Intelligent course personalized resource label classification and recommendation system based on teaching practical experience

Abstract

The invention discloses a smart course personalized resource label classification and recommendation system based on teaching practical experience, and relates to the technical field of smart education, and the system comprises a knowledge graph construction module, a resource labeling module, a user capability modeling module, a recommendation core module, a system interaction module and a parameter optimization module. The working process of the system is as follows: extracting a knowledge point set from a course standard, defining a relationship type between knowledge points, constructing a weighted directed graph, and calculating a shortest path distance between the knowledge points; marking a knowledge point set which is directly explained, a knowledge point set which needs to be mastered in advance, cognitive difficulty and a resource type for resources in the resource library; initializing a user knowledge state matrix, updating the mastery degree according to a test result, and calculating a current capability threshold value of the user; preliminarily screening a candidate resource set, calculating a knowledge migration intensity score, and grouping and sequencing according to resource types to generate a final recommendation list; and adjusting core parameters of a knowledge migration intensity algorithm according to user behavior data generated by a recommendation result.

Description

Technical Field

[0001] This invention belongs to the field of smart education technology, and in particular relates to a smart course personalized resource tag classification and recommendation system based on teaching practice experience. Background Technology

[0002] Existing smart course resource recommendation systems mostly rely on simple tag matching or collaborative filtering algorithms, lacking in-depth mining of the topological structure of subject knowledge systems and making it difficult to quantify the transfer relationships between knowledge points. Resource labeling dimensions are limited, typically only including content topic tags, without clearly defining prerequisite knowledge requirements and cognitive difficulty levels, leading to a mismatch between recommended resources and users' current abilities. User models are static and fixed, unable to dynamically update knowledge mastery based on learning behavior, and do not consider the impact of forgetting curves on knowledge status. Recommendation strategies lack quantitative evaluation of knowledge transfer effects, making it difficult to accurately predict the actual utility of resources for learning target knowledge points. System parameters are fixed, unable to adaptively optimize recommendation algorithms based on user feedback, making it difficult to continuously improve recommendation quality. These problems result in insufficient accuracy in resource recommendations, failing to meet personalized learning needs and impacting learning efficiency and effectiveness. Therefore, we propose a smart course personalized resource tag classification and recommendation system based on teaching practice experience. Summary of the Invention

[0003] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0004] This invention is a smart course personalized resource tag classification and recommendation system based on teaching practice experience, including a knowledge graph construction module, a resource annotation module, a user ability modeling module, a recommendation core module, a system interaction module, and a parameter optimization module;

[0005] The system's workflow is as follows:

[0006] Step S1: Construct a domain knowledge graph: Extract a set of knowledge points from the curriculum standards, define the types of relationships between knowledge points, construct a weighted directed graph, and calculate the shortest path distance between knowledge points;

[0007] Step S2: Learning Resource In-Depth Annotation: For each resource annotation in the resource library, the set of knowledge points directly explained, the set of knowledge points that need to be mastered in advance, the cognitive difficulty, and the resource type;

[0008] Step S3: Dynamic modeling of user capabilities: Initialize the user knowledge state matrix, update the mastery level in real time based on the test results, and calculate the user's current capability threshold;

[0009] Step S4: Transfer-aware resource recommendation generation: Based on knowledge graph, resource tags and user capability model, input user and target knowledge points, initially screen candidate resource set, calculate knowledge transfer strength score, and generate final recommendation list by grouping and sorting by resource type;

[0010] Step S5: Feedback-driven parameter optimization: Dynamically adjust the core parameters of the knowledge transfer intensity calculation algorithm based on user behavior data generated from the recommendation results.

[0011] Furthermore, the output ends of the knowledge graph construction module, the resource annotation module, and the user capability modeling module are all unidirectionally connected to the input end of the recommendation core module. The output end of the recommendation core module is unidirectionally connected to the input end of the system interaction module. The output end of the system interaction module is unidirectionally connected to the input end of the parameter optimization module. The output end of the parameter optimization module is unidirectionally connected to the input end of the recommendation core module.

[0012] The knowledge graph construction module is used to extract knowledge point elements from teaching standard documents, define the logical relationship types between knowledge points, establish topological relationships through expert annotation, and finally generate a weighted knowledge network structure and its distance data between nodes.

[0013] The resource annotation module is used to parse the content attributes of learning resources, identify the set of core knowledge points directly associated with the resources and the set of prerequisite knowledge points required for learning, assess the cognitive complexity level of the resources, and record the media presentation category of the resources.

[0014] The user capability modeling module is used to initialize the learner's knowledge status data, update the knowledge point mastery in real time based on the feedback results of the learner's completion of resource training tasks, and calculate the learning readiness threshold of the target knowledge point by comprehensively considering the mastery of prior knowledge.

[0015] The core recommendation module is used to receive user identifiers and target knowledge point inputs, filter candidate resource sets, quantify resource fit values ​​through an original transfer utility calculation model, and generate the final recommendation sequence after classifying and sorting according to teaching needs.

[0016] The system interaction module is used to coordinate the execution timing of various functional components, allocate computing task resources, process external request instructions, and maintain the version consistency of the algorithm parameter configuration library.

[0017] The parameter optimization module is used to collect user interaction logs with recommended resources, analyze resource usage performance metrics, and periodically iterate and update the key control parameters of the core algorithm.

[0018] Furthermore, step S1 includes the following steps:

[0019] Step S11: Extract the set of knowledge points

[0020] Step S12: Define three types of relations: prerequisite relations, derived relations, and analogical relations;

[0021] Step S13: Subject matter professionals label relation triples (k i ,relation,k j );

[0022] Step S14: Quantify the differences in the impact of different relation types on knowledge transfer and construct a weighted directed graph. And set the edge weights, as shown in the following formula:

[0023]

[0024] In the formula, w ij For knowledge point k i to k j The weights of the edges are: Prerequisite (prerequisite relation) with a weight of 1, Derivation (derivation relation) with a weight of 0.7, and Analogy (analogy relation) with a weight of 0.5.

[0025] Step S15: Support cross-knowledge point transfer strength calculation, and use the Floyd-Warshall algorithm to calculate the shortest path distance Dist between any two knowledge points. G (k i ,k j ), stored as Distance matrix.

[0026] Furthermore, step S2 includes the following steps:

[0027] Step S21: Label the set of knowledge points directly explained by resource r.

[0028] Step S22: The set of knowledge points that need to be mastered in advance for labeling resource r

[0029] Step S23: Quantify the cognitive complexity of resources and label the cognitive difficulty diff(r)∈[1,6] according to Bloom's Taxonomy.

[0030] Step S24: Label the resource type type(r)∈{video, exercise, case, experiment}.

[0031] Furthermore, step S3 includes the following steps:

[0032] Step S31: Initialize the user knowledge state matrix Mastery(u,k) = 0.5;

[0033] Step S32: Update mastery level when user completes resource test: To simulate the forgetting curve effect and dynamically adjust user knowledge status, incremental updates based on practice count and accuracy are adopted, as shown in the following formula:

[0034]

[0035] Mastery new =min(1,Mastery) old +Δm);

[0036] In the formula, Δm represents the change in mastery, the accuracy rate is the user's accuracy rate in completing resource-related test questions, n is the number of historical practice sessions for that knowledge point, and Mastery new For the updated mastery level, Mastery old The level of mastery before the update;

[0037] Step S33: Assess the user's readiness for the learning target knowledge points. Calculate the user's current ability threshold by weighting the mastery of directly prerequisite knowledge points, using the following formula:

[0038]

[0039] In the formula, c(u,k) t For user u, the target knowledge point k t capability threshold, For knowledge point k t The set of direct prerequisite knowledge points, Mastery(u,k) represents the degree of mastery of knowledge point k by user u, and w(k,k) represents the degree of mastery of knowledge point k by user u. t () represents knowledge point k to k t Relationship weights This refers to the number of directly leading knowledge points.

[0040] Furthermore, step S4 includes the following steps:

[0041] Step S41: Input user u and target knowledge point k t ;

[0042] Step S42: Balancing coverage and computational efficiency, initial screening of the candidate set, using the following formula:

[0043]

[0044] In the formula, Let r be the set of candidate resources, k be the learning resources, and k be the learning resources. t For the target knowledge points, k is the set of knowledge points directly explained for resource r. s Dist is any knowledge point directly explained by resource r. G (ks ,k t ) represents knowledge point k s and k t Shortest path distance in a knowledge graph

[0045] Step S43: Quantify the transfer utility value of resources to the target knowledge points, and calculate the score using the knowledge transfer strength model, as shown in the following formula:

[0046]

[0047] In the formula, KTIM(r,k t (u) represents resource r for user u's learning target knowledge point k. t The migration utility value, k is a set of knowledge points related to resource r. s For the relevant knowledge points of resource r, α(k) s ,k t ) represents the knowledge transfer attenuation factor, β(u,k) s ) represents the user mastery gain factor, γ(x) represents the difficulty adaptation function, and diff(r) represents the cognitive difficulty c(u,k) of resource r. t For user u, the target knowledge point k t Capability threshold;

[0048] in:

[0049]

[0050] x = diff(r) - c(u,k) t );

[0051] In the formula, λ is the attenuation coefficient, η is the sensitivity parameter, and σ is the tolerance parameter;

[0052] Step S44: Group resources by resource type and retrieve the Top-N resources to meet diverse teaching needs.

[0053] Control the proportion of single-type resources, and select the top-2 video resource groups;

[0054] Top 5 in the exercise resource group;

[0055] The top 3 cases in the resource group are selected.

[0056] The top two experimental resource groups are selected.

[0057] The resources in each group were sorted in descending order of KTIM score and then merged into the final recommendation list;

[0058] Step S45: Generate a globally sorted recommendation list

[0059] Furthermore, step S5 includes the following steps:

[0060] Step S51: Record user click-through rate (CTR) (r), completion rate (CR) (r), and test score improvement ΔScore (k). t );

[0061] Step S52: Based on the correlation between resource completion rate and knowledge point distance, dynamically update the attenuation coefficient to achieve self-optimization of algorithm parameters. The formula is as follows:

[0062]

[0063] In the formula, λ new Updated attenuation coefficient, λ old Here, μ is the decay coefficient before the update, μ is the learning rate, and CR(r) is the completion rate of resource r. Key knowledge points for resource r To the target knowledge point k t The distance in the knowledge graph. These are the main knowledge points for resource r.

[0064] The present invention has the following beneficial effects:

[0065] 1. This invention constructs a domain knowledge graph, extracts a set of knowledge points from curriculum standards, defines three relationship types: prerequisite, derivation, and analogy, constructs a weighted directed graph, and uses the Floyd-Warshall algorithm to calculate the shortest path distance. This solves the problem of existing systems lacking in-depth mining of the topological structure of subject knowledge systems, can accurately quantify the transfer associations between knowledge points, breaks through the limitations of simple label matching or collaborative filtering, makes recommended resources more in line with the internal logic of knowledge, avoids recommendation bias caused by ignoring knowledge associations, improves the relevance between resources and target knowledge points, and provides a knowledge structure foundation for personalized recommendations.

[0066] 2. This invention provides multi-dimensional, in-depth annotation of learning resources, including direct explanations and sets of knowledge points that need to be mastered in advance, cognitive difficulty labeled according to Bloom's Taxonomy, and resource types such as videos and exercises. At the same time, the user ability modeling module initializes the knowledge state matrix, and dynamically updates the mastery level based on the number of practice sessions and accuracy rate, taking into account the forgetting curve effect, and calculates the ability threshold. This solves the problems of single-dimensional resource annotation and static user model in existing systems, enabling recommended resources to accurately match the user's current ability, avoiding low learning efficiency due to ability mismatch, and allowing learners to obtain resources that match their own level.

[0067] 3. This invention utilizes a transfer-aware resource recommendation generation module to calculate knowledge transfer strength scores based on knowledge graphs, resource tags, and user capability models. It groups resources by type and selects the Top-N resources to meet diverse needs. Simultaneously, a parameter optimization module dynamically adjusts core parameters based on user click-through rates and completion rates, solving the problems of existing systems lacking quantitative assessment of transfer effects and having fixed parameters. This allows for accurate prediction of the actual utility of resources for learning target knowledge points, continuous optimization of recommendation quality, and improved efficiency and effectiveness of personalized learning, better meeting learners' individual needs.

[0068] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0069] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0070] Figure 1 This is a flowchart illustrating the intelligent course personalized resource tag classification and recommendation system based on teaching practice experience of the present invention.

[0071] Figure 2 This is a schematic diagram illustrating the steps of the intelligent course personalized resource tag classification and recommendation system based on teaching practice experience of the present invention. Detailed Implementation

[0072] 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.

[0073] Please see Figure 1 As shown, this invention is a smart course personalized resource tag classification and recommendation system based on teaching practice experience, including a knowledge graph construction module, a resource annotation module, a user ability modeling module, a recommendation core module, a system interaction module, and a parameter optimization module;

[0074] The system's workflow is as follows:

[0075] Step S1: Construct a domain knowledge graph: Extract a set of knowledge points from the curriculum standards, define the types of relationships between knowledge points, construct a weighted directed graph, and calculate the shortest path distance between knowledge points;

[0076] Step S2: Learning Resource In-Depth Annotation: For each resource annotation in the resource library, the set of knowledge points directly explained, the set of knowledge points that need to be mastered in advance, the cognitive difficulty, and the resource type;

[0077] Step S3: Dynamic modeling of user capabilities: Initialize the user knowledge state matrix, update the mastery level in real time based on the test results, and calculate the user's current capability threshold;

[0078] Step S4: Transfer-aware resource recommendation generation: Based on knowledge graph, resource tags and user capability model, input user and target knowledge points, initially screen candidate resource set, calculate knowledge transfer strength score, and generate final recommendation list by grouping and sorting by resource type;

[0079] Step S5: Feedback-driven parameter optimization: Dynamically adjust the core parameters of the knowledge transfer intensity calculation algorithm based on user behavior data generated from the recommendation results.

[0080] The outputs of the knowledge graph construction module, resource annotation module, and user capability modeling module are all unidirectionally connected to the input of the recommendation core module. The output of the recommendation core module is unidirectionally connected to the input of the system interaction module. The output of the system interaction module is unidirectionally connected to the input of the parameter optimization module. The output of the parameter optimization module is unidirectionally connected to the input of the recommendation core module.

[0081] The knowledge graph construction module is used to extract knowledge point elements from teaching standard documents, define the logical relationship types between knowledge points, establish topological relationships through expert annotation, and finally generate a weighted knowledge network structure and its distance data between nodes.

[0082] The resource annotation module is used to parse the content attributes of learning resources, identify the set of core knowledge points directly associated with the resources and the set of prerequisite knowledge points required for learning, assess the cognitive complexity level of the resources, and record the media presentation category of the resources.

[0083] The user capability modeling module is used to initialize learners' knowledge status data, update the mastery of knowledge points in real time based on the feedback results of their completion of resource training tasks, and calculate the learning readiness threshold of the target knowledge point by comprehensively considering the mastery of prior knowledge.

[0084] The core recommendation module receives user identifiers and target knowledge point inputs, filters candidate resource sets, quantifies resource fit values ​​through an original transfer utility calculation model, and generates the final recommendation sequence after classifying and sorting them according to teaching needs.

[0085] The system interaction module is used to coordinate the execution timing of various functional components, allocate computing task resources, process external request instructions, and maintain the version consistency of the algorithm parameter configuration library.

[0086] The parameter optimization module is used to collect user interaction logs with recommended resources, analyze resource usage performance metrics, and periodically iterate and update the key control parameters of the core algorithm.

[0087] Step S1 includes the following steps:

[0088] Step S11: Extract the set of knowledge points

[0089] Step S12: Define three types of relations: prerequisite relations, derived relations, and analogical relations;

[0090] Step S13: Subject matter professionals annotate relation triples (k i ,relation,k j );

[0091] Step S14: Quantify the differences in the impact of different relation types on knowledge transfer and construct a weighted directed graph. And set the edge weights, as shown in the following formula:

[0092]

[0093] In the formula, w ij For knowledge point k i to k j The weights of the edges are: Prerequisite (prerequisite relation) with a weight of 1, Derivation (derivation relation) with a weight of 0.7, and Analogy (analogy relation) with a weight of 0.5.

[0094] Step S15: Support cross-knowledge point transfer strength calculation, and use the Floyd-Warshall algorithm to calculate the shortest path distance Dist between any two knowledge points. G (k i ,k j ), stored as Distance matrix.

[0095] Step S2 includes the following steps:

[0096] Step S21: Label the set of knowledge points directly explained by resource r.

[0097] Step S22: The set of knowledge points that need to be mastered in advance for labeling resource r

[0098] Step S23: Quantify the cognitive complexity of resources and label the cognitive difficulty diff(r)∈[1,6] according to Bloom's Taxonomy.

[0099] Step S24: Label the resource type type(r)∈{video, exercise, case, experiment}.

[0100] Step S3 includes the following steps:

[0101] Step S31: Initialize the user knowledge state matrix Mastery(u,k) = 0.5;

[0102] Step S32: Update mastery level when user completes resource test: To simulate the forgetting curve effect and dynamically adjust user knowledge status, incremental updates based on practice count and accuracy are adopted, as shown in the following formula:

[0103]

[0104] Mastery new =min(1,Mastery) old +Δm);

[0105] In the formula, Δm represents the change in mastery, the accuracy rate is the user's accuracy rate in completing resource-related test questions, n is the number of historical practice sessions for that knowledge point, and Mastery new For the updated mastery level, Mastery old The level of mastery before the update;

[0106] Step S33: Assess the user's readiness for the learning target knowledge points. Calculate the user's current ability threshold by weighting the mastery of directly prerequisite knowledge points, using the following formula:

[0107]

[0108] In the formula, c(u,k) t For user u, the target knowledge point k t capability threshold, For knowledge point k t The set of direct prerequisite knowledge points, Mastery(u,k) represents the degree of mastery of knowledge point k by user u, and w(k,k) represents the degree of mastery of knowledge point k by user u. t () represents knowledge point k to k t Relationship weights This refers to the number of directly leading knowledge points.

[0109] Step S4 includes the following steps:

[0110] Step S41: Input user u and target knowledge point k t ;

[0111] Step S42: Balancing coverage and computational efficiency, initial screening of the candidate set, using the following formula:

[0112]

[0113] In the formula, Let r be the set of candidate resources, k be the learning resources, and k be the learning resources. t For the target knowledge points, k is the set of knowledge points directly explained for resource r. s Dist is any knowledge point directly explained by resource r. G (k s ,k t ) represents knowledge point k s and k t Shortest path distance in a knowledge graph

[0114] Step S43: Quantify the transfer utility value of resources to the target knowledge points, and calculate the score using the knowledge transfer strength model, as shown in the following formula:

[0115]

[0116] In the formula, KTIM(r,k t (u) represents resource r for user u's learning target knowledge point k. t The migration utility value, k is a set of knowledge points related to resource r. s For the relevant knowledge points of resource r, α(k) s ,k t ) represents the knowledge transfer attenuation factor, β(u,k) s ) represents the user mastery gain factor, γ(x) represents the difficulty adaptation function, and diff(r) represents the cognitive difficulty c(u,k) of resource r. t For user u, the target knowledge point k t Capability threshold;

[0117] in:

[0118]

[0119] x = diff(r) - c(u,k) t );

[0120] In the formula, λ is the attenuation coefficient, η is the sensitivity parameter, and σ is the tolerance parameter;

[0121] Step S44: Group resources by resource type and retrieve the Top-N resources to meet diverse teaching needs.

[0122] Control the proportion of single-type resources, and select the top-2 video resource groups;

[0123] Top 5 in the exercise resource group;

[0124] The top 3 cases in the resource group are selected.

[0125] The top two experimental resource groups are selected.

[0126] The resources in each group were sorted in descending order of KTIM score and then merged into the final recommendation list;

[0127] Step S45: Generate a globally sorted recommendation list

[0128] Step S5 includes the following steps:

[0129] Step S51: Record user click-through rate (CTR) (r), completion rate (CR) (r), and test score improvement ΔScore (k). t );

[0130] Step S52: Based on the correlation between resource completion rate and knowledge point distance, dynamically update the attenuation coefficient to achieve self-optimization of algorithm parameters. The formula is as follows:

[0131]

[0132] In the formula, λ new Updated attenuation coefficient, λ old Here, μ is the decay coefficient before the update, μ is the learning rate, and CR(r) is the completion rate of resource r. Key knowledge points for resource r To the target knowledge point k t The distance in the knowledge graph. These are the main knowledge points for resource r.

[0133] One specific application of this embodiment is:

[0134] Target users: First-year students in vocational colleges, whose current learning goal is to master "subnetting technology";

[0135] System input: User ID, target knowledge point "k" t =Subnetting techniques;

[0136] System output: A personalized list of recommended learning resources generated for the user;

[0137] Step 1

[0138] 1. Knowledge Point Extraction: Extract a set of knowledge points related to "Network Fundamentals" from "University Information Technology Fundamentals" using the format k = {k1 = "IP Address Format", k2 = "Binary Conversion", k3 = "CIDR Notation", k4 = "Subnetting Techniques", k t =“Subnetting technology”, k5 = “Routing configuration”, k6 = “MAC address function”;

[0139] 2. Prerequisite relationship: k1→k t (Prerequisite: Subnetting requires understanding the IP address structure); k2→k t(Prerequisite: Binary conversion is the basis for calculating subnet masks);

[0140] Derivational relation: k3→k t (Derivative, CIDR is an extension of subnetting); k t →k5 (derived from k5, routing needs to be configured after subnetting);

[0141] Analogy: k t →k6 (Analogous, both subnetting and MAC addresses involve network identification);

[0142] 3. Set edge weights according to relation type: prerequisite relation weight w = 1, derived relation weight w = 0.7, analogy relation weight w = 0.5, and construct a directed graph.

[0143] 4. Calculate the distance between any two knowledge points using the Floyd-Warshall algorithm:

[0144] Dist G (k1,k t ) = 1 (direct prerequisite relationship);

[0145] Dist G (k3,k t ) = 1; (k3→k4→k t (Path weight 7+1=1.7);

[0146] Dist G (k6,k t =1; (Analogy relationship directly connected);

[0147] Step Two

[0148] Five typical resources in the resource repository were labeled:

[0149] Resource r1 (video):

[0150] Directly explain the knowledge points

[0151] Knowledge points to be mastered in advance

[0152] Cognitive difficulty diff(r1) = 3; (Bloom's Taxonomy "Application Level");

[0153] type(r1) = video;

[0154] Resource r2 (Exercises):

[0155]

[0156] diff(r2) = 4; (analysis level);

[0157] type(r2) = Exercises;

[0158] Resource r3 (Example):

[0159]

[0160] diff(r3) = 3 (application level);

[0161] type(r3) = case;

[0162] Resource r4 (experimental):

[0163]

[0164] diff(r4) = 2; (Understanding level);

[0165] type(r4) = experiment;

[0166] Resource R5 (video):

[0167]

[0168] diff(r5) = 3 (application level);

[0169] type(r5) = video(and k) t As an analogy, the distance Dist = 1);

[0170] Step 3

[0171] 1. The student's knowledge state matrix is ​​initialized as Mastery(student, k) = 0.5;

[0172] 2. Students completed (k2) related tests: 5 practice sessions (n=5), with an accuracy rate of 85%;

[0173] calculate

[0174] After the update, Mastery(student, k2) = 0.599;

[0175] Similarly, complete the completing the square (k1) test: 3 practice sessions (n=3), 70% accuracy, updated as follows:

[0176] Mastery(student, k4)≈0.554;

[0177] 3. Target knowledge point k t The direct prerequisite knowledge point N(k) t )={k2,k4}, relation weight w(k2,k t )=1,w(k4,kt ) = 1;

[0178] Capability threshold:

[0179]

[0180] Step Four

[0181] 1. Filtering criteria: Resources directly explain k t or related knowledge points and k t Distance ≤ 3;

[0182] Candidate set (k6 and k) t Distance = 1 ≤ 3 (meets the condition);

[0183] 2. Relevant knowledge point r1: KTIM(r 1, k t (students) ≈ 0.008;

[0184] The calculated values ​​for r5 are: K(r5) = {k6}, KTIM = 0.002 (analogy decay);

[0185] 3. Video group (r1): Take the top 2 → r1 in descending order of KTIM;

[0186] Problem set (r2): Select the Top 5 (only 1, keep r2);

[0187] Case group (r3): Take the top 3 (only 1, keep r3);

[0188] Experimental group (r4): Take the top 2 (only 1, keep r4);

[0189] 4. Final Recommendation List:

[0190] Step 5

[0191] 1. Behavioral data collection:

[0192] Completion rates: CR(r1) = 100%, CR(r5) = 20%, CR(r4) = 60%;

[0193] Distance between key knowledge points: (such as r5) Dist = 1);

[0194] 2. Let the learning rate μ = 0.1, and the original λ old =0.8, then:

[0195]

[0196] After the update, the system is more sensitive to nearby knowledge resources, improving the accuracy of subsequent recommendations.

[0197] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0198] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A smart curriculum personalized resource tagging and recommendation system based on teaching practice experience, characterized by: It includes a knowledge graph construction module, a resource annotation module, a user capability modeling module, a core recommendation module, a system interaction module, and a parameter optimization module; The system's workflow is as follows: Step S1: Construct a domain knowledge graph: Extract a set of knowledge points from the curriculum standards, define the types of relationships between knowledge points, construct a weighted directed graph, and calculate the shortest path distance between knowledge points; Step S2: Learning Resource In-Depth Annotation: For each resource annotation in the resource library, the set of knowledge points directly explained, the set of knowledge points that need to be mastered in advance, the cognitive difficulty, and the resource type; Step S3: Dynamic modeling of user capabilities: Initialize the user knowledge state matrix, update the mastery level in real time based on the test results, and calculate the user's current capability threshold; Step S4: Transfer-aware resource recommendation generation: Based on knowledge graph, resource tags and user capability model, input user and target knowledge points, initially screen candidate resource set, calculate knowledge transfer strength score, and generate final recommendation list by grouping and sorting by resource type; Step S5: Feedback-driven parameter optimization: Dynamically adjust the core parameters of the knowledge transfer intensity calculation algorithm based on user behavior data generated from the recommendation results.

2. The intelligent course personalized resource tag classification and recommendation system based on teaching practice experience as described in claim 1, characterized in that, The output ends of the knowledge graph construction module, the resource annotation module, and the user capability modeling module are all unidirectionally connected to the input end of the recommendation core module. The output end of the recommendation core module is unidirectionally connected to the input end of the system interaction module. The output end of the system interaction module is unidirectionally connected to the input end of the parameter optimization module. The output end of the parameter optimization module is unidirectionally connected to the input end of the recommendation core module. The knowledge graph construction module is used to extract knowledge point elements from teaching standard documents, define the logical relationship types between knowledge points, establish topological relationships through expert annotation, and finally generate a weighted knowledge network structure and its distance data between nodes. The resource annotation module is used to parse the content attributes of learning resources, identify the set of core knowledge points directly associated with the resources and the set of prerequisite knowledge points required for learning, assess the cognitive complexity level of the resources, and record the media presentation category of the resources. The user capability modeling module is used to initialize the learner's knowledge status data, update the knowledge point mastery in real time based on the feedback results of the learner's completion of resource training tasks, and calculate the learning readiness threshold of the target knowledge point by comprehensively considering the mastery of prior knowledge. The core recommendation module is used to receive user identifiers and target knowledge point inputs, filter candidate resource sets, quantify resource fit values ​​through an original transfer utility calculation model, and generate the final recommendation sequence after classifying and sorting according to teaching needs. The system interaction module is used to coordinate the execution timing of various functional components, allocate computing task resources, process external request instructions, and maintain the version consistency of the algorithm parameter configuration library. The parameter optimization module is used to collect user interaction logs with recommended resources, analyze resource usage performance metrics, and periodically iterate and update the key control parameters of the core algorithm.

3. The intelligent course personalized resource tag classification and recommendation system based on teaching practice experience as described in claim 1, characterized in that, Step S1 includes the following steps: Step S11: Extract the set of knowledge points Step S12: Define three types of relations: prerequisite relations, derived relations, and analogical relations; Step S13: Subject matter professionals annotate relation triples (k i ,relation,k j ); Step S14: Quantify the differences in the impact of different relation types on knowledge transfer and construct a weighted directed graph. And set the edge weights, as shown in the following formula: In the formula, w ij For knowledge point k i to k j The weights of the edges are: Prerequisite (prerequisite relation) with a weight of 1, Derivation (derivation relation) with a weight of 0.7, and Analogy (analogy relation) with a weight of 0.

5. Step S15: Support cross-knowledge point transfer strength calculation, and use the Floyd-Warshall algorithm to calculate the shortest path distance Dist between any two knowledge points. G (k i ,k j ), stored as Distance matrix.

4. The intelligent course personalized resource tag classification and recommendation system based on teaching practice experience as described in claim 1, characterized in that, Step S2 includes the following steps: Step S21: Label the set of knowledge points directly explained by resource r. Step S22: The set of knowledge points that need to be mastered in advance for labeling resource r Step S23: Quantify the cognitive complexity of resources and label the cognitive difficulty diff(r)∈[1,6] according to Bloom's Taxonomy; Step S24: Label the resource type type(r)∈{video, exercise, case, experiment}.

5. The intelligent course personalized resource tag classification and recommendation system based on teaching practice experience according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Initialize the user knowledge state matrix Mastery(u,k) = 0.5; Step S32: Update mastery level when user completes resource test: To simulate the forgetting curve effect and dynamically adjust user knowledge status, incremental updates based on practice count and accuracy are adopted, as shown in the following formula: Mastery new =min(1,Mastery old +Δm); In the formula, Δm represents the change in mastery, the accuracy rate is the user's accuracy rate in completing resource-related test questions, n is the number of historical practice sessions for that knowledge point, and Mastery new For the updated mastery level, Mastery old The level of mastery before the update; Step S33: Assess the user's readiness for the learning target knowledge points. Calculate the user's current ability threshold by weighting the mastery of directly prerequisite knowledge points, using the following formula: In the formula, c(u,k) t For user u, the target knowledge point k t capability threshold, For knowledge point k t The set of direct prerequisite knowledge points, Mastery(u,k) represents the degree of mastery of knowledge point k by user u, and w(k,k) represents the degree of mastery of knowledge point k by user u. t () represents knowledge point k to k t Relationship weights This refers to the number of directly leading knowledge points.

6. The intelligent course personalized resource tag classification and recommendation system based on teaching practice experience according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Input user u and target knowledge point k t ; Step S42: Balancing coverage and computational efficiency, initial screening of the candidate set, using the following formula: In the formula, Let r be the set of candidate resources, k be the learning resources, and k be the learning resources. t For the target knowledge points, k is the set of knowledge points directly explained for resource r. s Dist is any knowledge point directly explained by resource r. G (k s ,k t ) represents knowledge point k s and k t Shortest path distance in a knowledge graph Step S43: Quantify the transfer utility value of resources to the target knowledge points, and calculate the score using the knowledge transfer strength model, as shown in the following formula: In the formula, KTIM(r,k t (u) represents resource r for user u's learning target knowledge point k. t The migration utility value, k is a set of knowledge points related to resource r. s For the relevant knowledge points of resource r, α(k) s ,k t ) represents the knowledge transfer attenuation factor, β(u,k) s ) represents the user mastery gain factor, γ(x) represents the difficulty adaptation function, and diff(r) represents the cognitive difficulty c(u,k) of resource r. t For user u, the target knowledge point k t Capability threshold; in: x=diff(r)-c(u,k t ); In the formula, λ is the attenuation coefficient, η is the sensitivity parameter, and σ is the tolerance parameter; Step S44: Group resources by resource type and retrieve the Top-N resources to meet diverse teaching needs: Control the proportion of single-type resources, and select the top-2 video resource groups; Top 5 in the exercise resource group; The top 3 cases in the resource group are selected. The top two experimental resource groups are selected. The resources in each group were sorted in descending order of KTIM score and then merged into the final recommendation list; Step S45: Generate a globally sorted recommendation list 7. The intelligent course personalized resource tag classification and recommendation system based on teaching practice experience according to claim 1, characterized in that, Step S5 includes the following steps: Step S51: Record user click-through rate (CTR) (r), completion rate (CR) (r), and test score improvement ΔScore (k). t ); Step S52: Based on the correlation between resource completion rate and knowledge point distance, dynamically update the attenuation coefficient to achieve self-optimization of algorithm parameters. The formula is as follows: In the formula, λ new Updated attenuation coefficient, λ old Here, μ is the decay coefficient before the update, μ is the learning rate, and CR(r) is the completion rate of resource r. Key knowledge points for resource r To the target knowledge point k t The distance in the knowledge graph. These are the main knowledge points for resource r.

Images