CN114065040A - Individual learning path and learning resource recommendation method based on discipline knowledge graph - Google Patents
Individual learning path and learning resource recommendation method based on discipline knowledge graph Download PDFInfo
- Publication number
- CN114065040A CN114065040A CN202111368109.7A CN202111368109A CN114065040A CN 114065040 A CN114065040 A CN 114065040A CN 202111368109 A CN202111368109 A CN 202111368109A CN 114065040 A CN114065040 A CN 114065040A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- learner
- knowledge point
- points
- resource
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Abstract
The invention discloses a discipline knowledge graph-based personalized learning path and learning resource recommendation method, which comprises the following steps of: 1) constructing a discipline knowledge graph; 2) preliminarily obtaining the knowledge point mastering condition of the learner through matrix evolution; 3) the knowledge point mastering condition of the learner is further obtained by defining knowledge reasoning; 4) calculating the similarity between the learner and the resource; 5) calculating the similarity between the knowledge point mastering condition of the learner and the resource knowledge point; 6) weighting and calculating the similarity between the learner and the resource and the similarity between the mastery condition of the knowledge point of the learner and the knowledge point of the resource; 7) and generating an individualized learning path and a learning resource recommendation result. According to the method, the accuracy of recommendation is effectively improved by diagnosing the learning condition of the learner; meanwhile, in the aspects of learning condition diagnosis and resource recommendation, the accuracy of personalized recommendation is increased by defining multiple attributes for knowledge points, and the method has high extensibility and application value.
Description
Technical Field
The invention relates to the field of personalized learning recommendation, in particular to a discipline knowledge graph-based personalized learning path and learning resource recommendation method.
Background
The continuous development of information technology and modern education theory changes the learning mode of people, and the education informatization urges a large number of online learning platforms, so that the learning of people is not limited by factors such as fields, time and the like, and is more flexible and free. However, the explosive growth of network information data brings great convenience to learners, and simultaneously inevitably brings trouble to some problems, especially to the problems of information overload and information lost. People are faced with explosive data that makes them difficult to effectively discern, accept, process, and utilize, which leads to information overload problems; the information lost navigation means that when a user collects information in a complex network environment, the user is often attracted by some irrelevant problems, loses the information searching direction or forgets an initial learning target.
Knowledge-graphs are intended to describe in a structured form the concepts, entities, events and relationships between them of the objective world. The knowledge map can store the relation between the knowledge points and the knowledge points, enrich the semantic relation of the knowledge points, integrate learning resources and form a subject knowledge network. The knowledge map is applied to personalized learning recommendation, so that a learner can be helped to know the position of a knowledge point in a knowledge system in the learning process, a clear learning path is planned for the learner by combining the learning state of the learner, and reasonable learning resources are recommended for the learner according to the interest and the ability of the learner.
Search engines are a typical solution to the information overload problem. However, the conventional search engine can only search resources for keywords provided by the user, and cannot filter resources according to the needs of the user, that is, cannot meet personalized requirements. The advent of recommendation systems provides an effective solution for personalization. The recommendation system is a technology capable of providing effective information services for users, does not need the users to provide clear requirements, but can mine the personalized requirements of the users according to the behavior information of the users on articles, and actively provides the required information for the users through the behavior models of the users. Therefore, providing personalized learning recommendations for learners by recommendation systems in order to solve the information overload problem is a research hotspot of related researchers nowadays.
At present, research on relevant aspects of personalized learning recommendation mainly focuses on recommendation methods based on learner interest modeling, but the methods have certain defects. The method is mainly used for constructing a knowledge portrait of the learner on the basis of historical behaviors or basic information of the learner. In the aspect of recommending learning resources, resources similar to the preference of the learner are recommended, and the influence of factors such as mastery of the knowledge points by the learner or incidence relations among the knowledge points on the learner is ignored. Considering that the association relationship between knowledge points plays an important role in meeting the learning requirements of learners, meeting the characteristics of learning process from shallow to deep and from easy to difficult, and the like, it is necessary to consider the relationship between knowledge points and the knowledge mastering condition of learners in teaching resource recommendation. In addition, the collaborative filtering algorithm is used as an important method in personalized recommendation, and when the cold start problem and the data sparsity problem are involved, the personalized recommendation method taking historical behaviors as the center cannot be effectively solved.
Aiming at the knowledge lost problem of learners, researchers provide a resource recommendation method based on a learning path by constructing learning characteristics based on a statistical method or learning history. Although the method is fit with the learning path of the learner to a certain extent, the knowledge framework of the whole learning path cannot be completely displayed, and knowledge point-based guidance is provided for the learner.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a discipline knowledge graph-based personalized learning path and learning resource recommendation method so as to deeply mine the knowledge mastering condition and knowledge point incidence relation of learners by utilizing matrix reasoning and hybrid recommendation and provide clear learning path recommendation and reasonable learning resource recommendation for learners.
The purpose of the invention is realized as follows: a disciplinary knowledge graph-based personalized learning path and learning resource recommendation method comprises the following steps:
step 1) establishing a discipline knowledge graph: defining a knowledge point structure according to the incidence relation between the subject knowledge points, representing the subject knowledge points by using a domain ontology and describing by using an ontology language to construct a subject knowledge graph;
step 2) preliminarily obtaining the knowledge point mastering condition of the learner through a matrix evolution model: testing the learner through the test questions to obtain a learner-knowledge point matrix and a test question-knowledge point matrix, and preliminarily obtaining knowledge point mastering conditions of the learner through evolution, wherein the knowledge point mastering conditions comprise mastered knowledge points and mastered knowledge points which are not mastered;
and 3) further obtaining the knowledge point mastering condition of the learner by defining knowledge reasoning: for the knowledge points which are judged to be under the mastery condition in the step 2), traversing the knowledge graph by using a knowledge inference technology according to the incidence relation between the knowledge points and the knowledge points which are judged to be under the mastery condition through a certain inference rule to further obtain the mastery condition of the knowledge points of the learner;
step 4) calculating the similarity between the learner and the resource: extracting key words by using a TextRank algorithm based on resource text data, performing word vector representation on the key words by using BERT (best effort test), constructing a resource knowledge base, forming a learner knowledge base by extracting knowledge points contained in learner historical resources, and finally calculating the similarity of texts by using word vectors as the similarity Interest of user interests and resources;
step 5) calculating the similarity between the knowledge point mastering condition of the learner and the resource knowledge point: defining the shortest path distance between the knowledge points, giving a formula to express the association degree between the knowledge points, and calculating the similarity DOC of the knowledge point mastering condition of the learner and the resource knowledge point through the characteristics of the learner;
step 6) weighting calculation of similarity between a learner and the resource and similarity between the mastery condition of the learner knowledge point and the resource knowledge point: carrying out linear weighted fusion on the user Interest and resource similarity Interest obtained in the step 4) and the learner knowledge point mastering condition and the resource knowledge point similarity DOC obtained in the step 5) to obtain a resource recommendation degree Sim, and introducing a parameter alpha to balance the weight relationship between the learner Interest and the knowledge point mastering condition;
step 7) generating an individual learning path and a learning resource recommendation result: according to the knowledge point mastering conditions of the learner obtained in the step 3), generating a knowledge point path to be learned of the learner from the knowledge points which are not mastered in the knowledge points according to the sequence from the base to the step, and recommending Top-N results with the highest recommendation degree to the learner according to the resource recommendation degree Sim obtained in the step 6).
As a further limitation of the present invention, the step 1) specifically comprises:
step 1.1) defining a knowledge point structure: defining four relations among knowledge points, wherein the four relations comprise a precondition and, a precondition or, a postamble and a postamble, and the structure of the knowledge points is recorded as G ═ K, R, wherein the set K ═ { K ═ K1，k2，k3，…，knDenotes knowledge points included in a subject, n denotes the total number of knowledge points, and the set R ═ R1，2，r2，3，…，ri，j，…，rn-1，nI is more than or equal to 1 and less than j and less than or equal to n, representing the incidence relation among the knowledge points, wherein n represents the total number of the knowledge points, ri，jRepresenting a knowledge point kiAnd knowledge point kjIn relation to each other, specifically, ri，jCan be expressed as<ki&kj>，<ki|kj>，<ki&kj>Represents kiAnd k isjIs a precondition and a relationship, kjAnd k isiThe relationship between the rear position and the front position,<ki|kj>represents kiAnd k isjIs a precondition or a relationship, kjAnd k isiIs a post or relationship;
step 1.2) using a domain ontology to express the knowledge point structure defined in the step 1.1), and modeling the domain ontology by using three elements of concept, relation and entity, wherein the knowledge points are concept elements, the association relation among the knowledge points is a relation element, and a specific knowledge point k is an entity element; the method comprises the following steps of describing a field ontology by using a triple in an ontology language OWL, wherein the triple comprises a class, an attribute and an object, the class corresponds to a concept element, the attribute is divided into an associated attribute and a mastered attribute, the associated attribute corresponds to a relation element, the mastered attribute corresponds to a mark of an entity element in the field ontology, the object corresponds to the entity element, and the structure of the triple can be represented as follows:
and step 1.3) visualizing the discipline knowledge graph by applying a knowledge graph representation technology.
As a further limitation of the present invention, the step 2) specifically comprises:
step 2.1) extracting test questions from the question bank, testing the learner A, and using a set Q ═ Q1，q2，…，qtDenotes test questions, t denotes the number of the test questions, and the set K ═ K is used1，k2，…，kmThe test question contains knowledge points, m represents the number of the knowledge points in the test question, and a vector S ═ S is used1，si，…，smDenotes the knowledge state, where siRepresents learner A to knowledge point kiThe grasping condition of (1) is specifically taken as 0, 1 and-1 to represent unknown, grasping and not grasping, and the weight number of the difficulty degree is added for each test question;
step 2.2) Using the test question-knowledge Point matrix
To represent the corresponding relation between the test question set Q and the knowledge point set K in the step 2.1), wherein cijShow the question qiAnd knowledge point kjThe relation between, in particular, the valuesThe mark is 0, 1, -1 and-2 respectively, the four cases of unknown, mastered, not mastered and not included are marked, countiRepresentative test question qiThe number of the knowledge points which are not mastered in the method and the mastering conditions of the initialized knowledge points are unknown;
and 2.3) evolving the matrix C in the step 2.2) to preliminarily obtain the knowledge point mastering condition of the learner.
As a further limitation of the present invention, the step 2.3) specifically includes:
step 2.3.1) if the learner A answers a certain question, temporarily judging that all the knowledge points covered by the question are mastered, and using a knowledge point use set R ═ { R ═ covered by the question answered by the learner A1，r2，…，rpExpressing that the row vector of the associated knowledge point in the answer pair test question is updated to 1 in the matrix C, the column vector corresponding to the knowledge point in the set R is deleted, and all the count values containing the knowledge point in the matrix C and the knowledge state vector S in the step 2.1) are updated;
step 2.3.2) if the learner A answers a wrong question, temporarily judging that the knowledge points covered by the question are not mastered, updating the row vectors of the associated knowledge points in the wrong answer question to-1 in the matrix C, and updating all the count values of the knowledge points in the matrix C and the knowledge state vector S in the step 2.1);
step 2.3.3) if the count value of a certain test question is 1, the values of other marks except the undetermined mark are 1, and the test question is answered incorrectly, determining that the knowledge point corresponding to the test question at the moment is not mastered, updating the knowledge point marked as undetermined at the moment of the test question to-1, updating the mark corresponding to the test question containing the knowledge point after the test question sequence in the matrix C to-1, and updating the count value of the matrix C and the knowledge state vector S in the step 2.1);
step 2.3.4) if the count value of a certain test question is 1, but the test question contains the unmastered knowledge points, the mastering condition of the unjudged knowledge points in the test question cannot be judged;
step 2.3.5) for knowledge points other than the situation mentioned in step 2.3.4), for a knowledge point, the difficulty weighting number of the mark 1 is markedMeasured as total1The difficulty-weighted number of tag-1 is total-1The difficulty weighting number for tag 0 is total0 if
And is
Beta is a self-defined attribute, the recommendation method provides open self-defined selection for a user, the meaning of the recommendation method is self-defined knowledge point mastering rate to detect the self knowledge point mastering degree, the knowledge point is judged to be mastered, and if the knowledge point mastering rate is self-defined, the knowledge point is judged to be mastered
And is
Judging the knowledge point as not mastered, otherwise classifying the knowledge point as not-determinable;
step 2.3.6) for the knowledge points whose mastery condition can not be determined in the evolution, use the set O ═ { O ═ O1，o2，…，oqAnd (c) represents.
As a further limitation of the present invention, the step 3) specifically includes:
step 3.1) for the knowledge points which can not be judged to be mastered in the step 2), making an inference rule by using possible relations between the knowledge points and the knowledge points which are judged to be mastered, wherein the possible relations are as follows: if the knowledge point k is judged to be mastered, all the knowledge points and the antecedent knowledge points of the k are judged to be mastered, or at least one of the knowledge points of the k or the antecedent knowledge points is judged to be mastered; otherwise, all or the antecedent knowledge points of k are judged to be not mastered, and at least one of the antecedent knowledge points of k is judged to be not mastered;
step 3.2) taking the set O in the step 2.3.6) as a current knowledge point set, and extracting a knowledge point kiAccording to kiThe grasping condition of the post-positioned knowledge point and the inference rule of the step 3.1) are inferred by kiIf the knowledge can not be inferred, go to the next knowledgePoint until finding a knowledge point, denoted k, from which the result can be inferredaTurning to the step 3.3), if all the knowledge points are taken, and the result cannot be deduced, exiting the cycle;
step 3.3) updating the knowledge state vector S in said step 2.1) and deleting knowledge points k from said set O of step 3.1)aWill be the knowledge point kaPremise knowledge points and knowledge points in the set O are extracted and put into the set X ═ X1，x2，…，xbReturning to the step 3.2) to be used as a new set O for circulation, wherein b is the number of the extracted knowledge points;
and 3.4) after all the reasoning is finished, all the knowledge points which can not judge the mastering conditions are treated as the masterless knowledge points, and the knowledge state vector S in the step 2.1) is updated to form the final knowledge point mastering conditions of the learners.
As a further limitation of the present invention, the step 4) specifically includes:
step 4.1) performing word segmentation and word removal operation on the resource text data, extracting 25 knowledge point keywords by using TextRank as semantic features of resources, expressing the keywords as 50-dimensional word vectors by using BERT, and using Si＝{si1，si2，…，sinDenotes resource SiWherein i is a subscript and n is the number of word vectors; for each resource SiMaintaining a list of entities, referred to as a knowledge repository KsourceThen resource SiCan be expressed as: ksource(Si)＝{si1，si2，…，sin}；
Step 4.2) taking all knowledge point entity sets contained in the historical resources learned by the learner as knowledge point sets learned by the user according to the historical behaviors of the learner, and including the knowledge point entity sets into a knowledge base K (user) { K ═ KsourceS1，KsourceS2，…，KsourceSNAdding knowledge points contained in the new resources into a knowledge base of the learner every time the learner learns the new resources;
and 4.3) calculating text similarity by using the word vector as the similarity between the interest of the learner and the resource, wherein the similarity is defined as follows:
where user represents the target learner, K (user) represents the set of historical resources for this learner,
and
respectively representing the word vector representation of the resource knowledge base resource i and the resource j that the learner has learned.
As a further limitation of the present invention, the step 5) specifically includes:
step 5.1) defining knowledge compactness: for any knowledge point k in the discipline knowledge graph G1And k is2，k2To k1Is knowledge point k2Is divided by k2To k1Shortest path distance SP (k)2，k1) Defining the postknowledge points of the same prerequisite knowledge point, defining that they do not have precedence order, setting shortest path distance as 1, if knowledge point k is2And k is1If there is no path relation or the same, the shortest path distance is set to-1, and the shortest path distances are the shortest path distances that the knowledge graph can search, except for the special case. The knowledge compactness is expressed by the following formula:
step 5.2) extracting the knowledge points which are not mastered by the learner in the step 2.1) from the knowledge state vector S obtained in the step 2.1 to form a knowledge point set DK, and using a knowledge resource base K (S) to represent the set of all the knowledge points of the resource S, so that the knowledge closeness from the resource S to the set of the knowledge points which are not mastered by the user represents the similarity between the mastered condition of the knowledge points of the learner and the knowledge points of the resource, and the similarity is expressed by the following formula:
wherein k isiRepresents an arbitrary knowledge point in the knowledge resource library K (S), in (k)i) I represents the importance of the knowledge points, and the shortest path distance SP represents the association between the knowledge points.
As a further limitation of the present invention, the step 6) specifically includes:
step 6.1) in order to simultaneously consider the balance between the learning interest of the user and the knowledge level of the user when recommending resources, linear weighted fusion of the two types of similarity is introduced as a basis for recommending the resources, and the method can be specifically expressed by the following formula:
Sim(user，i，S)＝η·Interest(user，)+(1-η)·DOC(S，user) (5)，
wherein the parameter η is a self-defined parameter to balance the weighted relationship between learner interest and knowledge point mastery level.
As a further limitation of the present invention, the step 7) specifically includes:
step 7.1) extracting knowledge points with the value of-1 in the knowledge state vector S of the learner as knowledge points which are not mastered by the learner according to the step 2.1), and storing the knowledge points into a set Kw＝{k1，k2，…，kwW is the total number of knowledge points which are not mastered by the learner;
step 7.2) from the set KwWhere the pre-selected knowledge point does not exist in KwKnowledge points with or without prior knowledge points are stored in a set Rr＝{k1，k2，…，krIn the equation, r is the number of knowledge points selected in this loop, from the set KwDeleting the selected knowledge points;
step 7.3) repeat step 7.2) until set KwIf the sequence is empty, the sequence to be learned L ═ { R } is finally generated1，R2，…，Rl} andusing a visualization method to represent;
and 7.4) generating a resource recommendation sequence of Top-N according to the resource recommendation Sim generated in the step 6.1), and finally integrating into learning path recommendation to form synchronous recommendation of the learning path and the learning resource.
By adopting the technical scheme, compared with the prior art, the invention has the beneficial effects that: 1) the knowledge point learning method defines a close logical relationship for knowledge points, establishes a knowledge system network based on a disciplinary knowledge map, considers rich knowledge semantic information, and effectively helps learners to know the positions of the knowledge points in a knowledge system; 2) aiming at the diagnosis of the learning condition of the learner, the invention adopts a matrix evolution method, combines knowledge reasoning, deeply excavates the mastering condition of the learner on the knowledge point and gives deep learning feedback to the learner; 3) aiming at a learning resource recommendation method, the invention adopts a mixed recommendation method based on learner interest and learner knowledge point mastering conditions, considers learner learning defects, combines learner interest, recommends learning resources really needed by a learner, and aims at a cold start problem.
Drawings
Figure 1 is an overall block diagram of the present invention.
FIG. 2 is a conceptual diagram of a discipline knowledge graph constructed in accordance with the present invention.
FIG. 3 is an exemplary diagram of the learner knowledge point mastery and resource knowledge point similarity calculation.
FIG. 4 is a diagram illustrating an exemplary learning path generation method according to the present invention.
Detailed Description
The method for recommending the personalized learning path and the learning resource based on the discipline knowledge graph as shown in FIG. 1 comprises the following steps:
step 1) establishing a discipline knowledge graph: defining a knowledge point structure according to the incidence relation between the subject knowledge points, representing the subject knowledge points by using a domain ontology and describing by using an ontology language to construct a subject knowledge graph;
step 1.1) defining a knowledge point structure: according to the relation structure of the knowledge points proposed by the related scholars, the relation between the knowledge points is expressed into four types including a precondition and, a precondition or, a postamble and a postamble. The method specifically comprises the following steps: if the premise of grasping the knowledge point k is to grasp the knowledge point k at the same time1，k2，k3，…，knAnd n is more than or equal to 1, then the knowledge point k1，k2，k3，…，knEach knowledge point and the knowledge point k are both a precondition and a relation, and the knowledge point k1，k2，k3，…，knEach knowledge point in the knowledge list is in a postpositional and relation; if knowledge point k is grasped, it is assumed that at least knowledge point k1, k is grasped2，k3，…，knOne of them, then knowledge point k1，k2，k3，…，knEach knowledge point and the knowledge point k are preconditions or relations, and the knowledge point k are1，k2，k3，…，knEach knowledge point in the knowledge list is in a postposition or relationship; the knowledge point structure is denoted as G ═ (K, R), where the set K ═ K1，k2，k3，…，knDenotes knowledge points included in a subject, n denotes the total number of knowledge points, and the set R ═ R1，2，r2，3，…，ri，j，…，rn-1，nI is more than or equal to 1 and less than j and less than or equal to n, representing the incidence relation among the knowledge points, wherein n represents the total number of the knowledge points, ri，jRepresenting a knowledge point kiAnd knowledge point kjIn relation to each other, specifically, ri，jCan be expressed as<ki&kj>，<ki|kj>，＜ki&kj>Represents kiAnd k isjIs a precondition and a relationship, kjAnd k isiThe relationship between the rear position and the front position,<ki|kj>represents kiAnd k isjIs a precondition or a relationship, kjAnd k isiIs a post or relationship;
step 1.2) using a domain ontology to express the knowledge point structure defined in the step 1.1), and modeling the domain ontology by using three elements of concept, relation and entity, wherein the knowledge points are concept elements, the association relation among the knowledge points is a relation element, and a specific knowledge point k is an entity element; the method comprises the following steps of describing a field ontology by using a triple in an ontology language OWL, wherein the triple comprises a class, an attribute and an object, the class corresponds to a concept element, the attribute is divided into an associated attribute and a mastered attribute, the associated attribute corresponds to a relation element, the mastered attribute corresponds to a mark of an entity element in the field ontology, the object corresponds to the entity element, and the structure of the triple can be represented as follows:
step 1.3) visualizing the discipline knowledge graph by applying knowledge graph representation technology; the constructed knowledge-graph conceptual diagram is shown in fig. 2.
Step 2) preliminarily obtaining the knowledge point mastering condition of the learner through a matrix evolution model: testing the learner through the test questions to obtain a learner-knowledge point matrix and a test question-knowledge point matrix, and preliminarily obtaining knowledge point mastering conditions of the learner through evolution, wherein the knowledge point mastering conditions comprise mastered knowledge points and mastered knowledge points which are not mastered;
step 2.1) extracting test questions from the question bank, testing the learner A, and using a set Q ═ Q1，q2，…，qtDenotes test questions, t denotes the number of the test questions, and the set K ═ K is used1，k2，…，kmThe test question contains knowledge points, m represents the number of the knowledge points in the test question, and a vector S ═ S is used1，si，…，smDenotes the knowledge state, where siRepresents learner A to knowledge point kiThe grasping condition of (1) is specifically taken as 0, 1 and-1 to represent unknown, grasping and not grasping, and the weight number of the difficulty degree is added for each test question;
step 2.2) Using the test question-knowledge Point matrix
To represent the pairs of the test question set Q and the knowledge point set K in the step 2.1)Should be connected, wherein cijShow the question qiAnd knowledge point kjThe values of the relationship are 0, 1, -1 and-2, respectively, and the values refer to the four conditions of unknown mark, mastered mark and no mastered mark, countiRepresentative test question qiThe number of the knowledge points which are not mastered in the method and the mastering conditions of the initialized knowledge points are unknown;
step 2.3) evolving the matrix C in the step 2.2) to preliminarily obtain the knowledge point mastering condition of the learner;
step 2.3.1) if the learner A answers a certain question, temporarily judging that all the knowledge points covered by the question are mastered, and using a knowledge point use set R ═ { R ═ covered by the question answered by the learner A1，r2，…，rpExpressing that the row vector of the associated knowledge point in the answer pair test question is updated to 1 in the matrix C, the column vector corresponding to the knowledge point in the set R is deleted, and all the count values containing the knowledge point in the matrix C and the knowledge state vector S in the step 2.1) are updated;
step 2.3.2) if the learner A answers a wrong question, temporarily judging that the knowledge points covered by the question are not mastered, updating the row vectors of the associated knowledge points in the wrong answer question to-1 in the matrix C, and updating all the count values of the knowledge points in the matrix C and the knowledge state vector S in the step 2.1);
step 2.3.3) if the count value of a certain test question is 1, the values of other marks except the undetermined mark are 1, and the test question is answered incorrectly, determining that the knowledge point corresponding to the test question at the moment is not mastered, updating the knowledge point marked as undetermined at the moment of the test question to-1, updating the mark corresponding to the test question containing the knowledge point after the test question sequence in the matrix C to-1, updating the count value of the matrix C and the knowledge state vector S in the step 2.1);
step 2.3.4) if the count value of a certain test question is 1, but the test question contains the unmastered knowledge points, the mastering condition of the unjudged knowledge points in the test question cannot be judged;
step 2.3.5) for knowledge points other than the case mentioned in step 2.3.4), for a knowledge point, the number of weighted difficulty levels of 1 is markedIs total1The difficulty-weighted number of tag-1 is total-1The difficulty weighting number for tag 0 is total0 if
And is
Beta is a self-defined attribute, the recommendation method provides open self-defined selection for a user, the meaning of the recommendation method is self-defined knowledge point mastering rate to detect the self knowledge point mastering degree, the knowledge point is judged to be mastered, and if the knowledge point mastering rate is self-defined, the knowledge point is judged to be mastered
And is
Judging the knowledge point as not mastered, otherwise classifying the knowledge point as not-determinable;
step 2.3.6) for the knowledge points whose mastery condition can not be determined in the evolution, use the set O ═ { O ═ O1，o2，…，oqAnd (c) represents.
And 3) further obtaining the knowledge point mastering condition of the learner by defining knowledge reasoning: for the knowledge points which are judged to be under the mastery condition in the step 2), traversing the knowledge graph by using a knowledge inference technology according to the incidence relation between the knowledge points and the knowledge points which are judged to be under the mastery condition through a certain inference rule to further obtain the mastery condition of the knowledge points of the learner;
step 3.1) for the knowledge points which can not be judged to be mastered in the step 2), making an inference rule by using possible relations between the knowledge points and the knowledge points which are judged to be mastered, wherein the possible relations are as follows: if the knowledge point k is judged to be mastered, all the knowledge points and the antecedent knowledge points of the k are judged to be mastered, or at least one of the knowledge points of the k or the antecedent knowledge points is judged to be mastered; otherwise, all or the antecedent knowledge points of k are judged to be not mastered, and at least one of the antecedent knowledge points of k is judged to be not mastered;
step 3.2) taking the set O in the step 2.3.6) as a current knowledge point set, and extracting a knowledge point kiAccording to kiPostsetting the mastery condition of the knowledge point and the inference rule in the step 3.1), inferring the mastery condition of ki, if inference cannot be carried out, switching to the next knowledge point until finding a knowledge point capable of inferring a result, and marking the knowledge point as kaTurning to the step 3.3), if all the knowledge points are taken, and the result cannot be deduced, exiting the cycle;
step 3.3) update the knowledge state vector S in step 2.1) and remove knowledge points k from the set O of step 3.1)aWill be the knowledge point kaPremise knowledge points and knowledge points in the set O are extracted and put into the set X ═ X1，x2，…，xbReturning to the step 3.2) to be used as a new set O for circulation, wherein b is the number of the extracted knowledge points;
and 3.4) after all the reasoning is finished, all the knowledge points which can not judge the mastering conditions are treated as the masterless knowledge points, and the knowledge state vector S in the step 2.1) is updated to form the final knowledge point mastering conditions of the learner.
Step 4) calculating the similarity between the learner and the resource: extracting key words by using a TextRank algorithm based on resource text data, performing word vector representation on the key words by using BERT (best effort test), constructing a resource knowledge base, forming a learner knowledge base by extracting knowledge points contained in learner historical resources, and finally calculating the similarity of texts by using word vectors as the similarity Interest of user interests and resources;
step 4.1) performing word segmentation and word removal operation on the resource text data, extracting 25 knowledge point keywords by using TextRank as semantic features of resources, expressing the keywords as 50-dimensional word vectors by using BERT, and using Si＝{si1，si2，…，sinDenotes resource SiWherein i is a subscript and n is the number of word vectors; for each resource SiMaintaining a list of entities, referred to as a knowledge repository KsourceThen resource SiCan be expressed as: ksource(Si)＝{si1，si2，…，sin}；
Step 4.2) taking all knowledge point entity sets contained in the historical resources learned by the learner as knowledge point sets learned by the user according to the historical behaviors of the learner, and including the knowledge point entity sets into a knowledge base K (user) { K ═ KsourceS1，KsourceS2，…，KsourceSNAdding knowledge points contained in the new resources into a knowledge base of the learner every time the learner learns the new resources;
and 4.3) calculating text similarity by using the word vector as the similarity between the interest of the learner and the resource, wherein the similarity is defined as follows:
where user represents the target learner, K (user) represents the set of historical resources for this learner,
and
respectively representing the word vector representation of the resource knowledge base resource i and the resource j that the learner has learned.
Step 5), calculating and learning the similarity between the knowledge point grasping condition and the resource knowledge point: defining the shortest path distance between the knowledge points, giving a formula to express the association degree between the knowledge points, and calculating the similarity DOC of the knowledge point mastering condition of the learner and the resource knowledge point through the characteristics of the learner;
step 5.1) defining knowledge compactness: for any knowledge point k in the discipline knowledge graph G1And k is2，k2To k1Is knowledge point k2Is divided by k2To k1Shortest path distance SP (k)2，k1) Defining the postknowledge points of the same prerequisite knowledge point, defining that they do not have precedence order, setting shortest path distance as 1, if knowledge point k is2And k is1Non-existent roadAnd if the path relations are the same, setting the shortest path distances of the paths as-1, and removing the special condition, wherein the shortest path distances are the shortest path distances which can be searched by the knowledge graph. The knowledge compactness is expressed by the following formula:
step 5.2) extracting the knowledge points which are not mastered by the learner in the step 2.1) from the knowledge state vector S obtained in the step 2.1 to form a knowledge point set DK, and using a knowledge resource base K (S) to represent the set of all the knowledge points of the resource S, so that the knowledge closeness from the resource S to the set of the knowledge points which are not mastered by the user represents the similarity between the mastered condition of the knowledge points of the learner and the knowledge points of the resource, and the similarity is expressed by the following formula:
wherein k isiRepresents an arbitrary knowledge point in the knowledge resource library K (S), in (k)i) I represents the importance of the knowledge points, and the shortest path distance SP represents the association between the knowledge points. As shown in the example of fig. 3, K (user) represents the learner knowledge base, K (1), K (2), and K (3) represent the resource knowledge base, and for the knowledge base B, the shortest path distance of D is 1, the degree of income is 1, and DOC is 1; the shortest path distance of F is 1, the degree of income is 2, and DOC is 2; e has a shortest path distance of 2, an introspection of 2, and DOC of 1, so that the resource corresponding to the resource repository F is recommended to the learner for the repository B. It is worth mentioning that in calculating the degree of income, the weight of the relation and/or the relation can be considered to obtain a more refined calculation result.
Step 6) weighting calculation of similarity between a learner and the resource and similarity between the mastery condition of the learner knowledge point and the resource knowledge point: carrying out linear weighted fusion on the user Interest and resource similarity Interest obtained in the step 4) and the learner knowledge point mastering condition and the resource knowledge point similarity DOC obtained in the step 5) to obtain a resource recommendation degree Sim, and introducing a parameter alpha to balance the weight relationship between the learner Interest and the knowledge point mastering condition;
step 6.1) in order to simultaneously consider the balance between the learning interest of the user and the knowledge level of the user when recommending resources, linear weighted fusion of the two types of similarity is introduced as a basis for recommending the resources, and the method can be specifically expressed by the following formula:
Sim(user，i，S)＝η·Interest(user，)+(1-η)·DOC(S，user) (5)，
wherein the parameter η is a self-defined parameter to balance the weighted relationship between learner interest and knowledge point mastery level.
Step 7) generating an individual learning path and a learning resource recommendation result: according to the knowledge point mastering conditions of the learner obtained in the step 3), generating a knowledge point path to be learned of the learner from the knowledge points which are not mastered in the knowledge points according to the sequence from the base to the step, and recommending Top-N results with the highest recommendation degree to the learner according to the resource recommendation degree Sim obtained in the step 6).
Step 7.1) extracting knowledge points with the value of-1 in the knowledge state vector S of the learner as knowledge points which are not mastered by the learner according to the step 2.1), and storing the knowledge points into a set Kw＝{k1，k2，…，kwW is the total number of knowledge points which are not mastered by the learner;
step 7.2) from the set KwWhere the pre-selected knowledge point does not exist in KwKnowledge points with or without prior knowledge points are stored in a set Rr＝{k1，k2，…，krIn the equation, r is the number of knowledge points selected in this loop, from the set KwDeleting the selected knowledge points;
step 7.3) repeat step 7.2) until set KwIf the sequence is empty, the sequence to be learned L ═ { R } is finally generated1，R2，…，RlAnd expressed by using a visualization method; as shown in fig. 4, a learning path generation example, where blue symbols indicate knowledge points not grasped, arrows indicate a pre-condition class relationship, and knowledge points k4 and k6 are sequentially extracted in evolution; knowledge point k 5; knowledge points k2, k 13; knowledge point k 7; knowledge ofThe points k1, k10, ultimately generate the knowledge point learning path shown in fig. 4 (g).
And 7.4) generating a resource recommendation sequence of Top-N according to the resource recommendation Sim generated in the step 6.1), and finally integrating into learning path recommendation to form synchronous recommendation of the learning path and the learning resource.
Aiming at the problems presented by the traditional method, the invention takes more attention to the logic relation between knowledge points and the mastery condition of learners on the knowledge points in the personalized learning recommendation. The knowledge point learning method defines a close logical relationship for knowledge points, establishes a knowledge system network based on a disciplinary knowledge map, considers rich knowledge semantic information, and effectively helps learners to know the positions of the knowledge points in a knowledge system; aiming at the diagnosis of the learning condition of the learner, the invention adopts a matrix evolution method, combines knowledge reasoning, deeply excavates the mastering condition of the learner on the knowledge point and gives deep learning feedback to the learner; aiming at a learning resource recommendation method, the invention adopts a mixed recommendation method based on learner interest and learner knowledge point mastering conditions, considers learner learning defects, combines learner interest, recommends learning resources really needed by a learner, and aims at a cold start problem. Meanwhile, in the aspects of learning condition diagnosis and resource recommendation, the method can increase the accuracy of personalized recommendation by defining multiple attributes for knowledge points, and has high extensibility and application value.
The present invention is not limited to the above-mentioned embodiments, and based on the technical solutions disclosed in the present invention, those skilled in the art can make some substitutions and modifications to some technical features without creative efforts according to the disclosed technical contents, and these substitutions and modifications are all within the protection scope of the present invention.
Claims (9)
1. A disciplinary knowledge graph-based personalized learning path and learning resource recommendation method is characterized by comprising the following steps:
step 1) establishing a discipline knowledge graph: defining a knowledge point structure according to the incidence relation between the subject knowledge points, representing the subject knowledge points by using a domain ontology and describing by using an ontology language to construct a subject knowledge graph;
step 2) preliminarily obtaining the knowledge point mastering condition of the learner through a matrix evolution model: testing the learner through the test questions to obtain a learner-knowledge point matrix and a test question-knowledge point matrix, and preliminarily obtaining knowledge point mastering conditions of the learner through evolution, wherein the knowledge point mastering conditions comprise mastered knowledge points and mastered knowledge points which are not mastered;
and 3) further obtaining the knowledge point mastering condition of the learner by defining knowledge reasoning: for the knowledge points which are judged to be under the mastery condition in the step 2), traversing the knowledge graph by using a knowledge inference technology according to the incidence relation between the knowledge points and the knowledge points which are judged to be under the mastery condition through a certain inference rule to further obtain the mastery condition of the knowledge points of the learner;
step 4) calculating the similarity between the learner and the resource: extracting key words by using a TextRank algorithm based on resource text data, performing word vector representation on the key words by using BERT (best effort test), constructing a resource knowledge base, forming a learner knowledge base by extracting knowledge points contained in learner historical resources, and finally calculating the similarity of texts by using word vectors as the similarity Interest of user interests and resources;
step 5) calculating the similarity between the knowledge point mastering condition of the learner and the resource knowledge point: defining the shortest path distance between the knowledge points, giving a formula to express the association degree between the knowledge points, and calculating the similarity DOC of the knowledge point mastering condition of the learner and the resource knowledge point through the characteristics of the learner;
step 6) weighting calculation of similarity between a learner and the resource and similarity between the mastery condition of the learner knowledge point and the resource knowledge point: carrying out linear weighted fusion on the user Interest and resource similarity Interest obtained in the step 4) and the learner knowledge point mastering condition and the resource knowledge point similarity DOC obtained in the step 5) to obtain a resource recommendation degree Sim, and introducing a parameter alpha to balance the weight relationship between the learner Interest and the knowledge point mastering condition;
step 7) generating an individual learning path and a learning resource recommendation result: according to the knowledge point mastering conditions of the learner obtained in the step 3), generating a knowledge point path to be learned of the learner from the knowledge points which are not mastered in the knowledge points according to the sequence from the base to the step, and recommending Top-N results with the highest recommendation degree to the learner according to the resource recommendation degree Sim obtained in the step 6).
2. The method for recommending personalized learning path and learning resource based on discipline knowledge graph according to claim 1, wherein the step 1) specifically comprises:
step 1.1) defining a knowledge point structure: defining four relations among knowledge points, wherein the four relations comprise a precondition and, a precondition or, a postamble and a postamble, and the structure of the knowledge points is recorded as G ═ K, R, wherein the set K ═ { K ═ K1，k2，k3，…，knDenotes knowledge points included in a subject, n denotes the total number of knowledge points, and the set R ═ R1，2，r2，3，…，ri，j，…，rn-1，nI is more than or equal to 1 and less than j and less than or equal to n, representing the incidence relation among the knowledge points, wherein n represents the total number of the knowledge points, ri，jRepresenting a knowledge point kiAnd knowledge point kjIn relation to each other, specifically, ri，jCan be expressed as < ki&kj＞，＜ki|kj＞，＜ki&kjIs > represents kiAnd k isjIs a precondition and a relationship, kjAnd k isiIs a postAND relationship, < ki|kjIs > represents kiAnd k isjIs a precondition or a relationship, kjAnd k isiIs a post or relationship;
step 1.2) using a domain ontology to express the knowledge point structure defined in the step 1.1), and modeling the domain ontology by using three elements of concept, relation and entity, wherein the knowledge points are concept elements, the association relation among the knowledge points is a relation element, and a specific knowledge point k is an entity element; the method comprises the following steps of describing a field ontology by using a triple in an ontology language OWL, wherein the triple comprises a class, an attribute and an object, the class corresponds to a concept element, the attribute is divided into an associated attribute and a mastered attribute, the associated attribute corresponds to a relation element, the mastered attribute corresponds to a mark of an entity element in the field ontology, the object corresponds to the entity element, and the structure of the triple can be represented as follows:
and step 1.3) visualizing the discipline knowledge graph by applying a knowledge graph representation technology.
3. The method for recommending personalized learning path and learning resource based on discipline knowledge graph according to claim 1, wherein the step 2) specifically comprises:
step 2.1) extracting test questions from the question bank, testing the learner A, and using a set Q ═ Q1，q2，…，qtDenotes test questions, t denotes the number of the test questions, and the set K ═ K is used1，k2，…，kmThe test question contains knowledge points, m represents the number of the knowledge points in the test question, and a vector S ═ S is used1，si，…，smDenotes the knowledge state, where siRepresents learner A to knowledge point kiThe grasping condition of (1) is specifically taken as 0, 1 and-1 to represent unknown, grasping and not grasping, and the weight number of the difficulty degree is added for each test question;
step 2.2) Using the test question-knowledge Point matrix
To represent the corresponding relation between the test question set Q and the knowledge point set K in the step 2.1), wherein cijShow the question qiAnd knowledge point kjThe values of the relationship are 0, 1, -1 and-2, respectively, and the values refer to the four conditions of unknown mark, mastered mark and no mastered mark, countiRepresentative test question qiThe number of the knowledge points not mastered in the middle, initializing the knowledge point mastering conditionsAre all unknown;
and 2.3) evolving the matrix C in the step 2.2) to preliminarily obtain the knowledge point mastering condition of the learner.
4. The discipline knowledge graph-based personalized learning path and learning resource recommendation method according to claim 3, wherein the step 2.3) specifically comprises:
step 2.3.1) if the learner A answers a certain question, temporarily judging that all the knowledge points covered by the question are mastered, and using a knowledge point use set R ═ { R ═ covered by the question answered by the learner A1，r2，…，rpExpressing that the row vector of the associated knowledge point in the answer pair test question is updated to 1 in the matrix C, the column vector corresponding to the knowledge point in the set R is deleted, and all the count values containing the knowledge point in the matrix C and the knowledge state vector S in the step 2.1) are updated;
step 2.3.2) if the learner A answers a wrong question, temporarily judging that the knowledge points covered by the question are not mastered, updating the row vectors of the associated knowledge points in the wrong answer question to-1 in the matrix C, and updating all the count values of the knowledge points in the matrix C and the knowledge state vector S in the step 2.1);
step 2.3.3) if the count value of a certain test question is 1, the values of other marks except the undetermined mark are 1, and the test question is answered incorrectly, determining that the knowledge point corresponding to the test question at the moment is not mastered, updating the knowledge point marked as undetermined at the moment of the test question to-1, updating the mark corresponding to the test question containing the knowledge point after the test question sequence in the matrix C to-1, and updating the count value of the matrix C and the knowledge state vector S in the step 2.1);
step 2.3.4) if the count value of a certain test question is 1, but the test question contains the unmastered knowledge points, the mastering condition of the unjudged knowledge points in the test question cannot be judged;
step 2.3.5) for knowledge points other than the situation mentioned in step 2.3.4), for a certain knowledge point, the weighted number of the difficulty degrees marked 1 is total1The difficulty-weighted number of tag-1 is total-1The difficulty weighting number for tag 0 is total0 if
And is
Beta is a self-defined attribute, the recommendation method provides open self-defined selection for a user, the meaning of the recommendation method is self-defined knowledge point mastering rate to detect the self knowledge point mastering degree, the knowledge point is judged to be mastered, and if the knowledge point mastering rate is self-defined, the knowledge point is judged to be mastered
And is
Judging the knowledge point as not mastered, otherwise classifying the knowledge point as not-determinable;
step 2.3.6) for the knowledge points whose mastery condition can not be determined in the evolution, use the set O ═ { O ═ O1，o2，…，oqAnd (c) represents.
5. The method for recommending personalized learning path and learning resource based on discipline knowledge graph according to claim 4, characterized in that the step 3) specifically comprises:
step 3.1) for the knowledge points which can not be judged to be mastered in the step 2), making an inference rule by using possible relations between the knowledge points and the knowledge points which are judged to be mastered, wherein the possible relations are as follows: if the knowledge point k is judged to be mastered, all the knowledge points and the antecedent knowledge points of the k are judged to be mastered, or at least one of the knowledge points of the k or the antecedent knowledge points is judged to be mastered; otherwise, all or the antecedent knowledge points of k are judged to be not mastered, and at least one of the antecedent knowledge points of k is judged to be not mastered;
step 3.2) taking the set O in the step 2.3.6) as a current knowledge point set, and extracting a knowledge point kiAccording to kiMastering of post-knowledge pointsThe situation is compared with the inference rule of said step 3.1), inference kiIf the user can not reason, the user goes to the next knowledge point until finding a knowledge point which can reason the result, and the knowledge point is recorded as kaTurning to the step 3.3), if all the knowledge points are taken, and the result cannot be deduced, exiting the cycle;
step 3.3) updating the knowledge state vector S in said step 2.1) and deleting knowledge points k from said set O of step 3.1)aWill be the knowledge point kaPremise knowledge points and knowledge points in the set O are extracted and put into the set X ═ X1，x2，…，xbReturning to the step 3.2) to be used as a new set O for circulation, wherein b is the number of the extracted knowledge points;
and 3.4) after all the reasoning is finished, all the knowledge points which can not judge the mastering conditions are treated as the masterless knowledge points, and the knowledge state vector S in the step 2.1) is updated to form the final knowledge point mastering conditions of the learners.
6. The method for recommending personalized learning path and learning resource based on discipline knowledge graph according to claim 1, characterized in that the step 4) specifically comprises:
step 4.1) performing word segmentation and word removal operation on the resource text data, extracting 25 knowledge point keywords by using TextRank as semantic features of resources, expressing the keywords as 50-dimensional word vectors by using BERT, and using Si＝{si1，si2，…，sinDenotes resource SiWherein i is a subscript and n is the number of word vectors; for each resource SiMaintaining a list of entities, referred to as a knowledge repository KsourceThen resource SiCan be expressed as: ksource(Si)＝{si1，si2，…，sin}；
Step 4.2) taking all knowledge point entity sets contained in the historical resources learned by the learner as knowledge point sets learned by the user according to the historical behaviors of the learner, and including the knowledge point entity sets into a knowledge base K (user) { K ═ KsourceS1，KsourceS2，…，KsourceSNAdding knowledge points contained in the new resources into a knowledge base of the learner every time the learner learns the new resources;
and 4.3) calculating text similarity by using the word vector as the similarity between the interest of the learner and the resource, wherein the similarity is defined as follows:
where user represents the target learner, K (user) represents the set of historical resources for this learner,
and
respectively representing the word vector representation of the resource knowledge base resource i and the resource j that the learner has learned.
7. The method for recommending personalized learning path and learning resource based on discipline knowledge graph according to claim 3, characterized in that said step 5) specifically comprises:
step 5.1) defining knowledge compactness: for any knowledge point k in the discipline knowledge graph G1And k is2，k2To k1Is knowledge point k2Is divided by k2To k1Shortest path distance SP (k)2，k1) Defining the postknowledge points of the same prerequisite knowledge point, defining that they do not have precedence order, setting shortest path distance as 1, if knowledge point k is2And k is1If there is no path relation or the same, the shortest path distance is set to-1, and the shortest path distances are the shortest path distances that the knowledge graph can search, except for the special case. The knowledge compactness is expressed by the following formula:
step 5.2) extracting the knowledge points which are not mastered by the learner in the step 2.1) from the knowledge state vector S obtained in the step 2.1 to form a knowledge point set DK, and using a knowledge resource base K (S) to represent the set of all the knowledge points of the resource S, so that the knowledge closeness from the resource S to the set of the knowledge points which are not mastered by the user represents the similarity between the mastered condition of the knowledge points of the learner and the knowledge points of the resource, and the similarity is expressed by the following formula:
wherein k isiRepresents an arbitrary knowledge point in the knowledge resource library K (S), in (k)i) I represents the importance of the knowledge points, and the shortest path distance SP represents the association between the knowledge points.
8. The method for recommending personalized learning path and learning resource based on discipline knowledge graph according to claim 3, characterized in that said step 6) specifically comprises:
step 6.1) in order to simultaneously consider the balance between the learning interest of the user and the knowledge level of the user when recommending resources, linear weighted fusion of the two types of similarity is introduced as a basis for recommending the resources, and the method can be specifically expressed by the following formula:
Sim(user，i，S)＝η·Interest(user，i)+(1-η)·DOC(S，user) (5)，
wherein the parameter η is a self-defined parameter to balance the weighted relationship between learner interest and knowledge point mastery level.
9. The method for recommending personalized learning path and learning resource based on discipline knowledge graph according to claim 8, characterized in that said step 7) specifically comprises:
step 7.1) extracting knowledge state of learner according to the step 2.1)The knowledge points with the quantity S median value of-1 are taken as the knowledge points which are not mastered by the learner and are stored into a set Kw＝{k1，k2，…，kwW is the total number of knowledge points which are not mastered by the learner;
step 7.2) from the set KwWhere the pre-selected knowledge point does not exist in KwKnowledge points with or without prior knowledge points are stored in a set Rr＝{k1，k2，…，krIn the equation, r is the number of knowledge points selected in this loop, from the set KwDeleting the selected knowledge points;
step 7.3) repeat step 7.2) until set KwIf the sequence is empty, the sequence to be learned L ═ { R } is finally generated1，R2，…，RlAnd expressed by using a visualization method;
and 7.4) generating a resource recommendation sequence of Top-N according to the resource recommendation Sim generated in the step 6.1), and finally integrating into learning path recommendation to form synchronous recommendation of the learning path and the learning resource.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111368109.7A CN114065040A (en) | 2021-11-18 | 2021-11-18 | Individual learning path and learning resource recommendation method based on discipline knowledge graph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111368109.7A CN114065040A (en) | 2021-11-18 | 2021-11-18 | Individual learning path and learning resource recommendation method based on discipline knowledge graph |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114065040A true CN114065040A (en) | 2022-02-18 |
Family
ID=80277761
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111368109.7A Pending CN114065040A (en) | 2021-11-18 | 2021-11-18 | Individual learning path and learning resource recommendation method based on discipline knowledge graph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114065040A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115329200A (en) * | 2022-08-26 | 2022-11-11 | 国家开放大学 | Teaching resource recommendation method based on knowledge graph and user similarity |
| CN116628339A (en) * | 2023-06-09 | 2023-08-22 | 国信蓝桥教育科技股份有限公司 | Educational resource recommendation method and system based on artificial intelligence |
| CN117131288A (en) * | 2023-09-07 | 2023-11-28 | 苏州椿之硕信息科技有限公司 | Recommendation system and method for obstetric and academic research |
| CN117672027A (en) * | 2024-02-01 | 2024-03-08 | 青岛培诺教育科技股份有限公司 | VR teaching method, device, equipment and medium |
-
2021
- 2021-11-18 CN CN202111368109.7A patent/CN114065040A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115329200A (en) * | 2022-08-26 | 2022-11-11 | 国家开放大学 | Teaching resource recommendation method based on knowledge graph and user similarity |
| CN115329200B (en) * | 2022-08-26 | 2024-04-26 | 国家开放大学 | Teaching resource recommendation method based on knowledge graph and user similarity |
| CN116628339A (en) * | 2023-06-09 | 2023-08-22 | 国信蓝桥教育科技股份有限公司 | Educational resource recommendation method and system based on artificial intelligence |
| CN116628339B (en) * | 2023-06-09 | 2023-11-17 | 国信蓝桥教育科技股份有限公司 | Educational resource recommendation method and system based on artificial intelligence |
| CN117131288A (en) * | 2023-09-07 | 2023-11-28 | 苏州椿之硕信息科技有限公司 | Recommendation system and method for obstetric and academic research |
| CN117672027A (en) * | 2024-02-01 | 2024-03-08 | 青岛培诺教育科技股份有限公司 | VR teaching method, device, equipment and medium |
| CN117672027B (en) * | 2024-02-01 | 2024-04-30 | 青岛培诺教育科技股份有限公司 | VR teaching method, device, equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114065040A (en) | Individual learning path and learning resource recommendation method based on discipline knowledge graph | |
| Liu et al. | Ekt: Exercise-aware knowledge tracing for student performance prediction | |
| Song et al. | Jkt: A joint graph convolutional network based deep knowledge tracing | |
| Gilbert et al. | Simulating societies: the computer simulation of social phenomena | |
| Niiniluoto | Realism, relativism, and constructivism | |
| CN110991645A (en) | Self-adaptive learning method, system and storage medium based on knowledge model | |
| CN113344053B (en) | Knowledge tracking method based on examination question different composition representation and learner embedding | |
| Wu et al. | SGKT: Session graph-based knowledge tracing for student performance prediction | |
| CN113033808A (en) | Deep embedded knowledge tracking method based on exercise difficulty and student ability | |
| Nkambou | Modeling the domain: An introduction to the expert module | |
| CN116594768A (en) | Large-model-oriented universal tool collaboration and refinement learning system and method | |
| CN113283488B (en) | Learning behavior-based cognitive diagnosis method and system | |
| Vesin et al. | Rule-based reasoning for building learner model in programming tutoring system | |
| Guan et al. | Kg4ex: An explainable knowledge graph-based approach for exercise recommendation | |
| Yue et al. | Augmenting interpretable knowledge tracing by ability attribute and attention mechanism | |
| Khenissi et al. | Automatic generation of fuzzy logic components for enhancing the mechanism of learner's modeling while using educational games | |
| Lofi | Just ask a human?-controlling quality in relational similarity and analogy processing using the crowd | |
| Liao et al. | Knowledge graph and its applications in MOOC and SPOC | |
| Pardos et al. | Distributed representation of misconceptions | |
| Ramírez-Noriega et al. | An ontology of the object orientation for intelligent tutoring systems | |
| Doran | Computer models as tools for archaeological hypothesis formation | |
| Abd El-Rady | An ontological model to predict dropout students using machine learning techniques | |
| Sun et al. | Adapting New Learners and New Resources to Micro Open Learning via Online Computation | |
| Liu et al. | SARLR: Self-adaptive Recommendation of Learning Resources. | |
| Irdesel et al. | Supplement for “development of a web-based strategic management expert system using knowledge graphs |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |