A differentiated educational knowledge graph construction method, system, medium and device

Abstract

The application discloses a kind of support differentiated educational knowledge graph construction method, comprising: S1: collection learner information and knowledge point information, constructs knowledge point knowledge graph;S2: the complex relationship between knowledge point and knowledge point in knowledge point knowledge graph is predicted and completed;S3: determine the center degree of each knowledge point in knowledge point knowledge graph;S4: according to the historical answer data of learner, the knowledge point acquisition probability of learner is calculated;S5: the importance of each knowledge point is calculated by the center degree of knowledge point and knowledge point acquisition probability value, and the minimum weight learning path is obtained according to the importance optimization user learning path, and the shortest path is used to construct the educational knowledge graph of user.The application judges the mastery of new knowledge point of learner and individualized optimal recommendation method of learning new knowledge point, improves the learning efficiency of learner.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent education, and in particular relates to a method, system, medium and device for constructing a differentiated education knowledge graph. BACKGROUND

[0002] With the continuous acceleration of the digitalization process in China, online education technology has also been more and more widely applied. At present, academic social networks and MOOCs and other education platforms have obtained a large amount of academic education resources. After the strict time and place restrictions are lifted, learners can obtain various needed learning resources at any time through various platforms. Therefore, in order to better improve the learning effect of learners, it is very necessary to construct a knowledge graph for each attribute of each learner and use it in an education recommendation system.

[0003] Most of the current related researches judge the degree of adaptation of learners and learning resources according to the cognitive level, learning style, learning tendency and other characteristics of learners, and sort and recommend learning resources according to the corresponding rules, but do not utilize the historical data of learners, the relationship between knowledge points and other information, and the generated learning path may have the disadvantages of violating the internal logic between knowledge points and being unable to match the changing knowledge level of learners.

[0004] A knowledge graph is a huge network knowledge base linked by entities, entity attributes and entity relationships. The set of knowledge points mastered by learners, target knowledge points and historical interactive learning sequences are collected, and various knowledge points and their attributes involved in each education resource are integrated and a knowledge graph is established, so as to better recommend learning content for different learners. The existing knowledge graph recommendation model randomly samples and selects a neighborhood, which may easily lead to inappropriate recommendation results and the learners cannot make good use of online learning resources. Therefore, it is very necessary to propose a differentiated education knowledge graph to better recommend learning resources for learners. SUMMARY

[0005] The purpose of the present application is to overcome the above-mentioned defects in the prior art, and to provide a differentiated education knowledge graph construction method. The present application first determines the mastery of new knowledge points by learners according to their own characteristics and other attributes, and then determines the optimal recommendation method for learning new knowledge points according to the previous problem-solving situation and knowledge point search situation of learners, thereby improving the learning efficiency of learners.

[0006] To achieve the above-mentioned purpose, the present application provides a differentiated education knowledge graph construction method, comprising:

[0007] S1: collecting learner information and knowledge point information, and constructing a knowledge point knowledge graph;

[0008] S2: predicting and completing the knowledge points and the complex relationships between the knowledge points in the knowledge point knowledge graph;

[0009] S3: determining the centrality of each knowledge point in the knowledge point knowledge graph;

[0010] S4: calculating the knowledge point acquisition probability of the learner according to the historical answer data of the learner;

[0011] S5: calculating the importance of each knowledge point by the centrality of the knowledge point and the knowledge point acquisition probability value, optimizing the user learning path according to the importance, obtaining the shortest learning path, and constructing the educational knowledge graph of the user according to the shortest path.

[0012] Further, the knowledge point knowledge graph in S1 uses a TransR model to predict and complete the knowledge point entities and the complex relationships between the knowledge point entities;

[0013] The TransR model is as follows:

[0014] First, the knowledge points and the relationships between the knowledge points are encoded and mapped to integer indexes to construct a knowledge graph triple (h, r, t), (h, r, t) representing the head knowledge point, the relationship, and the tail knowledge point, and the mapping relationship is h, t∈Z |E| and r∈Z |R| where |E| represents the number of knowledge points, |R| represents the number of relationships, and Z is an integer index set;

[0015] Then, the knowledge point vector the relationship vector and the projection matrix are defined and randomly initialized, where d e and d r represent the dimensions of the entity vector and the relationship vector, respectively;

[0016] For each relationship, the head knowledge point vector, the tail knowledge point vector, and the relationship vector are projected into the same relationship space using the projection matrix M r , and the parameters are optimized so that the projected head knowledge point vector plus the projected relationship vector is closer to the tail knowledge point vector in the same relationship space.

[0017] Further, the projection of the head knowledge point vector, the tail knowledge point vector, and the relationship vector into the same relationship space is represented as h r and t r , where h r = hM r and t r = tM rThe head knowledge point vector after projection plus the relationship vector after projection in the same relationship space should be closer to the tail knowledge point vector. The following formula is used to calculate its good and bad:

[0018] f r (h,r,t) = |h r +r-t r |

[0019] In order to make the real relationship score as high as possible and the non-existent relationship score as low as possible, and improve the accuracy of the model, the loss function of the model is set as:

[0020]

[0021] Where is the set of positive sample triplets, is the set of negative sample triplets, the positive sample triplet is the real knowledge point triplet from the knowledge point knowledge graph; the negative sample triplet is a false knowledge point triplet generated according to the real positive sample triplet; h' r , t' r represent the projection of the head knowledge point and the tail knowledge point in the relationship projection space; h r , r, t r represent the vector representation of the head knowledge point, the relationship and the tail knowledge point in the original space respectively; γ is a margin hyperparameter, used to measure the interval between positive samples and negative samples, to ensure that the model makes the positive sample score as high as possible and the negative sample score as low as possible;

[0022] Then use the Adam optimizer to minimize the optimization target and update the model parameters. In each iteration, a batch of positive and negative samples is randomly selected, the score is calculated, and the model parameters are updated according to the optimization target.

[0023] Further, the center degree of each knowledge point in the S3 is determined, that is, by defining and calculating the adaptability index Score(i,j), wherein i represents the center entity and j represents the neighbor entity, the adaptability index Score(i,j) includes the similarity measure Sim(i,j) and the structural hole measure Hole(i), wherein the similarity measure reflects the similarity between the center knowledge point i and the neighbor knowledge point j, and the structural hole measure reflects the structural relationship closeness between them, and the sum of the two is the value of the adaptability index. The calculation formula of Sim(i,j) is as follows:

[0024] Sim(i,j) = (|N(i)∩N(j)|+1) / |N(i)∪N(j)|

[0025] Wherein, N(i) and N(j) represent the number of adjacent nodes of knowledge point node i and knowledge point node j respectively, |N(i)∩N(j)| represents the number of common adjacent nodes of two entity nodes, |N(i)∪N(j)| represents the sum of adjacent nodes of node i and node j, and the plus one in (|N(i)∩N(j)|+1) is to avoid the case that the adaptability is zero when there is no common adjacent node between two entity nodes.

[0026] Further, the knowledge point learning probability in S4 is the judgment of the knowledge mastery of the learner by the time convolution knowledge tracking ATCKT model with fusion attention mechanism, and the learning probability value of the correct answer of the user to all unmastered knowledge points is predicted.

[0027] Further, the calculation formula of the importance in S5 is:

[0028] important=a*Score(i,j)+(1-a)*probability, wherein probability is the learning probability value in S4, and the parameter a in the formula is artificially given and adjusted according to the preference of the learner.

[0029] Further, in S5, the weighted shortest path from each mastered knowledge point to the target knowledge point in the personalized knowledge point directed graph G is calculated by using the Dijkstra algorithm, and the path with the minimum weight in all weighted shortest paths is selected to construct the educational knowledge graph of the user.

[0030] A differentiated educational knowledge graph construction system, comprising:

[0031] A data storage module for storing information of learners, a knowledge point set, educational resource information and attribute relationship information thereof;

[0032] A backend service layer responsible for processing data and requests passed from the front end and performing corresponding calculations and processing; should include: data acquisition function, mainly responsible for acquiring historical data such as user historical answering conditions; knowledge graph construction function, that is, constructing a perfect knowledge graph and using TransR to complete the knowledge graph, the relationship therebetween includes various knowledge points, concepts, associated relationships and the like of various disciplines, to support the depth understanding and recommendation of the system to knowledge; personalized recommendation function, integrating a personalized recommendation algorithm, using an ATCKT model to recommend suitable knowledge content for the learner according to the personalized characteristics and behavior data of the learner;

[0033] A front-end display layer, including a user login page, a data input interface, a knowledge graph display interface and a knowledge point viewing and answering page.

[0034] A computer storage medium is used to store computer instructions, when the computer instructions are executed by a processor, the above-mentioned support differentiated education knowledge graph construction method is completed.

[0035] An electronic device apparatus includes a memory, a processor, and computer instructions stored on the memory and running on the processor, when the computer instructions are run by the processor, the above-mentioned support differentiated education knowledge graph construction method is completed.

[0036] Compared with the prior art, the present application has the following advantages:

[0037] In the present application, the design of the recommendation model aims at the problem of relationship sparsity of the knowledge graph composed of knowledge points and the relationship between knowledge points, and the use of the TransR model to complete the knowledge graph can more accurately capture the semantic information of the relationship, predict the complex relationship between knowledge points, improve the accuracy of relationship reasoning and the generalization ability of the model, thereby improving the quality of the knowledge point knowledge graph and the completion effect, especially for the differences between learners, the use of the ATCKT model can better judge the knowledge mastery of learners, capture the learning characteristics and behavior patterns of each learner, and achieve higher prediction accuracy and personalized learning guidance.

[0038] Through testing and verification on multiple types and multiple source data sets, the accuracy, precision and AUC value of the present application are significantly improved compared with traditional models.

[0039] The present application helps to provide a personalized education recommendation system for learners, assists learners to better learn target knowledge points, provides personalized learning guidance, and improves the learning efficiency of learners. BRIEF DESCRIPTION OF DRAWINGS

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0041] Figure 1 The flow chart of the support differentiated education knowledge graph construction method is shown in Figure 1.

[0042] Figure 2 The knowledge graph of the knowledge points of the high number course is shown in Figure 2.

[0043] Figure 3 The design architecture diagram of the ATCKT model is shown in Figure 3.

[0044] Figure 4An architecture diagram of a residual block;

[0045] Figure 5 An architecture diagram of an educational knowledge graph construction system supporting learner differentiation. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are one of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0047] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present application are only used to explain the relative positional relationship, movement condition, etc. between components in a certain posture (as described in the drawings), and if the certain posture changes, the directional indications also change accordingly.

[0048] In addition, if the present application embodiments involve descriptions such as "first", "second", etc., the "first", "second", etc. are only for description purposes, and cannot be understood as indicating or implying the relative importance of the indicated technical features or implicitly indicating the number of the indicated technical features.

[0049] Embodiments of the present application provide a differentiated educational knowledge graph construction method, system, medium and device, specifically as follows:

[0050] A differentiated educational knowledge graph construction method for learners, comprising:

[0051] S1: Collecting learner information and knowledge point information, and constructing a knowledge point knowledge graph;

[0052] S2: Predicting and completing the knowledge points and the complex relationships between the knowledge points in the knowledge point knowledge graph;

[0053] S3: Determining the centrality of each knowledge point in the knowledge point knowledge graph;

[0054] S4: Calculating the knowledge point acquisition probability of the learner according to the historical answer data of the learner;

[0055] S5: Synthetically calculating the importance of each knowledge point through the centrality of the knowledge point and the knowledge point acquisition probability value, optimizing the user learning path according to the importance, obtaining the minimum weight learning path, and constructing the educational knowledge graph of the user with the shortest path.

[0056] The steps are as follows: Figure 1The following is a detailed description of each step:

[0057] Step S1: Constructing the knowledge point knowledge graph

[0058] The knowledge point knowledge graph is a graphical network representation of learning resources. It reflects the logical relationships and connections between different knowledge points, with user-learned and to-be-mastered knowledge points as nodes and relationships between knowledge points as edges. First, manually collect various knowledge points involved in textbooks, academic papers, course outlines, online education platforms, etc. and determine the attribute information of each knowledge point (definition of knowledge point, related concepts, related theorems or formulas, examples, etc.). Then analyze the relationships between the collected knowledge points and determine the relationships between the knowledge points, including prerequisite relationships, dependency relationships, similarity relationships, etc.

[0059] By constructing the knowledge point knowledge graph, students can better understand the relationships between knowledge points, discover the difficulty level of knowledge and the learning path, and provide support for personalized learning and intelligent recommendation. For example, the relationships between different knowledge points in the mathematics discipline are shown. The connections between nodes in the knowledge graph represent the relationships between these knowledge points, such as prerequisite relationships, dependency relationships, application relationships, etc. For example, linear equations and quadratic equations are the basis for equations and inequalities, systems of equations depend on linear equations and quadratic equations, and polynomial equations are derived from systems of equations, etc. Such a knowledge graph can help students better understand the structure and association of knowledge, helping them to more systematically learn and master the knowledge of the algebra discipline. Figure 2

[0060] Step S2: Predicting and completing the knowledge points and complex relationships between knowledge points in the knowledge point knowledge graph

[0061] In this embodiment, the TransR model is used to complete the knowledge graph, aiming to solve some problems in traditional knowledge graph representation learning methods, such as the diversity and complexity between entities and relationships. The solution is to fill in or predict missing knowledge points and relationships between knowledge points to improve the completeness and accuracy of the knowledge graph. The main idea is to project the relationships between knowledge points into different spaces to better capture the semantic relationships between knowledge points.

[0062] First, encode the knowledge points and relationships between knowledge points and map them to integer indices to construct a knowledge graph of relationships between knowledge points, i.e., knowledge graph triples (h, r, t), (h, r, t) represent the head knowledge point, the relationship, and the tail knowledge point, respectively, and the mapping relationship is h, t∈Z |E| and r∈Z |R| ​where |E| denotes the number of knowledge points, |R| denotes the number of relations, and Z is a set of integer indices. Then the knowledge point vectors are randomly initialized Relation vectors and projection matrix where d e and d r denote the dimensions of entity vectors and relation vectors, respectively.

[0063] For each relation, the head knowledge point vector, the tail knowledge point vector, and the relation vector are projected into the same relation space using the projection matrix M r , denoted as h r , t r , and r, respectively, where h r = hM r and t r = tM r . The projected head knowledge point vector plus the projected relation vector in the same relation space should be closer to the tail knowledge point vector. This can be measured by the following formula:

[0064] f r (h, r, t) = |j r +r-t r |

[0065] To make the score of real relations as high as possible and the score of non-existent relations as low as possible, and to improve the accuracy of the model, the loss function of the model is set as:

[0066]

[0067] where is the set of positive sample triplets, and is the set of negative sample triplets. The positive sample triplets are real knowledge point triplets from the knowledge graph. The negative sample triplets are false knowledge point triplets generated based on the real positive sample triplets. h' r and t' r denote the projections of the head knowledge point and the tail knowledge point in the relation projection space, and h r , r, t r denote the vector representations of the head knowledge point, the relation, and the tail knowledge point in the original space, respectively. γ is a margin hyperparameter that measures the gap between positive samples and negative samples, which can ensure that the model makes the score of positive samples as high as possible and the score of negative samples as low as possible, thereby improving the accuracy and performance of the model.

[0068] Then the Adam optimizer is used to minimize the optimization objective to update the model parameters. In each iteration, a batch of positive and negative samples is randomly selected, their scores are calculated, and the model parameters are updated according to the optimization objective.

[0069] The existing knowledge graph data and TransR model are used to train through the above steps, and then the trained model is used to predict and complete the relationship between knowledge points. Through the above steps, a TransR model is obtained to complete the knowledge graph of knowledge points. In this way, a more complete and accurate knowledge graph of knowledge points can be obtained, which helps to deeply understand the relationship and structure between knowledge.

[0070] Step S3: Calculate the centrality of each knowledge point in the knowledge graph of knowledge points

[0071] By analyzing the connection relationship between knowledge point nodes in the knowledge graph, especially the connection between the center knowledge point node and its neighbor knowledge point nodes, the centrality of each knowledge point is determined to improve the efficiency and quality of the recommendation results.

[0072] In order to determine the centrality of each knowledge point, an adaptive index Score(i,j) is proposed, where i represents the center knowledge point node and j represents its first-order neighbor knowledge point. This adaptive index is used to measure the importance of the neighbor knowledge point node to the center knowledge point node, that is, its influence on the center knowledge point. If a neighbor knowledge point node has a greater influence on the center knowledge point node, the value of its adaptive index will be higher, and vice versa.

[0073] The calculation of the adaptive index Score(i,j) uses a formula, which is as follows:

[0074]

[0075] In this formula, N(i) is the neighbor of node i, which includes two factors, similarity measure Sim(i,j) and structural hole measure Hole(i). Among them, the similarity measure is used to measure the similarity between the center knowledge point and the neighbor knowledge point, and the structural hole measure is used to measure the structural relationship between them. The product of these two factors is the value of the adaptive index, and the calculation formula of Sim(i,j) is as follows:

[0076] Sim(i,j) = (|N(i)∩N(j)|+1) / |N(i)∪N(j)|

[0077] In this formula, N(i) and N(j) represent the number of neighbor nodes of knowledge point node i and knowledge point node j respectively, |N(i)∩N(j)| represents the number of common neighbors of the two entity nodes, |N(i)∪N(j)| represents the sum of the neighbor nodes of node i and node j, and "+1" in (|N(i)∩N(j)|+1) is to avoid the case that the adaptability is zero when there is no common neighbor between the two entity nodes. Hole(i) represents the weight of the entity node, and the structural hole index is used to calculate the node weight, and the formula is as follows:

[0078] Hole(i) = (Σ(j = 1 N (d i -d j )) / (N-1)

[0079] Where d i represents the degree of knowledge point node i, and d j represents the degree of other knowledge point node j in the connected component. The larger the structural hole index, the more special or important the role of the entity knowledge point node in the graph structure. The knowledge point node with low structural hole index may mean that the knowledge point node plays a relatively small or relatively ordinary role in the graph structure, connects fewer other knowledge point nodes, is in the edge position, or links in a localized manner, and only plays a role in a specific field or topic.

[0080] Step S4: calculating the knowledge point learning probability of the learner according to the historical answer data of the learner

[0081] In the present application, the time convolution knowledge tracking (ATCKT) with fusion attention mechanism is used to judge the knowledge mastery of the learner. Compared with the traditional knowledge tracking method, ATCKT can better capture the learning characteristics and behavior patterns of students, realize higher prediction accuracy and personalized learning guidance. The following is a detailed description of the ATCKT model in the present system:

[0082] The specific process of the ATCKT model is shown in Figure 3 In the present system, the input of the ATCKT model is the historical answer sequence of the user, represented as a one-dimensional array X, where x i (q i , 0 / 1), q is the question, q i represents the question selected by the user at the i-th step, and the second item is correct / incorrect; the output of the model is the prediction result of the next answer action for each user.

[0083] First, input the historical answer sequence X and randomly initialize the question embedding matrix E K×Q)and each row vector e i i represents an embedding vector of a question q t , K represents the dimension of the vectors, and Q represents the number of questions. During the training of the model, the question embedding vectors e i will automatically encode the information of the questions through learning.

[0084] The unanswered question e t and the question e i answered by the user before are then linearly transformed as follows:

[0085] Q = e t W query

[0086] A = e i W answer , i ∈ (1, T)

[0087] where W query ∈ R( K×K ), W answer ∈ R( K×K ) are two transformation matrices used to transform e t and e i , and then perform dot product operations on Q and A, and use the softmax activation function to normalize to obtain the attention weight α. At the same time, the historical answer data H is spliced, and the behavior h i of the i-th step is represented as (q, h), q i is the encoding of the question answered at the i-th step, h i is whether the question answered at the i-th step is correct, and if it is correct, it is (1,...,1), and if it is incorrect, it is (0,...,0), and its length is K, and then the historical interaction data h i is weighted by the attention weight α i to represent the user's historical comprehensive performance P i .

[0088] The user's historical comprehensive performance P i is then input into a temporal convolutional network (TCN) to extract the knowledge state of the user at each time. Each convolutional layer of the TCN has a fixed size sliding window for performing convolution operations on the input sequence. The residual block structure in the model is shown in Figure 4 , which includes: causal dilated convolution, layer normalization to normalize all neurons in the layer, ReLU activation function, and Dropout layer to prevent overfitting.

[0089] The TCN extracts the knowledge state matrix S ∈ R N×K of the user at each time, where N represents the number of questions answered by the user. s t ∈ R K, represents the student's mastery of knowledge at that moment, and the student's knowledge state s t and the next question e t+1 to get y t+1 Then, y t+1 Using the sigmoid activation function, the probability of answering the next question correctly can be obtained. The model outputs the probability of each possible answer option, so the loss function L is defined as the cross-entropy loss function, a t is the true value, and p t is the predicted value, as shown in the following formula:

[0090]

[0091] The model parameters are updated using the backpropagation algorithm and the Adam optimizer to minimize the loss function.

[0092] The above model is used to comprehensively measure the student's individualized prior knowledge by training the user's historical data for practice and answering, that is, the model is used to predict the user's correct answer probability value for all unmastered knowledge points. Since each question only tests one knowledge point, the question answer probability value calculated by the model can be equivalent to the knowledge point student answer probability value (probability value) of the knowledge point related to the question. The larger the probability value of a knowledge point, the more sufficient the student's prior knowledge about the knowledge point, and the more likely it is that the student will learn and master the knowledge point in the near future.

[0093] Step S5: Calculate the importance of each knowledge point by integrating the centrality of the knowledge point and the knowledge point learning probability value, optimize the user's learning path according to the importance, obtain the shortest learning path, and construct the user's educational knowledge graph based on the shortest path.

[0094] The calculation method of the importance (important value) of each knowledge point in the knowledge point knowledge graph is shown in the following formula:

[0095] important=a*Score(i,j)+(1-a)*probability

[0096] Wherein the Score(i,j) value is obtained by the algorithm in step three, and the probability value is obtained by the ATCKT model. The parameter a in the formula is artificially assigned, and can be appropriately adjusted according to the learning preference of the student, for example, some students have limited energy and prefer to achieve the goal of learning knowledge points in a shorter period of time, so the system can appropriately reduce the a value of the student; or some students want to learn more complete knowledge points and more complete knowledge system, so the a value can be increased. Through experimental test, a is best when a is 0.5, and a is set to 0.5 in this paper.

[0097] After calculating the important value of the knowledge point v j , the important value is used as the weight of the directed edge <v i , v j > in the personalized knowledge point directed graph, and then the Dijkstra algorithm is used to calculate the weighted shortest path from each mastered knowledge point to the target knowledge point in the personalized knowledge point directed graph G, and the path with the minimum weight in all weighted shortest paths is selected to construct the knowledge graph of the user. The path recommended to the student in the learning process is important and easy for the learner to master, so that the learner can quickly learn and get feedback and complete the learning of the target knowledge according to the shortest path.

[0098] One embodiment of the present application also provides a differentiated education knowledge graph construction system, which comprises: through front-end and back-end interaction and data storage, user input data, calculation processing and result display functions are realized, and efficient and accurate education knowledge recommendation services are provided for users. The main architecture is shown in Figure 5 , which comprises:

[0099] A data storage module, the data storage layer can store various data required for system operation, and is used for storing the information of learners, knowledge point sets, education resource information and attribute relationship information, etc.

[0100] A back-end service layer, the back-end service layer is responsible for processing the data and requests transmitted by the front end, and performing corresponding calculation and processing; should include: data acquisition function, mainly responsible for acquiring historical data such as user historical answering conditions; knowledge graph construction function, that is, constructing a perfect knowledge graph and using TransR to complete the knowledge graph, the relationship therebetween includes various knowledge points, concepts, associated relationships and the like of various disciplines, to support the depth understanding and recommendation of the system to knowledge; personalized recommendation function, integrating a personalized recommendation algorithm, using the ATCKT model to recommend suitable knowledge content for the learner according to the personalized characteristics and behavior data of the learner.

[0101] Front-end display layer: the front-end display layer is the interface for users to interact with the system, and the main functions include user login page, data input interface, knowledge graph display interface and knowledge point viewing, answering page, etc., the front-end interface is simple and clear, easy to use, can attract users and provide good user experience; it should also support various forms of knowledge presentation, including text, pictures, videos, audio, etc., to meet the learning habits and needs of different learners.

[0102] One embodiment of the present application also provides a computer storage medium for storing computer instructions, which, when executed by a processor, complete the above-mentioned support differentiated educational knowledge graph construction method.

[0103] One embodiment of the present application also provides an electronic device apparatus, comprising a memory, a processor and computer instructions stored on the memory and running on the processor, which, when executed by the processor, complete the above-mentioned support differentiated educational knowledge graph construction method.

[0104] The above is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for constructing educational knowledge graphs that supports differentiation, characterized in that, include: S1: Collect learner information and knowledge point information to construct a knowledge point knowledge graph. The knowledge point knowledge graph uses the TransR model to predict and complete the complex relationships between knowledge point entities. The TransR model mentioned above is as follows: First, the knowledge points and the relationships between them are encoded and mapped to integer indices to construct a graph of relationships between knowledge points, i.e., knowledge graph triples. , These represent the head knowledge point, the relationship, and the tail knowledge point, respectively, and the mapping relationship is as follows: and ,in Indicates the number of knowledge points. Indicates the number of relations. A set of integer indices; Next, define and randomly initialize the knowledge point vector. relation vectors and projection matrix ,in and These represent the dimensions of the entity vector and the relation vector, respectively. For each relation, use the projection matrix. The head knowledge vector, tail knowledge vector, and relation vector are projected into the same relation space, and the parameters are optimized so that the projected head knowledge vector plus the projected relation vector is closer to the tail knowledge vector in the same relation space. S2: Predict and complete the knowledge points and complex relationships between them in the knowledge point knowledge graph; S3: Determine the centrality of each knowledge point in the knowledge graph; S4: Calculate the probability of a learner acquiring knowledge points based on the learner's historical response data; S5: The importance of each knowledge point is calculated by combining the centrality and the probability of acquiring the knowledge point. The user's learning path is optimized based on the importance to obtain the learning path with the minimum weight. The user's educational knowledge graph is then constructed using the learning path with the minimum weight.

2. The method for constructing a differentiated educational knowledge graph according to claim 1, characterized in that, The projection of the head knowledge point vector, tail knowledge point vector, and relation vector onto the same relation space is represented as follows: and ,in and The projected head knowledge vector plus the projected relation vector should be closer to the tail knowledge vector in the same relation space. Its quality can be calculated using the following formula: To maximize the score for true relationships and minimize the score for non-existent relationships, thereby improving model accuracy, the model's loss function is set as follows: in It is a set of positive sample triples. It is a set of negative sample triples, and positive sample triples are real knowledge point triples from the knowledge point knowledge graph; negative sample triples are fake knowledge point triples generated based on real positive sample triples. , This represents the projection of the head and tail knowledge points into the relation projection space. , , These represent the vector representations of the head knowledge points, relationships, and tail knowledge points in the original space, respectively. It is the margin hyperparameter, used to measure the margin between positive and negative samples, ensuring that the model makes the positive sample score as high as possible as the negative sample score; The Adam optimizer is then used to minimize the optimization objective and update the model parameters. In each iteration, a batch of positive and negative sample pairs are randomly selected, their scores are calculated, and the model parameters are updated according to the optimization objective.

3. The method for constructing a differentiated educational knowledge graph according to claim 1, characterized in that, In step S3, the centrality of each knowledge point is determined, that is, by defining and calculating the adaptability index. ,in Represents the central entity. The neighboring entity is represented by the fitness index. Includes similarity metrics and structural hole measurement The similarity metric reflects the central knowledge point. Knowledge points about neighbors The similarity between the structures is measured by the degree of similarity between them, while the structural void measure reflects the tightness of their structural relationship. The sum of the products of the two is the value of the fitness index. The calculation formula is as follows: and These represent knowledge point nodes. and knowledge point nodes The number of neighboring nodes, This represents the number of common neighbors between two entity nodes. Represents a node and nodes The sum of neighboring nodes, The addition of 1 in the formula is to avoid the case where the fitness is zero when two entity nodes have no common neighbors.

4. The method for constructing a differentiated educational knowledge graph according to claim 3, characterized in that, In S4, the knowledge point acquisition probability is determined by the Temporal Convolutional Knowledge Tracking (ATCKT) model, which uses an attention-integrated mechanism, to assess the learner's knowledge mastery and predict the probability of the user correctly answering questions corresponding to all unmastered knowledge points.

5. The method for constructing a differentiated educational knowledge graph according to claim 4, characterized in that, The formula for calculating importance in S5 is as follows: ,in The learned probability value in S4, and the parameters in the formula It is assigned manually and adjusted appropriately according to the learner's preferences.

6. The method for constructing a differentiated educational knowledge graph according to claim 5, characterized in that, In step S5, the weighted shortest path from each mastered knowledge point to the target knowledge point in the personalized knowledge point directed graph G is calculated using the Dijkstra algorithm. The path with the smallest weight among all weighted shortest paths is then selected to construct the user's educational knowledge graph.

7. A differentiated educational knowledge graph construction system, characterized in that, include: The data storage module is used to store learner information, knowledge point sets, educational resource information, and their attribute relationship information; The backend service layer is responsible for handling data and requests from the frontend and performing corresponding calculations and processing. It should include: data acquisition functionality, responsible for acquiring historical data including users' past answers; knowledge graph construction functionality, i.e., building a complete knowledge graph and using TransR for knowledge graph completion, with relationships including knowledge points, concepts, and connections across various disciplines to support the system's deep understanding and recommendation of knowledge; and personalized recommendation functionality, integrating personalized recommendation algorithms and using the ATCKT model to recommend suitable knowledge content based on learners' personalized characteristics and behavioral data. Specifically, the knowledge point knowledge graph uses the TransR model to predict and complete the complex relationships between knowledge point entities. The TransR model mentioned above is as follows: First, the knowledge points and the relationships between them are encoded and mapped to integer indices to construct a graph of relationships between knowledge points, i.e., knowledge graph triples. , These represent the head knowledge point, the relationship, and the tail knowledge point, respectively, and the mapping relationship is as follows: and ,in Indicates the number of knowledge points. Indicates the number of relations. A set of integer indices; Next, define and randomly initialize the knowledge point vector. relation vectors and projection matrix ,in and These represent the dimensions of the entity vector and the relation vector, respectively. For each relation, use the projection matrix. The head knowledge vector, tail knowledge vector, and relation vector are projected into the same relation space, and the parameters are optimized so that the projected head knowledge vector plus the projected relation vector is closer to the tail knowledge vector in the same relation space. The front-end presentation layer includes the user login page, data input interface, knowledge graph display interface, knowledge point viewing, and question answering page.

8. A computer storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete any one of the methods for constructing a differentiated educational knowledge graph according to claims 1-7.

9. An electronic device, characterized in that, It includes a memory, a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, complete the method for constructing a differentiated educational knowledge graph as described in any one of claims 1-7.

Images