, soft skill index, learning style preference vector, and learning history sequence; Teacher multi-dimensional portrait: Evaluate the teaching content adaptation degree through the teaching outline, courseware, and homework design scheme submitted by the teacher; Evaluate the effectiveness of teaching methods through student learning effect feedback, teaching evaluation data, and peer evaluation; Analyze the teaching style preference through the behavior data of the teacher using the system; Gather the above data to form a dynamic updated teacher multi-dimensional digital portrait containing teacher ID, teaching content adaptation degree index, teaching method effectiveness score, teaching style preference vector, teaching history record, and subject expertise label; S4, generate differentiated personalized paths: Student personalized learning path: Based on A search algorithm, build a learning path generation engine, model the learning path planning as a path search problem with minimum cognitive cost from the student's current state to the target state on the field knowledge graph; The nodes to be checked and the checked nodes are managed by open list and closed list respectively, the path cost is calculated by using cost function f(n)=g(n)+h(n), wherein g(n) is actual cost from starting point to node n, and the calculation formula is g(n)=∑(inherent difficulty of knowledge point_i(1-student's mastery of knowledge point_i)), and h(n) is estimated cost from node n to target point, that is, average difficulty of knowledge points involved in the shortest path from node n to target; input current state vector of student, target position / skill set, domain knowledge graph and algorithm weight parameter, and output optimal learning sequence; Teacher personalized teaching path: based on domain knowledge graph and teacher multi-dimensional portrait, combined with the overall learning profile of students in the class taught, a personalized teaching path is generated, which includes teaching focus allocation, teaching resource recommendation, teaching activity design, and learning profile monitoring nodes; input teacher portrait, class learning data, course objectives, and knowledge graph, and output teaching progress planning, phased teaching focus, and adapted teaching method suggestions; S5、Adaptive content recommendation: Student content recommendation: based on content filtering, context awareness and ε-greedy strategy, an adaptive content recommendation engine of hybrid recommendation algorithm is constructed; for each knowledge point in the learning path, resource content features, student portrait features and context information features are extracted, and comprehensive matching score S is calculated: S=W1*S_content+W2*S_context+W3*S_popularity, wherein S_content is the cosine similarity of resource features and student portrait, S_context is the context matching degree considering device type, available time and current learning state, S_popularity is the global average score and completion rate of the resource, and W1, W2 and W3 are weight coefficients; the ε-greedy strategy is used for decision-making, that is, with a high probability (1-ε), the resource with the highest comprehensive matching score is selected, and with a small probability (ε), other resources are randomly selected to discover potential high-quality resources and collect data; Teacher content recommendation: based on teaching path, teacher portrait and class learning profile, suitable teaching resources, learning profile analysis tools and teaching adjustment suggestions are recommended; when there is a lack of resources for a specific teaching scenario, AIGC tools automatically generate teaching outline templates, classroom interaction scripts or personalized learning profile reports; S6、Adaptive evaluation and dynamic adjustment: Student side adjustment: continuously monitor real-time behavior data stream and micro-evaluation results of students, set mastery threshold, error rate threshold and page dwell time threshold; when abnormal conditions such as student knowledge mastery not reaching the expected value, sudden increase of error rate or too long page dwell time are detected, trigger the dynamic adjustment mechanism, including backtracking recommended prerequisite knowledge points for consolidation, recommending alternative explanation resources, inserting AIGC generated targeted exercises, or re-invoking the personalized learning path generation engine to plan the subsequent path; Teacher-side adjustment: continuously collect learning feedback data of students in the class, teaching effectiveness evaluation results, and combine with the teacher's teaching behavior records; when it is detected that the common weakness of the class has not improved, the teaching resource adaptation degree is low, or the student learning participation is insufficient, trigger the teacher-side dynamic adjustment mechanism, including recommending optimized teaching resources, adjusting teaching focus allocation, providing differentiated teaching method suggestions, or regenerating personalized teaching paths. 2.The AIGC and knowledge graph based intelligent adaptive personalized learning method according to claim 1, characterized in that, The student multi-dimensional portrait in step S3 includes theoretical knowledge mastery, practical ability, communication and cooperation ability, project management ability, and learning style preference, and the teacher multi-dimensional portrait includes teaching content adaptation, teaching method effectiveness, teaching style preference, and subject expertise, and both types of portraits are updated in real time according to corresponding behavior data and evaluation results. 3.The AIGC and knowledge graph based intelligent adaptive personalized learning method according to claim 1, characterized in that, The A* search algorithm in step S4 optimizes node search efficiency through open list and closed list, and the weight parameters in the student path cost function can be dynamically adjusted according to subject characteristics, course difficulty, and student learning goals, and the teacher path generation can adjust the teaching focus allocation weight according to the class size and student foundation difference. 4.The AIGC and knowledge graph based intelligent adaptive personalized learning method according to claim 1, characterized in that, The exploration probability ε in the ε-greedy strategy in step S5 ranges from 0.05 to 0.2, which can be dynamically adjusted by the student side according to the resource library update frequency and student portrait perfection, and by the teacher side according to the teaching resource update frequency, teacher portrait perfection, and class learning situation changes.

5. An intelligent adaptive personalized learning system implementing the method of any of claims 1-4, characterized by, It includes: Knowledge graph library: stores structured domain knowledge entities and their relationships, including importance, difficulty coefficient, or correlation strength weight information of knowledge entities, providing structured knowledge support for system intelligent decision-making; Adaptive course resource library: stores learning resources such as videos, documents, code examples, interactive experiments, and evaluation questions, as well as teaching outline templates and learning situation analysis tools, each resource is associated with corresponding knowledge graph entities, has resource metadata labeling function, and dynamically supplements resources through AIGC content generation and enhancement module, and classifies and manages learning resources and teaching resources; Teacher and student portrait module: integrates online evaluation system, integrated development environment, project collaboration platform, behavior data collection SDK, and teaching evaluation tool, respectively continuously collects students' theoretical evaluation data, practical operation data, soft skill evaluation data, and learning behavior data, and teachers' teaching content data, teaching effectiveness evaluation data, peer evaluation data, and teaching behavior data, and calculates and maintains student and teacher multi-dimensional dynamic portraits; AIGC content generation and enhancement module: dynamically generates learning content such as text, code examples, illustrations, and simulated dialogues, and teaching content such as teaching outline templates and learning situation reports, based on resource gaps in the adaptive course resource library, student weaknesses, teacher teaching needs, and personalized portraits, to support adaptive content recommendation; Personalized path generation engine: includes student learning path generation sub-engine and teacher teaching path generation sub-engine; The student sub-engine adopts a knowledge graph-based A* search algorithm, receives a current state vector of a student, a target position / skill set, and knowledge graph library data, and generates an optimal learning path; the teacher sub-engine combines a knowledge graph, a teacher portrait, and class learning data to generate an individualized teaching path; The adaptive content recommendation engine includes a student content recommendation sub-engine and a teacher content recommendation sub-engine; The student sub-engine adopts a hybrid recommendation algorithm, receives a student individualized learning path, a student portrait, and learning resource data, and recommends or generates adaptive learning resources; The teacher sub-engine, based on a teacher teaching path, a teacher portrait, class learning data, and teaching resource data, recommends or generates adaptive teaching resources; The adaptive assessment and dynamic adjustment module includes a student-side adjustment sub-module and a teacher-side adjustment sub-module; the student-side sub-module collects learning behavior data and micro-assessment results of a student in real time, compares the data with preset threshold values, triggers a student-side adjustment mechanism, and updates a corresponding portrait, path, and recommended resources, forming a closed-loop feedback; the teacher-side sub-module collects class learning data, teaching effectiveness assessment results, and teacher teaching behavior data in real time, compares the data with preset threshold values, triggers a teacher-side adjustment mechanism, and updates a corresponding portrait, path, and recommended resources, forming a closed-loop feedback; The user interaction interface includes a student interaction interface and a teacher interaction interface; the student interface provides functions of personal portrait display, learning path viewing, resource learning, assessment submission, and feedback interaction; the teacher interface provides functions of personal portrait display, teaching path viewing, teaching resource calling, learning analysis viewing, teaching adjustment operation, and feedback interaction. 6.The AIGC and knowledge graph based intelligent adaptive personalized learning system according to claim 5, characterized in that, The knowledge graph library realizes an updating mechanism combining regular automatic updating and manual optimization through a natural language processing tool and an expert review system. 7.The AIGC and knowledge graph based intelligent adaptive personalized learning system according to claim 5, characterized in that, The student and teacher portrait modules and the adaptive assessment and dynamic adjustment module are linked to realize real-time updating of student and teacher portraits, and to provide dynamic data support for generation of student learning paths and teacher teaching paths and corresponding content recommendation.

Images