A knowledge graph-based adaptive learning path planning method and system

Abstract

This invention provides a knowledge graph-based adaptive learning path planning method and system. The method includes: acquiring a course knowledge graph and the target learner's cognitive mastery of knowledge points in the current course, generating a cognitive difference subgraph; extracting knowledge points and cognitive dependencies between them from the cognitive difference subgraph, generating an executable matrix including all executable learning paths; acquiring the target learner's path transfer value for each learning path, generating a path transfer matrix including all executable learning paths; generating a path benefit matrix including all executable learning paths based on the executable matrix and the path transfer matrix, obtaining the learning benefit value for each executable learning path; and obtaining the target learner's adaptive learning path based on the learning benefit value. This invention enhances the rationality of adaptive learning path planning in online learning, ensuring the connectivity and rationality of learners' online learning.

Description

Technical Field

[0001] This invention relates to the field of computer application technology, and in particular to an adaptive learning path planning method and system based on knowledge graphs. Background Technology

[0002] The explosive growth of digital resources brought about by the rapid development of information technology has led to problems such as "knowledge disorientation" and "information overload" for learners. Learners need to expend considerable energy searching for and identifying the information they need from complex knowledge systems and massive amounts of online resources, resulting in poor learning outcomes and low efficiency. Most existing online learning platforms are merely linear compilations of resources or only offer simple categorization or search functions, failing to provide learners with adaptive learning path planning and thus unable to effectively solve these problems.

[0003] Currently, mainstream learning path planning methods mainly include: individual feature-based learning path planning methods and group collaborative filtering-based learning path planning methods. Individual feature-based learning path planning methods plan learning paths based on learners' individual learning preferences, ability characteristics, and historical records, and according to the degree of matching between computational resources and individual characteristics, using algorithms such as genetic algorithms and particle swarm optimization. Group collaborative filtering-based learning path planning methods, on the other hand, are based on big data technology, using collaborative filtering algorithms to mine learning path data from similar student groups to generate learning paths.

[0004] However, while these methods can generate personalized learning paths to some extent, they ignore the fact that learning is a dynamic process, learners' cognitive characteristics evolve dynamically, and there are connections and cognitive dependencies between knowledge points in different subjects. This can result in redundant or irrelevant learning resources in the planned learning paths, violating learners' cognitive patterns and failing to guarantee the quality of the learning paths and the learners' learning benefits. Summary of the Invention

[0005] This invention provides an adaptive learning path planning method and system based on knowledge graphs, which solves the problem that existing path planning methods ignore the learner's cognitive dynamic development law and the correlation and cognitive dependence between subject knowledge points, resulting in low quality and inaccuracy of learning paths.

[0006] On one hand, this invention provides an adaptive learning path planning method based on knowledge graphs, the method comprising:

[0007] Obtain the course knowledge graph and the target learners' cognitive mastery of the knowledge points in the current course, and generate a cognitive difference subgraph;

[0008] Extract knowledge points and cognitive dependencies between knowledge points from the cognitive difference subgraph to generate an executable matrix that includes all executable learning paths;

[0009] Obtain the path transfer value of the target learner for each learning path, and generate a path transfer matrix that includes all executable learning paths;

[0010] Based on the executable matrix and the path migration matrix, a path benefit matrix including all executable learning paths is generated to obtain the learning benefit value of each executable learning path.

[0011] Based on the learning gains, an adaptive learning path can be obtained for the target learner.

[0012] Furthermore, the step of generating the executable matrix includes:

[0013] Based on the extraction rules, knowledge points and cognitive dependencies between knowledge points are extracted from the cognitive difference subgraph, resulting in a set of knowledge points and a set of dependencies.

[0014] A direct dependency matrix is ​​generated based on the set of knowledge points and the set of dependencies.

[0015] The direct dependency matrix is ​​added to the identity matrix of the same dimension, and n+1 Boolean powers are performed to obtain the total dependency matrix; where n is a positive integer.

[0016] The executable matrix is ​​generated based on the full dependency matrix.

[0017] Furthermore, the extraction rules include:

[0018] First rule: For knowledge entity K x K y If a priori-successor relationship exists <K x ,Requires,K y Parent-son relationship <K x SubClassof,K y Whole-part relationship <K x PartOf,K y If any one or more relations are present in the >, then the dependency relation is... <K x ,R,K y >Established;

[0019] Second rule: For knowledge point entity K x K y If there is a sibling relationship <K x BrotherClass,K y >Relationships <K xRelatedOf,K y If any one or more relations are present in the >, then the dependency relation is... <K x ,R,K y >、 <K y ,R,K x >Established;

[0020] Third rule: For knowledge cluster G m Knowledge Cluster G n and knowledge point K x If a hierarchical relationship exists <G m ,PartGsOf,G n > Inclusion relationship <K x ,BelongTo,G n Any one or more relations in >, then the dependency relation <K x ,BelongTo,G m >Established;

[0021] Fourth rule: For knowledge cluster G m and knowledge cluster G n If there is an order relationship <G m NextGsOf,G n > For all Dependency <K x ,R,K y Established.

[0022] Furthermore, the generation of the total dependency matrix includes:

[0023] The dimension of the direct dependency matrix A is determined based on the number of elements in the knowledge point set, and the value of the direct dependency matrix A is determined based on the dependency set.

[0024] The directly dependent matrix A is added to the identity matrix I of the same dimension (n×n), and then subjected to n+1 Boolean exponentiation operations; where n is a positive integer, starting from 1.

[0025] If there exists an n that satisfies the formula

[0026] Then the total dependency matrix M is (A+I). n .

[0027] Furthermore, generating the executable matrix S based on the full dependency matrix M includes:

[0028] Use the matrix of the same dimension as the full dependency matrix M as the initial executable matrix;

[0029] Obtain the total dependency matrix M based on the formula Update the value of each element in the initial executable matrix to obtain the executable matrix S.

[0030] Furthermore, the path migration value V a The methods for obtaining [x][y] include:

[0031]

[0032] Among them, V a [x][y] represents the transfer value of learner a along the learning path from knowledge point x to knowledge point y; sim ai The similarity between target learner a and reference learner i; This indicates that the set of learning paths included in the reference learning group contains the learning path x→y; score i [x][y] represents the learning value of reference learner i on the learning path x→y; h represents the number of reference learners i with the learning path x→y; This represents the mean of the transfer values ​​across all learning paths.

[0033] Furthermore, the path revenue matrix η a The generation method includes: converting the path migration matrix V a Calculate the Hadamard product with the executable matrix S.

[0034] Furthermore, the improved ant colony path planning algorithm utilizing the path reward matrix includes:

[0035] The path reward matrix is ​​used as the heuristic information matrix for the ant colony algorithm; a virtual node is set as the initial node of the ant colony algorithm, and the virtual node represents the cognitive mastery of the target learner at the current stage.

[0036] The pathfinding process is iterative. In each iteration, each ant searches for a path based on the pheromone concentration and the learning reward value of each executable learning path. If, during the pathfinding process, an ant is at knowledge node K at a certain moment... i At that time, the ant had not learned the set of knowledge points K. j set, and there is no knowledge node K in the executable matrix. i →Knowledge Section K j and / or no knowledge section K exists i →Knowledge Point Set K j If set, the executable path will be discarded, and no information concentration update or learning gain calculation will be performed;

[0037] In a single iteration, the learning reward value is calculated for all ants that have completed the current iteration;

[0038] The executable learning path with the highest learning benefit value is selected as the optimal learning path for the target learner.

[0039] Furthermore, the learning gain value Patc_val a The calculation methods include:

[0040]

[0041] Among them, Patc_val a This represents the learning gain value obtained by learner a after completing the learning of all knowledge points in the cognitive difference subgraph once; vis_kg a This represents the learning sequence of knowledge points in this iteration of the ant; vis_kg a [x] represents the knowledge point at position x; vis_kg a [x+1] represents the next knowledge point at position x; η[vis_kg] a [x]][vis_kg a [x+1] represents the knowledge point vis_kg a [x] to the knowledge point vis_kg a The learning efficiency of [x+1].

[0042] On the other hand, the present invention also provides an adaptive learning path planning system based on knowledge graphs, which includes at least a course knowledge graph construction module, a learning process data storage module, a learner dynamic cognitive management module, an adaptive learning path planning module, and an adaptive learning path recommendation module, for performing the steps of any of the above methods.

[0043] In general, the technical solution conceived in this invention can achieve the following beneficial effects compared with the prior art:

[0044] (1) The present invention also provides an adaptive learning path planning method and system based on knowledge graphs. It uses course knowledge graphs to model the relationship between knowledge, and based on this, it infers all learning paths that violate the cognitive order. Then, it uses these paths as taboo paths in path planning, thereby obtaining executable learning paths. This enhances the rationality of adaptive learning path planning in online learning and ensures the connectivity and rationality of learners' online learning.

[0045] (2) This invention also provides an adaptive learning path planning method and system based on knowledge graphs. It uses a method for calculating the learning benefit value of individual learners, and based on the calculated executable path, it uses the ant colony algorithm to calculate the knowledge learning path with the highest learning benefit value and recommends it to the learner. Based on the knowledge and online resource dependencies in the course knowledge graph, it uses the adaptive knowledge learning path calculated by the target learner's feature data to recommend adaptive resource learning paths to the learner. This solves the problems of knowledge confusion and resource overload faced by target learners in online learning. Furthermore, based on the dependencies between knowledge, it guides the planning of learning paths for online learning, solving the problems of weak correlation and insufficient rationality in the adaptive learning path planning of existing research. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating the adaptive learning path planning method and system based on knowledge graphs provided by the present invention.

[0048] Figure 2 This is a schematic diagram of a course knowledge graph for an adaptive learning path planning method and system based on knowledge graphs provided by the present invention.

[0049] Figure 3 This is a schematic diagram of the cognitive difference subgraph of a target learner in an adaptive learning path planning method and system based on knowledge graphs provided by the present invention.

[0050] Figure 4 This is a diagram illustrating the calculation process of the total dependency matrix of an adaptive learning path planning method and system based on knowledge graphs provided by this invention.

[0051] Figure 5 This is a schematic diagram of the system structure of an adaptive learning path planning method and system based on knowledge graphs provided by the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0053] It should be noted that, in the description of the embodiments of the present invention, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, or system. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, or system that includes said element.

[0054] This invention proposes an adaptive learning path planning method and system based on knowledge graphs. By establishing subject knowledge and multidimensional dependencies between knowledge in a course knowledge graph, and utilizing matrix reasoning and calculation of learning benefit values, it deeply mines the dependencies between knowledge points and correlates the cognitive characteristics of the target learner with the benefits of learning knowledge points. This provides the target learner with a reasonable, accurate, and highly adaptive executable learning path, thereby overcoming the technical shortcomings of existing learning path planning methods.

[0055] like Figure 1 As shown, the specific methods include:

[0056] Step S101: Obtain the course knowledge graph and the target learners' cognitive mastery of the knowledge points in the current course, and generate a cognitive difference subgraph.

[0057] As one embodiment of the present invention, such as Figure 2 As shown, the course knowledge graph includes a teaching organization layer based on knowledge cluster entities, a knowledge representation layer based on knowledge point entities, and a resource aggregation layer based on resource entities.

[0058] Specifically, based on the characteristics of the course and the laws of teaching, the three-layer entity system—knowledge cluster entity, knowledge point entity, and online learning resource entity—as well as the inter-layer and intra-layer relationships of each entity are extracted; and a course knowledge graph is constructed using the knowledge cluster entity, knowledge point entity, and online learning resource entity.

[0059] It should be noted that, based on the course characteristics and the learning needs of the target learners, the course knowledge graph can be denoted as KG = (E, R); where E is the set of entities and R is the set of relations; the course knowledge graph contains three types of entities: knowledge clusters G, knowledge points K, and online learning resources Res, so the set of entities is E = (G, K, Res).

[0060] The instructional organization layer models the instructional knowledge clusters and their sequence. This layer models abstract knowledge cluster entities and their inter- and intra-layer relationships. A knowledge cluster entity refers to a meaningful knowledge module composed of n knowledge points arranged according to a specific instructional logic. For example, the set of knowledge clusters contained in a course can be denoted as G = {g1, g2, ..., g...}. x}, where g i It represents a specific knowledge cluster entity.

[0061] The knowledge representation layer models the course knowledge and its inherent logical dependencies. The entities modeled in this layer are specific knowledge points and their inter- and intra-layer relationships. A knowledge point entity is an indivisible, independent unit with complete knowledge content representation. For example, the set of knowledge points included in a course can be denoted as K = {k1, k2, ..., k}. y}; where k i It represents a specific knowledge point entity.

[0062] The resource aggregation layer models the online learning resources related to this course. The entities modeled in this layer are multimodal online learning resource entities and their inter-layer and intra-layer relationships. Online learning resource entities refer to accessible digital resources on the internet. Res represents the set of online resources in this course domain, which can be denoted as Res = {res1, res2, ..., res...} z}; where res i It represents a specific resource entity.

[0063] Based on the dependencies and sequence characteristics between course knowledge points, the relationships between three types of entities in the course knowledge graph of this course domain are extracted.

[0064] For example, the set of relations contained in a course is denoted as R = {r1, r2, ..., r3}. m}. Wherein, each association r i Each can be represented as a relation triple <entity, relation, entity>.

[0065] Specifically, the relationships include any one or more of the following: intra-layer relationships and inter-layer relationships.

[0066] As a specific embodiment of the present invention, the intra-layer relationship includes: the sequential relationship between knowledge clusters, denoted as... <g iNextGsOf,g j >, where knowledge cluster g j The pre-existing knowledge cluster is g i The hierarchical relationship between knowledge clusters is denoted as... <g i ,PartGsOf,g j >, where knowledge cluster g i Includes knowledge cluster g j The parent-child relationship between knowledge points is denoted as... <k i ,SubClassof,k j >, where knowledge point k j It is knowledge point k i Sub-knowledge points; the whole-part relationship between knowledge points, denoted as... <k i ,PartOf,k j >, where knowledge point k j It is knowledge point k i The components; the prior-successor relationships between knowledge points, denoted as <k i ,Requires,k j >, where knowledge point k i It is knowledge point k j Prior knowledge; the sibling relationship between knowledge points, denoted as <k i BrotherClass,k j >, where knowledge point k i and knowledge point k j They are siblings; the relationship between knowledge points is denoted as... <k i RelatedOf,k j >, where knowledge point k i and knowledge point k j There is a correlation; the order in which online learning resources are arranged within the existing online platform is denoted as... <res i NextS,res j >, where resource res i Arranged in resources res j Before.

[0067] Inter-layer relationships include the inclusion relationship between knowledge clusters and knowledge points, denoted as... <g i ,BelongTo,k j >, where knowledge cluster g i Includes knowledge point k j The relationship between knowledge points and online learning resources is denoted as... <k i ,HasRelatesOf,res j >, where knowledge point k iThe associated resource is res j .

[0068] Based on the course knowledge graph, and considering the learners' cognitive mastery of the knowledge points in the current course, a cognitive difference subgraph is generated. This leads to the identification of the most suitable cognitive regions for student learning within the course knowledge graph, such as... Figure 3 The diagram shown is a schematic representation of cognitive differences among learners of a specific target group.

[0069] As a preferred embodiment of the present invention, the steps for generating a cognitive difference subgraph based on cognitive mastery include: obtaining the learner's current stage of cognitive development knowledge cluster based on cognitive mastery, the teaching organization layer, and the knowledge representation layer; and generating a cognitive difference subgraph based on the current stage of cognitive development knowledge cluster and the dependencies between entities in the course knowledge graph. That is, according to inference rules, based on the cognitive development knowledge cluster, through steps such as full dependency subgraph inference and subgraph pruning, the required prior knowledge is inferred, and gaps in knowledge points not mastered in the previous stage are filled to generate the target learner's cognitive difference subgraph.

[0070] Step S102: Extract knowledge points and cognitive dependencies between knowledge points from the cognitive difference subgraph, and generate an executable matrix that includes all executable learning paths.

[0071] In other words, by extracting knowledge points and cognitive dependencies between knowledge points from the cognitive difference subgraph, a set of knowledge points and a set of dependencies are obtained. Based on the executable matrix generation algorithm, the knowledge points are mathematically modeled at the teaching organization layer and the knowledge representation layer. Through matrix operations and Boolean operations, all path information that does not violate cognitive laws is inferred, thereby avoiding all learning paths that violate the teaching order, and thus generating an executable matrix of executable learning paths.

[0072] Specifically, the steps for generating the executable matrix include:

[0073] Step S201: Extract knowledge points and cognitive dependencies between knowledge points from the cognitive difference subgraph based on extraction rules, and obtain the corresponding set of knowledge points and set of dependencies.

[0074] It should be noted that the extraction rules include one or more of the first rule, the second rule, the third rule, and the fourth rule.

[0075] First rule: For knowledge entity K x K y If a priori-successor relationship exists <K x ,Requires,K y Parent-son relationship <K x SubClassof,K yWhole-part relationship <K x PartOf,K y If any one or more relations are present in the >, then the dependency relation is... <K x ,R,K y >Established;

[0076] Second rule: For knowledge point entity K x K y If there is a sibling relationship <K x BrotherClass,K y >Relationships <K x RelatedOf,K y If any one or more relations are present in the >, then the dependency relation is... <K x ,R,K y >、 <K y ,R,K x >Established;

[0077] Third rule: For knowledge cluster G m Knowledge Cluster G n and knowledge point K x If a hierarchical relationship exists <G m ,PartGsOf,G n > Inclusion relationship <K x ,BelongTo,G n Any one or more relations in >, then the dependency relation <K x ,BelongTo,G m >Established;

[0078] Fourth rule: For knowledge cluster G m and knowledge cluster G n If there is an order relationship <G m NextGsOf,G n > For all Dependency <K x ,R,K y Established.

[0079] As a specific embodiment of the present invention, such as Figure 3 The image shows a cognitive difference subgraph for a specific learner. Knowledge points are extracted from this subgraph, resulting in a set of knowledge points: Set(K) = {K1, K2, K3, K4, K5, K6, K7}. Cognitive dependencies between these knowledge points are then extracted and denoted as relation triples. <K x ,R,K y>, thus obtaining the dependency set set(R), as shown in Table 1.

[0080] Table 1 Extraction of Dependency Sets

[0081]

[0082] Step S202: Generate a direct dependency matrix based on the knowledge point set and the dependency relationship set.

[0083] As an embodiment of the present invention, the generation of the direct dependency matrix includes: determining the dimension of the direct dependency matrix A as m×m based on the number of elements m of the knowledge point set set(K); and determining the value of the direct dependency matrix A based on the dependency relationship set set(R).

[0084] It should be noted that if it exists <K x ,R,K y If the x][y] value is greater than or equal to 1, then A[x][y] is 1; otherwise, it is 0. This is used to generate the corresponding direct dependency matrix A.

[0085] For example, according to Figure 3 The obtained knowledge point set set(K) = {K1, K2, K3, K4, K5, K6, K7} has 7 elements, and the dimension of the direct dependency matrix A is determined to be 7×7. Based on the dependency relationship set set(R) between knowledge points, as shown in Table 1, the values ​​of the corresponding original elements in the direct dependency matrix A are determined; where, for element A[x][y], if there exists... <K x ,R,K y If the x-axis is greater than or equal to y-axis, then A[x][y] is 1; otherwise, it is 0. Based on this, the corresponding direct dependency matrix A is generated as follows:

[0086]

[0087] Step S203: Add the direct dependency matrix A to the identity matrix I of the same dimension n×n, and perform n+1 Boolean power operations to obtain the fully dependent matrix M; where n is a positive integer, starting from 1.

[0088] Additionally, if there exists an n that satisfies

[0089] Then the total dependency matrix M is (A+I). n In the total dependency matrix M, excluding the diagonal elements, for each non-zero element A[x][y], its x-coordinate is K. x All of these are the knowledge points K corresponding to their vertical axes. y The knowledge points that are directly or indirectly dependent on.

[0090] It should be noted that for all data in the full dependency matrix M where M[x][y] = 1, if M[x][y] = 1 and M[y][x] ≠ 1, then it represents knowledge point K. x It is knowledge point K y The knowledge points of dependency relationships, therefore, for learning path K y →K x This does not conform to the cognitive laws of learning, and such paths must be avoided when planning learning paths; if M[x][y]=1 and M[y][x]=1, then it means that for knowledge point K y and knowledge point K x There is a closed-loop path, with no order of learning, so there is no need to avoid it.

[0091] For example, according to Figure 3 The cognitive difference subgraph can be calculated to obtain n=2. The detailed calculation process of the total dependence matrix M is as follows: Figure 4 As shown.

[0092] Step S204: Generate an executable matrix based on the full dependency matrix, thereby obtaining all executable paths that do not violate cognitive laws.

[0093] As an embodiment of the present invention, generating an executable matrix S based on a full dependency matrix M includes: using an n×n matrix of the same dimension as the full dependency matrix M as an initial executable matrix; obtaining the full dependency matrix M, and based on the formula... Calculate the value of each element S[x][y] in the initial executable matrix to obtain the executable matrix S. It should be noted that in the executable matrix S, all paths with a value of 1 represent executable paths; all paths with a value of 0 represent taboo paths, indicating that the corresponding path violates cognitive rules and cannot be passed, thereby achieving the purpose of avoiding the path.

[0094] For example, using a 7×7 matrix of the same dimension as the full dependency matrix M as the initial executable matrix, and calculating the value of each element S[x][y] in the initial executable matrix, the executable matrix S is obtained as follows:

[0095]

[0096] Step S103: Obtain the path transfer value of the target learner on each learning path, and generate a path transfer matrix that includes all executable learning paths.

[0097] That is, the path transfer matrix is ​​obtained based on the learning path data of the reference learners.

[0098] As an embodiment of the present invention, the learning path data and score set of the reference learning group in the reference group are obtained to obtain the path transfer value of the target learner in each learning path.

[0099] Among them, the path migration value V a The methods for obtaining [x][y] include:

[0100]

[0101] Among them, V a [x][y] represents the transfer value of learner a along the learning path from knowledge point x to knowledge point y; sim ai The similarity between target learner a and reference learner i; This indicates that the set of learning paths included in the reference learning group contains the learning path x→y; score i [x][y] represents the learning value of reference learner i on the learning path x→y; h represents the number of reference learners i with the learning path x→y; This represents the mean of the transfer values ​​across all learning paths.

[0102] Furthermore, reference learners are learners who have already completed learning all the knowledge points in the cognitive difference submap. The reference group is constructed based on the similarity of learning data between the target learner and the reference learners. For example, the similarity score is calculated using the similarity score... ab The values ​​are sorted from high to low, and the learners ranked before the set threshold are added to the reference group, thus constructing the reference group.

[0103] As an embodiment of the present invention, similarity sim is calculated based on learner data. ab More specifically, the calculation methods include: sim ab =λ1B ab +λ2F ab +λ3C ab Where a represents the target learner; b represents the reference learner; B ab F represents the normalized similarity between the target learner and the reference learner in terms of basic information dimensions such as age and education level; ab C represents the normalized similarity between the target learner and the reference learner across feature dimensions; ab λ1 represents the normalized similarity of the cognitive dimensions between the target learner and the reference learner; λ2, λ3 represent the weights of the similarity in the three dimensions.

[0104] Step S104: Based on the executable matrix and the path transition matrix, generate a path benefit matrix that includes all executable learning paths, and obtain the learning benefit value of each executable learning path.

[0105] Furthermore, the path revenue matrix η a The generation methods include: using the path migration matrix V a Calculate the Hadamard product between the executable matrix S and the executable matrix S. The formula for calculation is: ηa =V a ·S.

[0106] Based on the learning path data of the reference learners, a path transfer matrix is ​​obtained. Then, based on the path transfer matrix and the executable matrix, a path reward matrix for the target learner is obtained. An improved ant colony path planning algorithm that incorporates learning reward calculation is used. That is, the ant colony algorithm is used to use the path reward matrix as the learning reward value for each path, and the learning path with the highest learning reward value is identified as the optimal learning path for the target learner and recommended to the learner.

[0107] As an embodiment of the present invention, the improved ant colony path planning algorithm utilizing the introduction of a path reward value matrix includes:

[0108] Step 301: Use the path reward value matrix as the heuristic information matrix for the ant colony algorithm; set a virtual node as the initial node of the ant colony algorithm, where the virtual node represents the cognitive mastery of the target learner at the current stage.

[0109] It should be noted that since the number of knowledge points in the target learner's current cognitive difference subgraph is n, both the target learner's path transfer matrix and initial path benefit matrix are n×n two-dimensional matrices. For example, according to... Figure 3 The difference subgraph shown has 7 knowledge points. Therefore, the path transfer matrix and the initial path benefit value matrix of the target learner a are both 7×7 two-dimensional matrices.

[0110] It should be noted that for the shortest path problem in existing technologies, traditional ant colony optimization (ACO) algorithms assume that each node is homogeneous, and the selection of the initial node is completely random, without affecting the actual path planning effect. However, in learning path planning in the field of education, the selection of the first knowledge point is not completely random. Because the target learner has their own level of knowledge and cognition, choosing different knowledge points in the cognitive difference subgraph will produce different learning benefits. Therefore, this invention sets a virtual node representing the target learner's current level of cognitive mastery as the initial node of the ACO algorithm, thereby more realistically simulating the target learner's online learning process and generating a learning path that better matches the target learner's actual cognitive process.

[0111] Step 302: Perform pathfinding iterations. In each iteration, each ant finds its own path based on the pheromone concentration and the learning reward value of each executable learning path.

[0112] In this case, if the ant is at knowledge node K at a certain moment during the pathfinding process... i At that time, the ant had not learned the set of knowledge points K. jset, and there is no knowledge node K in the executable matrix. i →Knowledge Section K j and / or no knowledge section K exists i →Knowledge Point Set K j If `set` is used, the executable path will be discarded, and no information concentration update or learning gain calculation will be performed.

[0113] Step 303: In a single iteration, calculate the learning reward value for all ants that have completed the current iteration.

[0114] Furthermore, the learning benefit value Path_val a The calculation methods include:

[0115]

[0116] Wherein, Path_val a This represents the learning gain value obtained by learner a after completing the learning of all knowledge points in the cognitive difference subgraph; vis_kg a This represents the learning sequence of knowledge points in this iteration of the ant; vis_kg a [x] represents the knowledge point at position x; vis_kg a [x+1] represents the next knowledge point at position x; η[vis_kg] a [x]][vis_kg a [x+1] represents the knowledge point vis_kg a [x] to the knowledge point vis_kg a The learning efficiency of [x+1].

[0117] Step S105: Obtain the adaptive learning path for the target learner based on the learning gains.

[0118] Step 304: Select the executable learning path with the highest learning benefit value as the optimal learning path for the target learner.

[0119] Preferably, when the maximum number of iterations or the optimal learning path exceeding the set number of iterations is reached, the learning gain value will no longer change, and the iteration will stop. Then, the learning path with the highest learning gain value among all current iterations is selected as the optimal learning path for the target learner and recommended to the learner.

[0120] On the other hand, the present invention also provides an adaptive learning path planning system based on knowledge graphs, which includes at least a course knowledge graph construction module, a learning process data storage module, a learner dynamic cognitive management module, an adaptive learning path planning module, and an adaptive learning path recommendation module, for performing the steps of any of the above methods.

[0121] The module includes: a course knowledge graph construction module for building and storing the course knowledge graph for the course domain; a learning process data storage module for storing the learning process data of individual learners in online learning; a learner dynamic cognition management module for managing, updating, and storing the dynamic cognitive development of learners based on the learning process data; an adaptive learning path planning module for generating adaptive knowledge learning paths based on the course knowledge graph, learning process data, and learner dynamic cognition; and an adaptive learning path recommendation module for recommending adaptive online resource learning paths for target learners based on the resource and knowledge associations in the course knowledge graph and learner characteristics.

[0122] Furthermore, the adaptive learning path planning system also includes basic modules such as a login module and a registration module.

[0123] As a specific embodiment of the present invention, such as Figure 5 As shown, during system operation, the target learner first logs into the system module. Based on the target learner information, the system retrieves the target learner's dynamic cognitive characteristic data from the learner dynamic management module, retrieves and obtains the cognitive difference subgraph that matches the target learner's developmental zone from the course knowledge graph module, and then generates an executable matrix for the learner through the knowledge and dependency extraction and executable matrix generation functions in the adaptive learning path planning module. The system also retrieves reference group learners through learner data in the learner dynamic management module and the learning process storage module, and obtains the learning benefit value matrix of the target learner through the path learning benefit value matrix calculation module. By introducing the adaptive learning path planning algorithm for benefit value calculation, an adaptive knowledge learning path is generated. Furthermore, based on the association between online resources and knowledge points in the subject knowledge graph management module, and using the adaptive knowledge learning path as a basis, and the learner data in the learner dynamic cognitive management module and the learning process storage module as a foundation, an adaptive online resource learning path is generated for the target learner through a collaborative filtering algorithm and recommended to the learner. It should be noted that during this process, all learning process data of the target learner is stored in the learning process storage module, and the dynamic cognitive characteristics of the learner's dynamic management module are updated in real time based on the dynamic cognitive characteristics.

[0124] This system recommends adaptive online resource learning paths for learners based on the relationships between online resources and knowledge points in the course knowledge graph, and on learners' historical learning data and knowledge learning paths. This enables adaptive knowledge and resource learning path planning for online learning, constrains adaptive learning path planning in online learning by using the dependencies between knowledge points to improve the rationality of adaptive learning paths, and uses path learning benefit values ​​as the basis for path selection to enhance the adaptability and intelligence of adaptive learning paths.

[0125] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications 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.

Claims

1. An adaptive learning path planning method based on knowledge graphs, characterized in that, The method includes: Obtain the course knowledge graph and the target learners' cognitive mastery of the knowledge points in the current course, and generate a cognitive difference subgraph; Extract knowledge points and cognitive dependencies between knowledge points from the cognitive difference subgraph to generate an executable matrix that includes all executable learning paths; Obtain the path transfer value of the target learner for each learning path, and generate a path transfer matrix that includes all executable learning paths; Based on the executable matrix and the path migration matrix, a path benefit matrix including all executable learning paths is generated to obtain the learning benefit value of each executable learning path. Based on the learning gains, an adaptive learning path can be obtained for the target learner; The step of generating the executable matrix includes: Based on the extraction rules, knowledge points and cognitive dependencies between knowledge points are extracted from the cognitive difference subgraph, resulting in a set of knowledge points and a set of dependencies. A direct dependency matrix is ​​generated based on the set of knowledge points and the set of dependencies. Add the directly dependent matrix to the identity matrix of the same dimension, and then... The total dependency matrix is ​​obtained by performing a power-law Boolean operation; where, It is a positive integer; The executable matrix is ​​generated based on the full dependency matrix.

2. The adaptive learning path planning method based on knowledge graphs as described in claim 1, characterized in that, The extraction rules include: First rule: For knowledge entity K x K y If a priori-successor relationship exists <K x , Requires, K y Parent-son relationship <K x SubClassof, K y Whole-part relationship <K x PartOf, K y If any one or more relations are present in the >, then the dependency relation is... <K x , R, K y >Established; Second rule: For knowledge point entity K x K y If there is a sibling relationship <K x BrotherClass, K y >Relationships <K x RelatedOf, K y If any one or more relations are present in the >, then the dependency relation is... <K x , R, K y >、 <K y , R, K x >Established; Third rule: For knowledge cluster G m Knowledge Cluster G n and knowledge point K x If a hierarchical relationship exists <G m , PartGsOf, G n > Inclusion relationship <K x BelongTo, G n Any one or more relations in >, then the dependency relation <K x BelongTo, G m >Established; Fourth rule: For knowledge cluster G m and knowledge cluster G n If there is an order relationship <G m NextGsOf, G n > For all K x {K | <K , BelongTo , G m >}, K y {K | <K , BelongTo , G n >}, then the dependency relationship <K x , R, K y Established.

3. The adaptive learning path planning method based on knowledge graphs as described in claim 1, characterized in that, The generation of the full dependency matrix includes: The direct dependency matrix is ​​determined based on the number of elements in the knowledge point set. The dimension is used to determine the direct dependency matrix based on the set of dependencies. The value; The direct dependency matrix Compared with the same dimension identity matrix Add, proceed Boolean operations of exponentiation; where, It is a positive integer, starting from 1; If it exists Satisfy the formula Then the total dependency matrix That is .

4. The adaptive learning path planning method based on knowledge graphs as described in claim 1, characterized in that, Based on the full dependency matrix Generate the executable matrix include: The full dependency matrix A matrix of the same dimension is used as the initial executable matrix; Obtain the full dependency matrix Based on the formula Update the value of each element in the initial executable matrix to obtain the executable matrix. .

5. The adaptive learning path planning method based on knowledge graphs as described in claim 1, characterized in that, The path migration value The methods of obtaining it include: ； in, Indicates target learners The learning path is based on knowledge points. To knowledge points The migration value; For target learners With reference learners Similarity; This indicates that the set of learning paths included in the reference learning group contains learning paths. ; Indicates reference learner In the learning path The learning value; This indicates the existence of a learning path. Reference learners The number of; This represents the mean of the transfer values ​​across all learning paths.

6. The adaptive learning path planning method based on knowledge graphs as described in claim 1, characterized in that, The path revenue matrix The generation method includes: converting the path migration matrix With the executable matrix Find the Hadamard product.

7. The adaptive learning path planning method based on knowledge graphs as described in claim 1, characterized in that, The improved ant colony path planning algorithm that utilizes the path reward matrix includes: The path reward matrix is ​​used as the heuristic information matrix for the ant colony algorithm; a virtual node is set as the initial node of the ant colony algorithm, and the virtual node represents the cognitive mastery of the target learner at the current stage. The pathfinding process is iterative. In each iteration, each ant searches for a path based on the pheromone concentration and the learning reward value of each executable learning path. If an ant is at a knowledge node at any point during the pathfinding process... At that time, a set of knowledge points that ants had not yet learned. Furthermore, there are no knowledge nodes in the executable matrix. →Knowledge Section and / or no knowledge section exists → Knowledge Points If so, the executable learning path will be discarded, and no information concentration update or learning benefit calculation will be performed; In a single iteration, the learning reward value is calculated for all ants that have completed the current iteration; The executable learning path with the highest learning benefit value is selected as the optimal learning path for the target learner.

8. The adaptive learning path planning method based on knowledge graphs as described in claim 1 or 7, characterized in that, The learning benefit value The calculation methods include: ； in, Indicates target learners The learning benefit value obtained by completing the learning of all knowledge points in the cognitive difference subgraph once; This represents the sequence of knowledge points learned by the ant in this iteration; Indicates the position as Knowledge points; Indicates the position as The next knowledge point after the knowledge point; Representing knowledge points To knowledge points The learning benefits.

9. An adaptive learning path planning system based on knowledge graphs, characterized in that, The system includes at least a course knowledge graph construction module, a learning process data storage module, a learner dynamic cognitive management module, an adaptive learning path planning module, and an adaptive learning path recommendation module, for performing the steps of the method described in any one of claims 1 to 8.

Images