Self-adaptive learning path recommendation method and system for artificial intelligence general recognition education

Abstract

The invention relates to the technical field of education artificial intelligence, and provides a self-adaptive learning path recommendation method for artificial intelligence general recognition education, which comprises the following steps of: 1, calling a large language model to analyze unstructured course resources of AI general recognition education, mining semantic association and implicit dependency of knowledge points, and constructing a learning path; generating a space-time knowledge graph containing confidence scores, automatically capturing the latest literature and tool update of the field every 72 hours, and iteratively optimizing the topological structure of the graph; 2, deploying data acquisition points to acquire multi-source learning data of a user in real time, and calculating a dynamic cognitive state vector with knowledge points as dimensions by mastering an entropy model; the knowledge graph is automatically iterated through LLM, and the problem of update lag is solved; precise personalized recommendation is realized according to multi-dimensional data and a dynamic algorithm, and cognitive differences are adapted; visually presenting decision logic, and cracking a decision black box; the method can be migrated to multiple fields, supports multi-terminal and offline learning, exceeds an expected adaptive scene, and improves the learning efficiency and credibility.

Description

Technical Field

[0001] This invention relates to the field of educational artificial intelligence technology, specifically to an adaptive learning path recommendation method and system for general education in artificial intelligence. Background Technology

[0002] In the field of general education in artificial intelligence, the subject knowledge is updated rapidly, the concepts are abstract, and the practical aspects are highly relevant. Learners need to accurately identify their weaknesses within a dynamic knowledge system and obtain personalized learning guidance in order to efficiently master the core content. This is especially true for learners from non-technical backgrounds, as traditional standardized learning paths are difficult to adapt to their cognitive differences, necessitating a recommendation system that can respond to individual needs and dynamically adjust accordingly.

[0003] Existing AI-powered general education learning recommendation technologies mainly fall into two categories: one is recommendation systems based on static knowledge graphs, which rely on domain experts to manually construct knowledge point association structures and push fixed paths by matching the user's initial profile; the other is recommendation systems based on simple behavioral analysis, which only recommend similar resources based on the user's historical learning records without deeply integrating the logic of the knowledge system. Both types of technologies can achieve basic recommendation functions and are already being used in some online education platforms.

[0004] Existing technologies have significant drawbacks: static knowledge graphs rely on manual updates, making it difficult to keep up with the pace of knowledge iteration in the AI ​​field, resulting in lagging recommended content; recommendations based on group profiles or simple behaviors cannot accurately depict the real-time cognitive state of individuals, easily leading to homogenization issues such as "one path for everyone"; although some systems can dynamically adjust paths, the adjustment logic is not transparent, and users cannot know the reasons for path changes, reducing their trust and enthusiasm for learning; in addition, existing technologies have not designed specific recommendation strategies for the "theory-practice integration" characteristics of AI general education, making it difficult to balance conceptual understanding and practical skills development, and failing to meet the specific learning needs of each subject. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an adaptive learning path recommendation method and system for AI-based general education, solving several defects in existing AI general education recommendation technologies: knowledge graphs rely on manual updates, making it difficult to keep up with subject iterations and leading to lag; recommendations are based on group profiles or simple behaviors, failing to accurately match individual real-time cognition and resulting in severe homogenization; the logic of path adjustment in some systems is opaque, reducing learning trust; and it fails to adapt to the "theory and practice integration" characteristics of subjects, making it difficult to balance the needs of conceptual and practical training.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an adaptive learning path recommendation method for general education in artificial intelligence, comprising the following steps: Step 1: Call a large language model to analyze unstructured course resources for AI general education, mine semantic relationships and implicit dependencies of knowledge points, generate a spatiotemporal knowledge graph with confidence scores, and automatically capture the latest literature and tool updates in the field every 72 hours to iteratively optimize the graph topology. Step 2: Deploy data collection points to collect multi-source learning data from users in real time, and calculate dynamic cognitive state vectors with knowledge points as the dimension by mastering the entropy model; Step 3: Based on the dynamic cognitive state vector and spatiotemporal knowledge graph, the learning path is planned using the dynamically weighted A* algorithm. The algorithm cost function is: ; in Cost of the learned path To estimate the remaining cost, weighting factors Dynamic calibration based on the knowledge mastery level in the cognitive state vector; Step 4: In the WebGL dynamic knowledge graph interface, highlight the recommended path, adjust the nodes and decision basis, and adjust the causal logic by marking the path in the pop-up window; Step 5: Collect users' test scores and subjective evaluations after completing the recommended path, input them into LLM for semantic analysis, and correct the parameters of the knowledge point association weights and cognitive state vector calculation model in the knowledge graph.

[0007] Preferably, in step one, the large-scale language model is a GPT-4 level or higher model, the knowledge graph is stored in the Neo4j graph database, each knowledge node includes the attributes of "concept definition, prerequisite dependencies, subsequent applications, and typical cases", and the confidence score is determined by combining LLM semantic similarity calculation with domain expert sampling verification.

[0008] Preferably, in step two, the entropy model is used to calculate cognitive uncertainty using the following formula: ; in Let be the probability of mastering the i-th knowledge point. The higher the entropy value, the more ambiguous the cognitive state, and the more necessary it is to strengthen the learning of this knowledge point.

[0009] Preferably, the path rerouting triggering conditions of the dynamically weighted A* algorithm in step three include: the mastery of any knowledge point in the cognitive state vector is lower than a threshold, the user triggers the same knowledge point-related question twice in a row, and the knowledge graph update involves core prerequisite knowledge points.

[0010] Preferably, the visualization interface in step four supports user interaction: hovering over a path node displays the mastery curve of the knowledge point at that node, clicking to adjust the record allows viewing historical decision logs, and dragging nodes allows customizing the learning order.

[0011] Preferably, step six is ​​also included: for each knowledge point in the recommended path, match 3-5 different learning resources, and prioritize pushing suitable resources based on the user's historical resource preferences.

[0012] Preferably, the resource preference analysis in step six is ​​implemented through a collaborative filtering algorithm, which calculates the similarity of resource selection between the user and users with similar cognitive states, and optimizes the recommendation priority by combining the characteristics of knowledge points.

[0013] An adaptive learning path recommendation system for general education in artificial intelligence includes: The knowledge graph management module includes an LLM interface unit, a resource parsing unit, a graph iteration unit, and a graph database storage unit. The LLM interface unit calls the model to parse resources, the resource parsing unit extracts knowledge triples, the graph iteration unit updates the graph periodically, and the storage unit implements the persistence of nodes and relationships. State perception and modeling module: includes a data acquisition unit, a data preprocessing unit, and a cognitive state calculation unit, outputting a dynamic cognitive state vector; The path decision engine module has a built-in dynamic weighted A* algorithm unit and a path evaluation unit. The algorithm unit receives cognitive vectors and graph data to calculate the optimal path, and the evaluation unit verifies the feasibility of the path. Visualization and Interaction Module: Includes a WebGL rendering unit, an interactive response unit, and an explanation information generation unit. The rendering unit draws dynamic graphs, the interactive unit receives user operations, and the explanation unit generates decision-making basis pop-ups. Feedback Iteration Module: Includes a data acquisition unit, a semantic analysis unit, and a model optimization unit.

[0014] Preferably, the knowledge graph management module and the path decision engine module communicate in real time via RESTful API. When the graph is updated, the decision engine automatically triggers path recalculation to ensure that the recommended path is synchronized with the latest knowledge system.

[0015] Preferably, the visualization interaction module supports multi-terminal adaptation, the mobile interface adopts a lightweight graph rendering scheme, retains the core path and explanation information, and supports offline caching of generated recommended paths, and synchronizes learning data and updated content after connecting to the network.

[0016] This invention provides an adaptive learning path recommendation method and system for general education in artificial intelligence. It has the following beneficial effects: 1. This invention uses LLM to automatically construct and periodically iterate knowledge graphs, replacing traditional manual updates. This not only significantly improves the efficiency of knowledge system updates, but also captures the latest tools and theories in the field, ensuring that the recommended path always adapts to the dynamic knowledge structure of AI general education.

[0017] 2. Based on multi-dimensional learning data and dynamic weighted A* algorithm, this invention can perceive changes in the user's cognitive state in real time, prioritize the supplementation of weak knowledge points, and avoid homogeneous recommendations. It can specifically lower the learning threshold for learners without technical backgrounds.

[0018] 3. This invention presents the causal logic of path adjustment intuitively through a visual interface, allowing users to clearly understand the basis for recommendations, eliminating doubts about AI recommendations, and enhancing the sense of learning goals. In practice, user path completion rate and learning enthusiasm have been significantly improved.

[0019] 4. This invention is not only applicable to AI general education, but can also be transferred to other rapidly iterating general education fields such as big data and blockchain by adjusting the knowledge graph construction rules and cognitive state calculation models; it also supports multi-terminal adaptation and offline learning, breaking through the scenario limitations of traditional online recommendation systems, providing flexible solutions for fragmented learning, and demonstrating an application scope beyond expectations. Attached Figure Description

[0020] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example: As one aspect of the present invention, please refer to the appendix. Figure 1 This invention provides an adaptive learning path recommendation method for general education in artificial intelligence, comprising the following steps: Step 1: Use a GPT-4 or higher language model to analyze unstructured course resources for AI general education, mine semantic relationships and implicit dependencies of knowledge points, and generate a spatiotemporal knowledge graph with confidence scores. The knowledge graph is stored in the Neo4j graph database. Each knowledge node includes the attributes of "concept definition, prerequisite dependencies, subsequent applications, and typical cases". The confidence score is determined by combining LLM semantic similarity calculation and domain expert sampling verification. The graph topology is automatically updated every 72 hours by crawling the latest literature and tools in the field. Step Two: Deploy data collection points to collect multi-source learning data from users in real time. Utilize the entropy model to calculate a dynamic cognitive state vector with knowledge points as the dimension. The entropy model calculates cognitive uncertainty using the following formula: ; in Let be the probability of mastering the i-th knowledge point. The higher the entropy value, the more ambiguous the cognitive state, and the more necessary it is to strengthen the learning of this knowledge point. Step 3: Based on the dynamic cognitive state vector and spatiotemporal knowledge graph, the learning path is planned using the dynamically weighted A* algorithm. The algorithm cost function is: ; in Cost of the learned path To estimate the remaining cost, weighting factors The path rerouting trigger conditions of the dynamically weighted A* algorithm are dynamically calibrated by the mastery of knowledge points in the cognitive state vector. These conditions include: the mastery of any knowledge point in the cognitive state vector is lower than the threshold, the user triggers the same knowledge point-related question twice in a row, and the knowledge graph update involves core prerequisite knowledge points. Step 4: In the WebGL dynamic knowledge graph interface, the recommended path, adjustment nodes and decision basis are highlighted. The causal logic can be adjusted by marking the path in the pop-up window. The visual interface supports user interaction: hovering over a path node displays the mastery curve of the knowledge point of that node, clicking on the adjustment record can view the historical decision log, and dragging nodes can customize the learning order. Step 5: Collect users' test scores and subjective evaluations after completing the recommended path, input them into LLM for semantic analysis, and correct the parameters of the knowledge point association weights and cognitive state vector calculation model in the knowledge graph; Step Six: For each knowledge point in the recommended path, match 3-5 different learning resources, prioritize the push of suitable resources based on the user's historical resource preferences, and implement resource preference analysis through collaborative filtering algorithm to calculate the similarity of resource selection between the user and users with similar cognitive states, and optimize the recommendation priority in combination with the characteristics of knowledge points.

[0023] As another aspect of the present invention, embodiments of the present invention provide an adaptive learning path recommendation system for general education in artificial intelligence, comprising: The knowledge graph management module includes an LLM interface unit, a resource parsing unit, a graph iteration unit, and a graph database storage unit. The LLM interface unit calls the model to parse resources, the resource parsing unit extracts knowledge triples, the graph iteration unit updates the graph periodically, and the storage unit implements the persistence of nodes and relationships. State perception and modeling module: includes a data acquisition unit, a data preprocessing unit, and a cognitive state calculation unit, outputting a dynamic cognitive state vector; The path decision engine module has a built-in dynamic weighted A* algorithm unit and a path evaluation unit. The algorithm unit receives cognitive vectors and graph data to calculate the optimal path, and the evaluation unit verifies the feasibility of the path. Visualization and Interaction Module: Includes a WebGL rendering unit, an interactive response unit, and an explanation information generation unit. The rendering unit draws dynamic graphs, the interactive unit receives user operations, and the explanation unit generates decision-making basis pop-ups. Feedback Iteration Module: Includes a data acquisition unit, a semantic analysis unit, and a model optimization unit.

[0024] The knowledge graph management module and the path decision engine module communicate in real time via RESTful API. When the graph is updated, the decision engine automatically triggers path recalculation to ensure that the recommended paths are synchronized with the latest knowledge system. The visualization interaction module supports multi-terminal adaptation. The mobile interface adopts a lightweight graph rendering scheme, retains core paths and explanation information, and supports offline caching of generated recommended paths. After connecting to the network, it synchronizes learning data and updated content.

[0025] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An adaptive learning path recommendation method for general education in artificial intelligence, characterized in that, Includes the following steps: Step 1: Call a large language model to analyze unstructured course resources for AI general education, mine semantic relationships and implicit dependencies of knowledge points, generate a spatiotemporal knowledge graph with confidence scores, and automatically capture the latest literature and tool updates in the field every 72 hours to iteratively optimize the graph topology. Step 2: Deploy data collection points to collect multi-source learning data from users in real time, and calculate dynamic cognitive state vectors with knowledge points as the dimension by mastering the entropy model; Step 3: Based on the dynamic cognitive state vector and spatiotemporal knowledge graph, the learning path is planned using the dynamically weighted A* algorithm. The algorithm cost function is: ； in Cost of the learned path To estimate the remaining cost, weighting factors Dynamic calibration based on the knowledge mastery level in the cognitive state vector; Step 4: In the WebGL dynamic knowledge graph interface, highlight the recommended path, adjust the nodes and decision basis, and adjust the causal logic by marking the path in the pop-up window; Step 5: Collect users' test scores and subjective evaluations after completing the recommended path, input them into LLM for semantic analysis, and correct the parameters of the knowledge point association weights and cognitive state vector calculation model in the knowledge graph.

2. The adaptive learning path recommendation method for general education in artificial intelligence according to claim 1, characterized in that, In step one, the large-scale language model selected is a GPT-4 level or higher model, and the knowledge graph is stored in the Neo4j graph database. Each knowledge node contains the attributes of "concept definition, prerequisite dependencies, subsequent applications, and typical cases". The confidence score is determined by combining LLM semantic similarity calculation with domain expert sampling verification.

3. The adaptive learning path recommendation method for general education in artificial intelligence according to claim 1, characterized in that, In step two, the entropy model is used to calculate cognitive uncertainty using the following formula: ; in Let be the probability of mastering the i-th knowledge point. The higher the entropy value, the more ambiguous the cognitive state, and the more necessary it is to strengthen the learning of this knowledge point.

4. The adaptive learning path recommendation method for general education in artificial intelligence according to claim 1, characterized in that, The path rerouting trigger conditions for the dynamic weighted A* algorithm in step three include: the mastery of any knowledge point in the cognitive state vector is lower than the threshold, the user triggers the same knowledge point-related question twice in a row, and the knowledge graph update involves core prerequisite knowledge points.

5. The adaptive learning path recommendation method for general education in artificial intelligence according to claim 1, characterized in that, In step four, the visualization interface supports user interaction: hovering over a path node displays the mastery curve of the knowledge point at that node, clicking to adjust the record allows viewing historical decision logs, and dragging nodes allows customizing the learning order.

6. The adaptive learning path recommendation method for general education in artificial intelligence according to claim 1, characterized in that, It also includes step six: for each knowledge point in the recommended path, match 3-5 different learning resources, and prioritize pushing suitable resources based on the user's historical resource preferences.

7. The adaptive learning path recommendation method for general education in artificial intelligence according to claim 6, characterized in that, In step six, the resource preference analysis is implemented through a collaborative filtering algorithm. This algorithm calculates the similarity of resource selection between a user and other users with similar cognitive states, and optimizes the recommendation priority by combining the characteristics of knowledge points.

8. An adaptive learning path recommendation system for general education in artificial intelligence, using the adaptive learning path recommendation method for general education in artificial intelligence as described in any one of claims 1-7, characterized in that, include: The knowledge graph management module includes an LLM interface unit, a resource parsing unit, a graph iteration unit, and a graph database storage unit. The LLM interface unit calls the model to parse resources, the resource parsing unit extracts knowledge triples, the graph iteration unit updates the graph periodically, and the storage unit implements the persistence of nodes and relationships. State perception and modeling module: includes a data acquisition unit, a data preprocessing unit, and a cognitive state calculation unit, outputting a dynamic cognitive state vector; The path decision engine module has a built-in dynamic weighted A* algorithm unit and a path evaluation unit. The algorithm unit receives cognitive vectors and graph data to calculate the optimal path, and the evaluation unit verifies the feasibility of the path. Visualization and Interaction Module: Includes a WebGL rendering unit, an interactive response unit, and an explanation information generation unit. The rendering unit draws dynamic graphs, the interactive unit receives user operations, and the explanation unit generates decision-making basis pop-ups. Feedback Iteration Module: Includes a data acquisition unit, a semantic analysis unit, and a model optimization unit.

9. The adaptive learning path recommendation system for general education in artificial intelligence according to claim 8, characterized in that, The knowledge graph management module and the path decision engine module communicate in real time via RESTful API. When the graph is updated, the decision engine automatically triggers path recalculation to ensure that the recommended path is synchronized with the latest knowledge system.

10. The adaptive learning path recommendation system for general education in artificial intelligence according to claim 8, characterized in that, The visualization interaction module supports multi-terminal adaptation. The mobile interface adopts a lightweight graph rendering scheme, retains the core paths and explanation information, and supports offline caching of generated recommended paths. After connecting to the network, it synchronizes learning data and updated content.

Images