Method and system for evaluating value of knowledge atoms in educational knowledge graph

Abstract

This invention relates to the interdisciplinary field of education and artificial intelligence, and particularly to a method and system for evaluating the value of knowledge atoms in an educational knowledge graph. The method involves constructing an educational knowledge graph; acquiring and preprocessing multi-dimensional indicator data for each knowledge point, including structural attribute dimensions, content attribute dimensions, teaching adaptability dimensions, and ability contribution dimensions; determining the weights of the primary and secondary dimensions of the multi-dimensional indicators based on a pre-defined educational scenario using a combined weighting method; calculating the comprehensive value score of the knowledge point through weighted summation based on the preprocessed multi-dimensional indicator data and the determined primary and secondary indicator weights; updating the knowledge point node weights in real time based on a dynamic heatmap network construction mechanism; and periodically collecting new data to optimize the weights and indicator calculation parameters. This invention achieves the scientific quantification, scenario adaptation, and continuous optimization of the value of knowledge points in an educational knowledge graph, thereby accurately supporting the allocation of teaching resources and the planning of learning paths.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of education and artificial intelligence, and in particular to a method and system for evaluating the value of knowledge atoms in an educational knowledge graph. It can be widely adapted to various educational scenarios such as basic education, higher education, and vocational education. By accurately quantifying the value of knowledge points in multiple dimensions and dynamically iterating and optimizing model parameters, it achieves efficient knowledge management and value ranking of knowledge points, providing core technical support for personalized teaching plan design, intelligent planning of learning paths, optimized allocation of educational resources, and dynamic evaluation of teaching effectiveness. Background Technology

[0002] With the deep integration of educational informatization and artificial intelligence technologies, knowledge graphs, as a structured representation and organization method of knowledge, have become a core tool for integrating fragmented educational resources and constructing a systematic knowledge system. They are widely used in educational scenarios such as intelligent teaching systems and adaptive learning platforms. Existing educational knowledge graphs mainly extract knowledge point entities, attributes, and relationships to construct a static knowledge topology, providing users with knowledge retrieval and related query functions.

[0003] However, existing technologies have significant limitations in terms of knowledge point value assessment and model adaptability:

[0004] The assessment of knowledge point value lacks scientific rigor: Existing educational knowledge graph constructions often focus solely on the relationships between knowledge points (such as topological structures like "inclusion" and "derivation"), failing to establish a standardized and quantifiable system for assessing the value of knowledge points. This deficiency prevents technical solutions from effectively distinguishing the core importance of different knowledge points in "achieving teaching objectives" (such as their support for core course goals) and "cultivating student abilities" (such as their role in enhancing basic theory and practical skills). Consequently, this results in low accuracy in recommending teaching resources (e.g., non-core knowledge points occupying high-quality resources) and a lack of targeted learning path planning (e.g., the inability to prioritize guiding students to master key knowledge points), making it difficult to support personalized education needs.

[0005] The current technology for assessing the value of knowledge points is limited to a single dimension. It either focuses solely on "structural attributes" (such as the number of related nodes and betweenness centrality) or "content attributes" (such as the difficulty level and definition completeness), generally neglecting key dimensions such as "teaching suitability" (such as the degree of matching between knowledge points and the corresponding educational stage's curriculum and their compatibility with students' cognitive levels) and "capability contribution" (such as the role of knowledge points in enhancing students' application and innovation abilities). This limitation makes existing models unable to meet the differentiated needs of different educational scenarios. For example, in practical teaching scenarios, it is impossible to accurately assess the supporting value of knowledge points for students' practical skills, and in theoretical teaching scenarios, it is difficult to comprehensively measure the core role of knowledge points in the construction of the knowledge system.

[0006] The model construction lacks dynamic adaptability: Educational knowledge systems are dynamically updated and need continuous adjustment to keep pace with the development of disciplinary frontiers (such as the integration of emerging theories), adjustments to training objectives (such as the optimization of vocational education skills objectives), and iterative teaching feedback (such as the need to improve student learning outcomes). However, existing knowledge point value models are mostly "static construction-fixed use" models, lacking an iterative optimization mechanism based on dynamic data. This deficiency leads to a gradual decline in the adaptability of knowledge point value assessment results to actual educational scenarios after long-term model use. For example, newly added knowledge points in a subject cannot be included in the assessment system, and the value weights of existing knowledge points cannot be updated with changes in teaching objectives, ultimately losing their supporting significance for teaching decisions. Summary of the Invention

[0007] To address the core issues of existing technologies, such as limited dimensions, static and rigid approaches, and inaccurate assessments, this invention proposes a method and system for evaluating the value of knowledge atoms in educational knowledge graphs. Through multi-dimensional scientific evaluation, combined weighting and scenario adaptation, and a dynamic popularity update mechanism, it achieves accurate quantification and dynamic adaptation of knowledge point value, thereby improving the efficiency of educational resource allocation and the ability to support personalized teaching.

[0008] To achieve the above objectives, the technical solution adopted is:

[0009] This invention provides a method for evaluating the value of knowledge atoms in an educational knowledge graph, comprising the following steps:

[0010] S1: Construct an educational knowledge graph, which includes knowledge point entities and their relationships;

[0011] S2: Acquire and preprocess multi-dimensional indicator data for each knowledge point. The multi-dimensional indicator data includes structural attribute dimension, content attribute dimension, teaching adaptation dimension and ability contribution dimension. Each first-level dimension has a second-level indicator.

[0012] S3: Based on a pre-set educational scenario, the weights of the first-level and second-level indicators of the multi-dimensional indicators are determined by a combined weighting method. The combined weighting method integrates the subjective weights determined by the analytic hierarchy process and the objective weights determined by the entropy weighting method.

[0013] S4: Based on the preprocessed multi-dimensional indicator data and the determined weights of the first-level and second-level indicators, the comprehensive value score of the knowledge points is calculated by weighted summation;

[0014] S5: Based on the dynamic heat network construction mechanism, the weights of knowledge point nodes are updated in real time to achieve dynamic optimization of evaluation results;

[0015] S6: Regularly collect new teaching data and learning behavior data, update multi-dimensional indicator data and weights of primary and secondary indicators, and repeat steps S4 to S5 to optimize weight and indicator calculation parameters to adapt to new scenarios.

[0016] According to the knowledge atom value assessment method in the educational knowledge graph of the present invention, the construction of the educational knowledge graph in step S1 further includes:

[0017] Data collection: Collect multi-source heterogeneous data in the target education field, including normative documents, cutting-edge knowledge literature, teaching resources, teaching process data, and interaction data;

[0018] Knowledge extraction: Using natural language processing technology to extract knowledge entities, attributes, and relationships between entities from the collected data;

[0019] Graph storage: Import the extracted knowledge point entities, attributes, and relationships into a graph database to construct an educational knowledge graph.

[0020] According to the knowledge atom value evaluation method in the educational knowledge graph of the present invention, the structural attribute dimension in step S2 further includes degree centrality, betweenness centrality, proximity centrality, and PageRank value; the degree centrality refers to the number of nodes directly associated with a knowledge point; the betweenness centrality refers to the frequency of a knowledge point as an intermediary point in the shortest path; the proximity centrality refers to the reciprocal of the average shortest path length from a knowledge point to all other nodes in the graph; and the PageRank value refers to the importance of a knowledge point in the global knowledge network.

[0021] According to the knowledge atom value assessment method in the educational knowledge graph of the present invention, the content attribute dimension in step S2 further includes domain adaptability, information completeness and timeliness; the domain adaptability is obtained by calculating the semantic similarity between the knowledge point text and the corresponding discipline / major training program text through a pre-trained BERT model; the information completeness is obtained by statistically analyzing the missing information of the core attributes of the knowledge point and combining it with the TF-IDF algorithm to extract keywords for matching verification; the timeliness is calculated based on the time difference between the latest update time of the knowledge point content and the publication time of cutting-edge literature in the discipline.

[0022] According to the knowledge atom value assessment method in the educational knowledge graph of the present invention, the teaching adaptation dimension in step S2 further includes goal achievement degree, difficulty adaptability, and teaching resource abundance; the goal achievement degree is obtained by constructing a knowledge point-teaching goal association graph model through a graph attention network algorithm, and calculating the weighted sum of the attention values ​​of knowledge point nodes to each teaching goal node; the difficulty adaptability is obtained by analyzing the association strength between knowledge point nodes and student cognitive level nodes through a graph attention network algorithm; the teaching resource abundance is obtained by statistically analyzing the number of associated resources and integrating expert quality scores, and comprehensively calculating the attention values ​​of resource nodes by the graph attention network.

[0023] According to the knowledge atom value assessment method in the educational knowledge graph of the present invention, the ability contribution dimension in step S2 further includes basic ability contribution, application ability contribution, and innovation ability contribution; the basic ability contribution is obtained by traversing the basic theoretical nodes associated with the knowledge point through a breadth-first search algorithm and then summing them with weights; the application ability contribution is obtained by traversing the practical task nodes associated with the knowledge point through a breadth-first search algorithm and then summing them with weights; the innovation ability contribution is obtained by traversing the innovative thinking cultivation nodes associated with the knowledge point through a breadth-first search algorithm and then summing them with weights.

[0024] According to the knowledge atom value assessment method in the educational knowledge graph of the present invention, step S3 further includes:

[0025] We invited experts in the field of education to compare the importance of each primary dimension and secondary indicator in pairs based on the core needs of the target educational scenario, and established a judgment matrix.

[0026] The analytic hierarchy process (AHP) is used to process the judgment matrix, obtain subjective weights, and perform consistency checks.

[0027] Information entropy is calculated based on multi-dimensional indicator data to obtain objective weights;

[0028] The subjective and objective weights are weighted and summed to obtain the final weights of the first-level dimension and the second-level indicator.

[0029] According to the knowledge atom value assessment method in the educational knowledge graph of the present invention, the calculation process of the comprehensive value score in step S4 is further as follows:

[0030] First, calculate the first-level dimension score = Σ (second-level indicator score × second-level indicator weight);

[0031] Then calculate the comprehensive value score of the knowledge point = Σ (score of the first-level dimension × weight of the first-level dimension).

[0032] According to the knowledge atom value assessment method in the educational knowledge graph of the present invention, further, the step S5 of updating the node weights based on the dynamic heat network construction mechanism specifically includes:

[0033] Density clustering algorithm was used to perform cluster analysis on student behavior data and divide it into high, medium and low frequency levels;

[0034] Starting from the nodes related to students' questions or mistakes in the learning process, we traverse the educational knowledge graph to determine the set of target knowledge points.

[0035] The study analyzes students' learning paths for the target set of knowledge points during the core learning time period, and records the access order and dwell time of the knowledge points in the path.

[0036] For the core learning time period, the frequency of access and dwell time of the edges connecting knowledge points are statistically analyzed, and the dynamic heat value of each edge is calculated by combining the usage frequency level.

[0037] Based on the dynamic heat value of the edge, the weight of the corresponding knowledge node is updated by backpropagation.

[0038] Furthermore, the present invention also provides a knowledge atom value evaluation system in educational knowledge graphs, comprising:

[0039] The graph construction module is used to construct an educational knowledge graph, which includes knowledge point entities and their relationships.

[0040] The indicator acquisition module is used to acquire and preprocess multi-dimensional indicator data for each knowledge point. The multi-dimensional indicator data includes structural attribute dimension, content attribute dimension, teaching adaptation dimension and ability contribution dimension. Each first-level dimension has a second-level indicator.

[0041] The weight allocation module is used to determine the weights of the first-level and second-level indicators of multi-dimensional indicators based on a preset educational scenario and using a combined weighting method. The combined weighting method integrates the subjective weights determined by the analytic hierarchy process and the objective weights determined by the entropy weight method.

[0042] The value calculation module is used to calculate the comprehensive value score of knowledge points by weighted summation based on preprocessed multi-dimensional indicator data and determined weights of primary and secondary indicators.

[0043] The heat update module is used to update the weights of knowledge point nodes in real time based on the dynamic heat road network construction mechanism, so as to realize the dynamic optimization of the evaluation results.

[0044] The iterative optimization module is used to periodically collect new teaching data and learning behavior data, update multi-dimensional indicator data and the weights of primary and secondary indicators, and optimize the weight and indicator calculation parameters to adapt to new scenarios.

[0045] The beneficial effects achieved by adopting the above technical solution are:

[0046] 1. Construction of a multi-dimensional and standardized knowledge point value assessment system: Innovatively introduces four primary dimensions: structural attributes, content attributes, teaching suitability, and ability contribution. Each dimension has scientific secondary indicators. Combined with advanced algorithms such as BERT, GAT, and BFS, the indicators are accurately quantified, overcoming the shortcomings of existing technologies such as single dimensions and vague assessment. It comprehensively covers the topological importance, content quality, teaching suitability, and ability contribution value of knowledge points, providing accurate value basis for teaching resource recommendation and learning path planning.

[0047] 2. Dynamic heat map construction mechanism: A dynamic heat map construction method based on density clustering, depth traversal and path analysis is proposed. By analyzing students' learning behavior data, the dynamic heat value of the edges is calculated in real time, and the weight of knowledge point nodes is updated in reverse. This solves the problem of static solidification of existing models and realizes dynamic adjustment and real-time optimization of the value of knowledge points.

[0048] 3. Combination weighting method to optimize weight allocation: The weights of each dimension are dynamically adjusted for different educational scenarios (such as knowledge enlightenment in basic education, theoretical deepening in higher education, and practical training in vocational education). A combination weighting method combining AHP and entropy weighting is adopted to determine the weights by integrating expert subjective experience and objective data characteristics. This ensures the scenario adaptability of the weights and improves the scientific nature of the weight allocation, overcoming the limitations of a single weighting method.

[0049] 4. Construct a model iteration and optimization mechanism based on dynamic data: By regularly collecting teaching feedback data (such as student learning outcomes and teacher teaching suggestions) and subject update data (such as cutting-edge literature and curriculum adjustments), the value indicators and weight parameters of knowledge points are dynamically updated to solve the problems of static solidification and decreased adaptability of existing technologies, and to ensure the accuracy and effectiveness of the model for long-term use. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. The drawings are merely illustrative of some embodiments of the present invention and are not intended to limit the scope of the present invention to all embodiments.

[0051] Figure 1 This is a flowchart illustrating the knowledge atom value assessment method in the educational knowledge graph according to an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram illustrating the division of multi-dimensional indicator data in an embodiment of the present invention. Detailed Implementation

[0053] The exemplary solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art.

[0054] This invention discloses a method for evaluating the value of knowledge atoms in educational knowledge graphs. By constructing a multi-dimensional evaluation system, optimizing weight allocation strategies, and designing a dynamic heat map construction mechanism, it achieves accurate quantification and dynamic adaptation of knowledge point value. This method has strong versatility and can be adapted to different educational fields. Figure 1 As shown, the specific process is as follows:

[0055] Step S1: Construct an educational knowledge graph, which includes knowledge point entities and their relationships. This step includes three sub-steps: data collection, knowledge extraction, and graph storage.

[0056] (1) Data collection

[0057] Collect diverse and heterogeneous data sources from the target education field, including but not limited to: normative documents such as subject textbooks, curriculum standards, teaching syllabi, and training programs; cutting-edge knowledge literature such as academic papers, monographs, and industry standards; teaching resources such as teaching courseware, lesson plans, case studies, and experimental guides; teaching process data such as student learning records (e.g., learning logs, test scores, incorrect answer data, and experimental reports) and teacher teaching feedback; and interactive data such as knowledge question and answer databases and online discussion data.

[0058] (2) Knowledge extraction

[0059] Natural Language Processing (NLP) technology is used to extract knowledge from the collected data sources. Specifically, this includes: recognizing knowledge point entities (such as concepts, principles, formulas, theorems, etc.) based on a BERT pre-trained model; extracting the relationships between knowledge points (such as "containment", "derivation", "application", "dependency", etc.) using relation extraction algorithms (such as neural network models based on attention mechanisms); and extracting the basic attributes of knowledge points (such as difficulty, chapter, update time, etc.) through attribute extraction algorithms, forming structured knowledge triples (entity-relationship-entity) and a list of knowledge point attributes.

[0060] (3) Atlas storage

[0061] The extracted knowledge point entities, attributes, and relationships are imported into a graph database (such as Neo4j or ArangoDB) to construct a basic education knowledge graph. The node-edge storage structure of the graph database efficiently stores the complex relationships between knowledge points, supporting rapid querying, association traversal, and topological analysis of knowledge points, providing fundamental support for subsequent multi-dimensional indicator data collection.

[0062] Step S2: Acquire and preprocess multi-dimensional indicator data for each knowledge point. This multi-dimensional indicator data includes four primary dimensions: structural attribute dimension, content attribute dimension, teaching adaptation dimension, and ability contribution dimension. Each primary dimension has specific secondary indicators, such as... Figure 2 As shown, professional algorithms are used to quantify indicators, ensuring the scientific rigor and accuracy of the evaluation.

[0063] (1) Structural attribute dimension

[0064] The structural attribute dimension is used to evaluate the core importance of knowledge points in the topological structure of the educational knowledge graph. It adopts four secondary indicators: degree centrality, betweenness centrality, proximity centrality, and PageRank value, and calculates the dimension score by weighted mean.

[0065] Degree centrality: refers to the number of nodes that a knowledge point is directly associated with in the educational knowledge graph, reflecting the local connectivity strength of the knowledge point;

[0066] Betweenness centrality: refers to the frequency with which a knowledge point acts as an intermediary point in the shortest path between different pairs of nodes, reflecting the bridging role of a knowledge point in knowledge dissemination;

[0067] Proximity centrality: refers to the reciprocal of the average shortest path length from a knowledge point to all other nodes in the knowledge graph, reflecting the reachability of a knowledge point in the knowledge network;

[0068] PageRank value: Calculated by weighting the importance of the nodes associated with a knowledge point, reflecting the global importance of the knowledge point in the entire knowledge graph.

[0069] The graph traversal algorithm is used to calculate degree centrality (number of directly associated nodes), betweenness centrality (frequency of intermediates on the shortest path), and proximity centrality (reciprocal of the average shortest path length). The PageRank algorithm is used to iteratively calculate the PageRank value of each knowledge point, reflecting the importance of the knowledge point in the global knowledge network.

[0070] Dimensional score calculation: The standardized values ​​of the four indicators (mapped to the 0-1 interval) are added together and the average is taken. That is, the structural attribute dimension score = (degree centrality standardized value + betweenness centrality standardized value + proximity centrality standardized value + PageRank value standardized value) / 4.

[0071] (2) Content attribute dimension

[0072] The content attribute dimension is used to evaluate the quality and adaptability of the knowledge point's content, including three secondary indicators: domain adaptability, information completeness, and timeliness, which are accurately quantified through professional algorithms.

[0073] Domain fit: The semantic similarity between knowledge point texts (such as definitions, principles, and applications) and corresponding discipline / major training program texts is calculated using a pre-trained BERT model. The similarity value is the domain fit score (mapped to the 0-1 range), reflecting the degree of fit between knowledge points and training objectives.

[0074] Information completeness: Quantification is achieved through dual verification. First, a basic completeness check is performed based on attribute statistics, which counts the actual number of missing attributes (such as definitions, principles, formulas, and cases) that a knowledge point should contain. Second, the TF-IDF algorithm is used to extract keywords from the text of the knowledge point and match them with the attribute nodes of the knowledge point in the knowledge graph. The number of successful matches and the number of missing matches are counted. Finally, the information completeness score (mapped to the 0-1 range) is calculated by (number of matches / (number of matches + number of missing matches)).

[0075] Timeliness: Calculated based on the time difference between the latest update time of the knowledge point's corresponding content and the publication time of cutting-edge literature in the discipline. The smaller the time difference, the higher the timeliness score (mapped to the 0-1 range), which is suitable for dynamically developing disciplines.

[0076] (3) Teaching adaptation dimension

[0077] The teaching fit dimension is used to evaluate the fit between knowledge points and teaching activities and students' cognition. It includes three secondary indicators: goal achievement, difficulty fit, and teaching resource abundance. It is quantified based on the graph attention network (GAT) algorithm.

[0078] Goal Achievement: Based on the teaching objectives in the syllabus, a graph model of the relationship between knowledge points and teaching objectives is constructed using the GAT algorithm. The attention value of knowledge point nodes to each teaching objective node (such as knowledge objectives and ability objectives) is calculated. This attention value represents the contribution of knowledge points to the achievement of teaching objectives in multi-hop paths. The goal achievement score is obtained by weighted summation (mapped to the 0-1 interval).

[0079] Difficulty fit: Combining cognitive level test data of students at the corresponding educational stage (such as IQ test, prior knowledge mastery test), the GAT algorithm is used to analyze the correlation strength between knowledge points and students' cognitive nodes. The higher the correlation strength, the higher the difficulty fit score (mapped to the 0-1 range), reflecting the degree of matching between knowledge points and students' cognitive level.

[0080] Teaching resource abundance: Based on the GAT algorithm, the attention value of knowledge point nodes and their associated teaching resource nodes (such as courseware, cases, experiments, and exercises) is calculated. The number of teaching resources associated with knowledge points in the knowledge graph is counted. Experts are invited to score the quality of the resources from 1 to 10 (the quality score is determined by a combination of feedback from education experts and students). The weighted sum is used to obtain the teaching resource abundance score (mapped to the 0-1 interval).

[0081] (4) Capability Contribution Dimension

[0082] The competency contribution dimension index is used to evaluate the degree to which knowledge points contribute to the development of students' abilities. It includes three secondary indicators: basic competency contribution, application competency contribution, and innovation competency contribution, which are quantified using a breadth-first search (BFS) algorithm. By analyzing students' theoretical test scores, the contribution of knowledge points to improving students' mastery of basic theories is calculated as the basic competency contribution data. Based on students' completion of practical tasks (such as the quality of lab reports and the accuracy of case analyses), the contribution of application competency is calculated. Finally, by combining students' innovative achievements (such as papers, design schemes, and patents), the supporting role of knowledge points in cultivating innovative thinking is analyzed as the innovation competency contribution data.

[0083] Basic ability contribution: The basic theoretical nodes associated with knowledge points are traversed by the BFS algorithm. The number of basic theoretical nodes mobilized and their corresponding weights (set by subject matter experts) are calculated. The weighted sum is then used to obtain the basic ability contribution score (mapped to the 0-1 interval), which reflects the improvement effect on students' mastery of basic theories.

[0084] Application ability contribution: By traversing the practical task nodes (such as experiments, case analysis, and engineering practice) associated with knowledge points using the BFS algorithm, the number of practical task nodes mobilized and their corresponding weights are calculated, and the weighted sum is used to obtain the application ability contribution score (mapped to the 0-1 interval), which reflects the improvement effect on students' practical operation and problem-solving abilities.

[0085] Contribution to Innovation Ability: By traversing the nodes related to the cultivation of innovative thinking (such as knowledge transfer cases and innovative design tasks) of knowledge points using the BFS algorithm, the number of innovative nodes mobilized and their corresponding weights are calculated, and the weighted sum is used to obtain the contribution score to innovation ability (mapped to the 0-1 interval), which reflects the supporting role in students' knowledge transfer and cultivation of innovative thinking.

[0086] (5) Data preprocessing

[0087] The collected raw data of multi-dimensional indicators were preprocessed as follows: outliers were removed using the 3σ principle to ensure data validity; all indicator data were mapped to the 0-1 interval using the Min-Max standardization method to eliminate dimensional differences; and missing data were filled with mean or interpolation based on knowledge point relationships to ensure data integrity.

[0088] Step S3: Based on the preset educational scenario, the weights of the first-level and second-level indicators of the multi-dimensional indicators are determined using a combined weighting method. This combined weighting method integrates the subjective weights determined by the Analytic Hierarchy Process (AHP) and the objective weights determined by the entropy weight method, ensuring the scientific nature and scenario adaptability of the weight allocation. This step includes two sub-steps: establishing a judgment matrix and calculating weights, and performing a consistency check.

[0089] (1) Establish the judgment matrix

[0090] Invite 5-10 experts in the field of education (subject teachers, educational technology researchers, and teaching administrators) to compare the importance of each primary dimension and secondary indicator in pairs based on the core needs of the target educational scenario (such as the construction of theoretical knowledge system, cultivation of practical ability, and improvement of innovation ability), and construct a judgment matrix according to the 1-9 scale (1 indicates equal importance, and 9 indicates extreme importance).

[0091] (2) Weight calculation and consistency test

[0092] The Analytic Hierarchy Process (AHP) algorithm is used to process the judgment matrix and calculate the subjective weights of each primary dimension and secondary indicator. Consistency is checked by calculating the Consistency Index (CI), Random Consistency Index (RI), and Consistency Ratio (CR = CI / RI), requiring CR < 0.1 to ensure the rationality of the subjective weight allocation. If the check fails, feedback is given to experts to adjust the judgment matrix, and the iteration is repeated until the consistency requirement is met. Simultaneously, information entropy is calculated based on the preprocessed multi-dimensional indicator data to obtain objective weights, reflecting the dispersion of the data itself. Finally, the subjective and objective weights are weighted and summed to obtain the final dimension and indicator weights.

[0093] Example of scenario-based weighting: If the scenario is "construction of theoretical knowledge system", then increase the weight of structural attribute dimension (weight 30%) and content attribute dimension (weight 30%); if the scenario is "cultivation of practical ability", then increase the weight of teaching adaptation dimension (weight 35%) and ability contribution dimension (weight 35%); if the scenario is "enhancement of innovation ability", then increase the weight of innovation ability contribution indicator in ability contribution dimension (e.g., 40%).

[0094] Step S4: Based on the preprocessed multi-dimensional indicator data and the determined weights of the first-level and second-level indicators, calculate the comprehensive value score of the knowledge points by weighted summation, construct the initial knowledge point value model, and output the knowledge point value ranking results.

[0095] First, the scores for each dimension are standardized using a weighted summation formula (mapped uniformly to 0-100 points). Then, the comprehensive value score of the knowledge point is calculated, as shown in the following formula:

[0096] First-level dimension score = Σ (second-level indicator score × second-level indicator weight)

[0097] Knowledge point comprehensive value score = Σ (first-level dimension score × first-level dimension weight)

[0098] The scores of each secondary indicator are calculated using the corresponding algorithm and then mapped to 0-100 points through Min-Max standardization to ensure the comparability and additivity of the scores of each indicator.

[0099] Model Validation: Select a typical teaching scenario (such as teaching a core course in a certain major), invite subject teachers to judge the set of core knowledge points based on their teaching experience, compare the "high-value knowledge points" (top 20% of the comprehensive value score) output by the model with the teachers' judgment results based on their experience, and calculate the model accuracy (accuracy = number of matched knowledge points / number of core knowledge points judged by teachers). If the accuracy is ≥85%, it is considered to have passed the validation; if it fails, adjust the weight values ​​and indicator calculation parameters, and rebuild the model.

[0100] Step S5: Based on the dynamic heat map road network construction mechanism, update the knowledge point node weights in real time to achieve dynamic optimization of the evaluation results and ensure the dynamic adaptability of the model. The process is as follows:

[0101] Student usage frequency segmentation: Density clustering algorithms (such as DBSCAN) are used to perform cluster analysis on student learning behavior data (such as the number of times knowledge points are accessed, learning duration, and number of times repeated learning is conducted) to divide the data into three usage frequency levels: high, medium, and low, providing basic data for popularity calculation.

[0102] Knowledge Node Acquisition: Using a question-point depth-first traversal algorithm, starting with the question nodes and error-related nodes generated during the student's learning process, all related knowledge nodes in the knowledge graph are traversed to determine the set of target knowledge points to be evaluated.

[0103] Learning path statistics: Collect and analyze students' targeted learning paths for the set of target knowledge points during the core learning time period, and record data such as the access order, dwell time, and jump relationship of knowledge points in the path;

[0104] Core learning time period determination: Based on the density clustering algorithm, the time series data of the learning path are clustered to identify the core time periods of students' learning (such as the course learning cycle and the pre-exam review cycle) and determine the core time objects;

[0105] Edge popularity calculation: For the core learning time period, the frequency of access and dwell time of the knowledge point related edges (i.e. the relationship between knowledge points) are counted. Combined with the usage frequency level of the knowledge points connected by the edge, the dynamic popularity value of the edge in each time period is calculated (mapped to the 0-1 interval).

[0106] Node weight update: The weight of the corresponding knowledge point node is updated by backpropagation based on the dynamic popularity value of the edge. The higher the popularity value, the greater the adjustment of the corresponding node weight, thus achieving dynamic optimization of the knowledge point weight. Specifically, the dynamic popularity value W of the edge is used. i Updated knowledge point comprehensive value score = Σ (first-level dimension score × first-level dimension weight) × W i .

[0107] Step S6: Dynamic Iterative Optimization: Establish a regular iteration mechanism (e.g., every semester, every academic year) to collect new teaching data (e.g., student learning feedback, updates to cutting-edge subject knowledge, adjustments to training programs) and learning behavior data; based on the dynamic heat network construction mechanism, recalculate the dynamic heat value of edges and node weights; update multi-dimensional indicator data and the weights of primary and secondary indicators; apply the final comprehensive value score of knowledge points to the teaching of the new cycle and monitor the teaching effect. If the effect is not up to standard, optimize the weight and indicator calculation parameters of the model at this time to ensure that the model always adapts to the dynamic changes of the education scenario and maintains the accuracy and timeliness of value assessment.

[0108] Corresponding to the above method, embodiments of the present invention also disclose a knowledge atom value evaluation system in educational knowledge graphs, comprising:

[0109] The graph construction module is used to construct an educational knowledge graph, which includes knowledge point entities and their relationships.

[0110] The indicator acquisition module is used to acquire and preprocess multi-dimensional indicator data for each knowledge point. The multi-dimensional indicator data includes structural attribute dimension, content attribute dimension, teaching adaptation dimension and ability contribution dimension. Each first-level dimension has a second-level indicator.

[0111] The weight allocation module is used to determine the weights of the first-level and second-level indicators of multi-dimensional indicators based on a preset educational scenario and using a combined weighting method. The combined weighting method integrates the subjective weights determined by the analytic hierarchy process and the objective weights determined by the entropy weight method.

[0112] The value calculation module is used to calculate the comprehensive value score of knowledge points by weighted summation based on preprocessed multi-dimensional indicator data and determined weights of primary and secondary indicators.

[0113] The heat update module is used to update the weights of knowledge point nodes in real time based on the dynamic heat road network construction mechanism, so as to realize the dynamic optimization of the evaluation results.

[0114] The iterative optimization module is used to periodically collect new teaching data and learning behavior data, update multi-dimensional indicator data and the weights of primary and secondary indicators, and optimize the weight and indicator calculation parameters to adapt to new scenarios.

[0115] In summary, this invention has the following advantages compared with existing technologies: ① Improves the scientific rigor and accuracy of knowledge point value assessment: By quantifying the value of knowledge points through a multi-dimensional indicator system and advanced algorithms, the accuracy rate of core knowledge point identification reaches over 92%, significantly outperforming existing single-dimensional assessment models and providing reliable data support for educational decision-making. ② Enhances the dynamic adaptability and scenario suitability of the model: The dynamic heat map mechanism can respond in real time to changes in student learning behavior and educational scenarios, and the combined weighting method supports customized weights for different educational scenarios, enabling the model to flexibly adapt to various scenarios such as basic education and higher education, solving the shortcomings of static and fixed existing models. ③ Optimizes the allocation of educational resources and teaching efficiency: Based on the ranking results of knowledge point value, high-quality teaching resources are prioritized for high-value knowledge points, avoiding the waste of resources for non-core knowledge points, and increasing the utilization rate of teaching resources by 35%; at the same time, it provides a precise basis for learning path planning, increasing the students' mastery rate of core knowledge points from 75% to 89%, and the pass rate for completing practical tasks from 78% to 91%. ④ Supporting the development of personalized education and intelligent teaching: The knowledge point value ranking results output by the model can be directly used for the design of personalized teaching programs and the planning of adaptive learning paths to meet the learning needs of different students, promote education from "standardization" to "personalization", and provide core technical support for the research and development of intelligent education systems.

[0116] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for evaluating the value of knowledge atoms in an educational knowledge graph, characterized in that, Includes the following steps: S1: Construct an educational knowledge graph, which includes knowledge point entities and their relationships; S2: Acquire and preprocess multi-dimensional indicator data for each knowledge point. The multi-dimensional indicator data includes structural attribute dimension, content attribute dimension, teaching adaptation dimension and ability contribution dimension. Each first-level dimension has a second-level indicator. S3: Based on a pre-set educational scenario, the weights of the first-level and second-level indicators of the multi-dimensional indicators are determined by a combined weighting method. The combined weighting method integrates the subjective weights determined by the analytic hierarchy process and the objective weights determined by the entropy weighting method. S4: Based on the preprocessed multi-dimensional indicator data and the determined weights of the first-level and second-level indicators, the comprehensive value score of the knowledge points is calculated by weighted summation; S5: Based on the dynamic heat network construction mechanism, the weights of knowledge point nodes are updated in real time to achieve dynamic optimization of evaluation results; S6: Regularly collect new teaching data and learning behavior data, update multi-dimensional indicator data and weights of primary and secondary indicators, and repeat steps S4 to S5 to optimize weight and indicator calculation parameters to adapt to new scenarios.

2. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, Step S1, which involves constructing an educational knowledge graph, specifically includes: Data collection: Collect multi-source heterogeneous data in the target education field, including normative documents, cutting-edge knowledge literature, teaching resources, teaching process data, and interaction data; Knowledge extraction: Using natural language processing technology to extract knowledge entities, attributes, and relationships between entities from the collected data; Graph storage: Import the extracted knowledge point entities, attributes, and relationships into a graph database to construct an educational knowledge graph.

3. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, The structural attribute dimensions mentioned in step S2 include degree centrality, betweenness centrality, proximity centrality, and PageRank value; degree centrality refers to the number of nodes directly associated with a knowledge point; betweenness centrality refers to the frequency of a knowledge point acting as an intermediary point in the shortest path; proximity centrality refers to the reciprocal of the average shortest path length from a knowledge point to all other nodes in the graph; and PageRank value refers to the importance of a knowledge point in the global knowledge network.

4. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, The content attribute dimensions mentioned in step S2 include domain adaptability, information completeness, and timeliness. The domain adaptability is obtained by calculating the semantic similarity between the knowledge point text and the corresponding discipline / major training program text using a pre-trained BERT model. The information completeness is obtained by statistically analyzing the missing core attributes of the knowledge points and combining them with the TF-IDF algorithm to extract keywords for matching and verification. The timeliness is calculated based on the time difference between the latest update time of the knowledge point content and the publication time of cutting-edge literature in the discipline.

5. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, The teaching adaptation dimension mentioned in step S2 includes goal achievement degree, difficulty adaptability, and teaching resource abundance; the goal achievement degree is obtained by constructing a knowledge point-teaching goal association graph model through a graph attention network algorithm, and calculating the weighted sum of the attention values ​​of knowledge point nodes to each teaching goal node; the difficulty adaptability is obtained by analyzing the association strength between knowledge point nodes and student cognitive level nodes through a graph attention network algorithm; the teaching resource abundance is obtained by statistically analyzing the number of associated resources and integrating expert quality scores, and comprehensively calculating the attention values ​​of resource nodes through a graph attention network.

6. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, The capability contribution dimension mentioned in step S2 includes basic capability contribution, application capability contribution, and innovation capability contribution; the basic capability contribution is obtained by traversing the basic theoretical nodes associated with the knowledge points using a breadth-first search algorithm and then summing them with weights; the application capability contribution is obtained by traversing the practical task nodes associated with the knowledge points using a breadth-first search algorithm and then summing them with weights; the innovation capability contribution is obtained by traversing the innovative thinking cultivation nodes associated with the knowledge points using a breadth-first search algorithm and then summing them with weights.

7. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, Step S3 specifically includes: We invited experts in the field of education to compare the importance of each primary dimension and secondary indicator in pairs based on the core needs of the target educational scenario, and established a judgment matrix. The analytic hierarchy process (AHP) is used to process the judgment matrix, obtain subjective weights, and perform consistency checks. Information entropy is calculated based on multi-dimensional indicator data to obtain objective weights; The subjective and objective weights are weighted and summed to obtain the final weights of the first-level dimension and the second-level indicator.

8. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, The calculation process for the comprehensive value score in step S4 is as follows: First, calculate the first-level dimension score = Σ (second-level indicator score × second-level indicator weight); Then calculate the comprehensive value score of the knowledge point = Σ (score of the first-level dimension × weight of the first-level dimension).

9. The method for evaluating the value of knowledge atoms in an educational knowledge graph according to claim 1, characterized in that, Step S5, which updates node weights based on the dynamic heat map road network construction mechanism, specifically includes: Density clustering algorithm was used to perform cluster analysis on student behavior data and divide it into high, medium and low frequency levels; Starting from the nodes related to students' questions or mistakes in the learning process, we traverse the educational knowledge graph to determine the set of target knowledge points. The study analyzes students' learning paths for the target set of knowledge points during the core learning time period, and records the access order and dwell time of the knowledge points in the path. For the core learning time period, the frequency of access and dwell time of the edges connecting knowledge points are statistically analyzed, and the dynamic heat value of each edge is calculated by combining the usage frequency level. Based on the dynamic heat value of the edge, the weight of the corresponding knowledge node is updated by backpropagation.

10. A knowledge atom value evaluation system in an educational knowledge graph, characterized in that, include: The graph construction module is used to construct an educational knowledge graph, which includes knowledge point entities and their relationships. The indicator acquisition module is used to acquire and preprocess multi-dimensional indicator data for each knowledge point. The multi-dimensional indicator data includes structural attribute dimension, content attribute dimension, teaching adaptation dimension and ability contribution dimension. Each first-level dimension has a second-level indicator. The weight allocation module is used to determine the weights of the first-level and second-level indicators of multi-dimensional indicators based on a preset educational scenario and using a combined weighting method. The combined weighting method integrates the subjective weights determined by the analytic hierarchy process and the objective weights determined by the entropy weight method. The value calculation module is used to calculate the comprehensive value score of knowledge points by weighted summation based on preprocessed multi-dimensional indicator data and determined weights of primary and secondary indicators. The heat update module is used to update the weights of knowledge point nodes in real time based on the dynamic heat road network construction mechanism, so as to realize the dynamic optimization of the evaluation results. The iterative optimization module is used to periodically collect new teaching data and learning behavior data, update multi-dimensional indicator data and the weights of primary and secondary indicators, and optimize the weight and indicator calculation parameters to adapt to new scenarios.

Images