CN114925610A - Learner knowledge structure and level modeling method, system, equipment and terminal - Google Patents
Learner knowledge structure and level modeling method, system, equipment and terminal Download PDFInfo
- Publication number
- CN114925610A CN114925610A CN202210570381.1A CN202210570381A CN114925610A CN 114925610 A CN114925610 A CN 114925610A CN 202210570381 A CN202210570381 A CN 202210570381A CN 114925610 A CN114925610 A CN 114925610A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- knowledge point
- student
- association
- learning
- 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
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
Abstract
The invention belongs to the technical field of education data mining, and discloses a learner knowledge structure and level modeling method, a system, equipment and a terminal, wherein the learner knowledge structure and level modeling method, the system, the equipment and the terminal are used for collecting answer data of student subjects and constructing a knowledge association graph, a knowledge point transfer probability graph and a knowledge point co-occurrence probability graph; constructing a knowledge association graph and carrying out association classification on knowledge points; introducing a long-time and short-time memory network, fusing a space propagation mechanism, designing a learning analysis model based on long-time dependence of sequence, and modeling the learning condition of students; updating the learning state of the student, analyzing the association information of the knowledge points, and updating the state of the knowledge points; and modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level. The learner knowledge structure and level modeling method is beneficial to accurately and effectively modeling the knowledge structure and the knowledge level of the student, thereby promoting the individual learning of the student and providing a new idea for the diagnosis and the prediction of the knowledge structure and the knowledge level of the student in the online learning platform.
Description
Technical Field
The invention belongs to the technical field of education data mining, and particularly relates to a learner knowledge structure and level modeling method, system, equipment and terminal.
Background
In recent years, with the leap of internet and communication technology, various online learning platforms are continuously developed, and particularly after a new crown epidemic situation is developed, online teaching is developed in various places in order to respond to a call of 'stopping class and not stopping learning', and further development of the online learning platform is promoted. The system is characterized in that various learning activities and learning performances of students are modeled by data mining technology, psychological and educational measurement technology and the like, so that the students are helped to realize personalized learning, the system is a particularly key part in an online learning platform, and knowledge structure and level analysis are key researches for assisting the realization of the personalized learning.
At present, many relevant workers have conducted various researches on the problem, and various methods and models are proposed, and the main problems to be solved can be summarized as follows: the method comprises the steps of firstly collecting answer data of students in a certain period of time from an online learning platform, wherein the answer data comprises relevant information of questions made and answer results of the questions, then modeling sequence data for recording student answer information by utilizing a cyclic neural network, a convolutional neural network and the like, modeling knowledge levels and structures of the students into a plurality of hidden state vectors which can change along with time, and updating the hidden state vectors correspondingly according to the answer results when the students finish answering once, so that the learning condition of the students can be observed in real time, the future performance of the students can be predicted, and corresponding learning guidance is provided for the students. In this regard, models and methods of some knowledge structure and level analysis techniques proposed by related workers can be classified into two categories according to whether the knowledge state of students at the knowledge point level and the relation between knowledge points are considered in the modeling process: the knowledge structure and level analysis method based on the modeling of the whole knowledge state of the student and the knowledge structure and level analysis method based on the modeling of the knowledge structure of the knowledge point hierarchy of the student. The knowledge structure and level analysis method based on the student's whole knowledge state modeling is generally less in network parameter, higher in model operation efficiency and free from additional prior information support. However, the knowledge state of the student is modeled into an integral vector in a general way, in an actual learning scene, the mastery levels of the student on different knowledge points of a certain subject are different and mutually dependent, and the knowledge structure and level analysis method based on the modeling of the integral knowledge state of the student cannot observe the levels of the student on different knowledge points in a more detailed way, so that the interpretability is low; the knowledge structure and level analysis method based on knowledge structure modeling of the knowledge point hierarchy of the students is closer to the actual learning scene, and has better interpretability by considering the knowledge structure state of the students in learning and different learning states of the students on different knowledge points. However, in such methods, the definition of the knowledge structure is usually single, or the knowledge structure is generated by completely depending on expert labels, or by a relevant statistical method, and the obtained knowledge structure information has a certain degree of sidedness.
How to model and characterize the mastery state of the student on each knowledge point independently and consider the relevance between the knowledge points; if multi-dimensional and accurate knowledge structure information is defined, the adverse effect of a subjective and single knowledge structure on the subsequent student knowledge state modeling is avoided; how to construct an interpretable knowledge state updating method in the process of modeling the knowledge state of a student, so that the reason for the change of the knowledge state of the student can be observed, and the dependence on a neural network is relieved to a certain extent. Therefore, it is desirable to design a new knowledge structure and level modeling method and system for learners.
Through the above analysis, the problems and defects of the prior art are as follows:
(1) the traditional knowledge structure and level analysis technology neglects the common modeling of the whole knowledge structure state of the student and the mastered state of the student on a single knowledge point, generally models the whole state of the student by using a hidden state vector in a general way, and the knowledge point is not isolated under the actual learning situation, so that the model accuracy is low.
(2) The definition method of knowledge structure information introduced by the traditional knowledge structure and level analysis technology is generally single, and the characterization and description of the association between knowledge points have certain one-sidedness.
(3) The traditional modeling method of the knowledge structure and level analysis technology for the knowledge state of the student relies on a neural network to automatically update the knowledge state of the student, the generation reason of the knowledge state of the student cannot be observed, and the interpretability is poor.
Disclosure of Invention
The invention provides a learner knowledge structure and level modeling method, a learner knowledge structure and level modeling system, learner knowledge structure and level modeling equipment and a learner knowledge structure and level modeling terminal, and particularly relates to a learner knowledge structure and level modeling method, a learner knowledge structure and level modeling system, a learner knowledge structure and level modeling medium, learner knowledge structure and level modeling terminal integrating a statistical chart and a subject knowledge chart.
The invention is realized by a learner knowledge structure and level modeling method, which comprises the following steps:
collecting answer data of student subjects, and constructing a knowledge association graph, a knowledge point transition probability graph and a knowledge point co-occurrence probability graph; constructing a knowledge association graph and associating and classifying knowledge points; introducing a long-time and short-time memory network, fusing a space propagation mechanism, designing a learning analysis model based on long-time dependence of sequence, and modeling the learning condition of students; updating the learning state of the student, analyzing the association information of the knowledge points, and updating the state of the knowledge points; and modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level.
Further, the learner knowledge structure and level modeling method comprises the following steps:
step one, collecting answer data of students on a certain subject from an online learning platform, wherein the answer data comprises answering questions, answering errors, answering time, answering contents and corresponding subject knowledge points, carrying out data preprocessing operation, generating a time sequence knowledge point-answering condition answer pair sequence for each student, and if a certain question comprises a plurality of knowledge points, resolving the question into answer pairs corresponding to the number of the knowledge points;
step two, making an answer pair sequence according to the time sequence of students, and constructing a knowledge association diagram facing to the actual answer situation; counting the frequency of successive occurrence of any two knowledge points and the frequency of successive correct answers to obtain a directed knowledge point transition probability graph and a undirected knowledge point co-occurrence probability graph;
thirdly, constructing a knowledge association diagram facing subject knowledge based on the question content and the knowledge point meanings in the answer data, dividing the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtaining a corresponding knowledge point association diagram by a manual labeling method;
introducing a long-time memory network, fusing a self-defined space propagation mechanism into the long-time memory network, designing a learning analysis model based on long-time dependence between sequences, and modeling the learning condition of students; and analyzing the knowledge point associated information facing the actual answer condition and subject knowledge while updating the learning state of the student at each moment, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
Further, the collecting the answer data of the students on a certain subject from the online learning platform and performing data preprocessing operation in the step one, generating a time-series knowledge point-answer condition answer pair sequence for each student, and if a certain question contains a plurality of knowledge points, resolving the question into answer pairs corresponding to the number of the knowledge points comprises:
(1) preprocessing the collected answer record data to generate a student set:
S={s 1 ,s 2 ,...s M };
topic set:
T={t 1 ,t 2 ,...,t N };
and (3) knowledge point set:
Q={q 1 ,q 2 ,...,q C };
(2) obtaining the answer sequence of each student according to the time sequence:
R s ={x 1 ,x 2 ,...,x t };
wherein x is t Questions to be answered by the student at time t and answer conditions, t t Questions answered at time t, t t ∈T,r t Indicating the condition of answering question, a t ∈{0,1},a t 0 means correct answer, r t 1 indicates an error in the answer;
x t =(t t ,r t );
(3) processing the answering records of each student into time sequence answering pairs of knowledge point levels according to knowledge points corresponding to the questions made by the students at each moment:
R s ={(q 1 ,r 1 ),(q 2 ,r 2 ),...,(q t ,r t )}。
further, constructing a knowledge association diagram facing to the actual answering situation through the time sequence answer pair sequence of the students in the second step; the method comprises the following steps of counting the frequency number of any two successive knowledge points and the frequency number of successive correct answers to obtain a directed knowledge point transition probability map and an undirected knowledge point co-occurrence probability map, wherein the method comprises the following steps:
(1) and traversing the answer records of each student, and counting the frequency of the successive occurrence of any two knowledge points and the successive correct answer:
COUNT ij =count((q i ,r i ),(q j ,r j ));
(2) fixing knowledge points i, counting the frequency of correct answers of the knowledge points i and all the knowledge points:
(3) calculating the probability of transferring from the knowledge point i to the knowledge point j to obtain a knowledge point transfer probability graph Prob transition :
(4) The undirected association is bidirectional association, and the frequency number transferred from the knowledge point i to the knowledge point j and the frequency number transferred from the knowledge point j to the knowledge point i are analyzed; if the undirected correlation between i and j is stronger, COUNT ij And COUNT ji The smaller the absolute value of the difference between, and COUNT ij And COUNT ji The larger the sum of (c) is, the more undirected relevance matrix Prob is calculated therefrom cooccurrence And adding an offset value Δ to the divisor:
(5) and (3) normalizing the calculated values:
further, in the third step, a knowledge association diagram facing subject knowledge is constructed based on the question content and the knowledge point meaning in the answer data, the knowledge point association is divided into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and the corresponding knowledge point association diagram is obtained by a manual labeling method, which includes:
(1) in the ICAI system, the association of parts of knowledge with the whole is called organization association; the composite knowledge points are composed of a plurality of knowledge points, one composite knowledge point comprises other composite knowledge points or meta knowledge points, and the knowledge point set is represented by one tree according to the division basis; the organization association is expressed by a function CR (i, j), if knowledge point j is a part of knowledge point i and knowledge points i and j are associated with the organization, CR (i, j) is 1, otherwise CR (i, j) is 0:
CR ij =CR(i,j);
(2) besides artificially dividing organization association, the meaning expressed by the content of each knowledge point has internal association, and the association between the knowledge points is embodied from another aspect; before learning the knowledge point j, the knowledge point i must be mastered, and the knowledge point i is called as a supporting knowledge point of the knowledge point j, namely the knowledge point i is a preliminary knowledge point of the knowledge point j; the association is represented by a function DR (i, j), where DR (i, j) ═ 1 represents that if a student wants to grasp a knowledge point j, the student needs to grasp the knowledge point i first, otherwise, DR (i, j) ═ 0, so that:
DR ij =DR(i,j);
(3) another association is extended by organizational associations: all knowledge points forming the same composite knowledge point mutually form a brother relationship and are represented by a function BR (i, j); if it isCR (k, j) equals 1, BR (i, j) equals 1, otherwise BR (i, j) equals 0:
BR ij =BR(i,j);
(4) if the knowledge point i and the related content of the knowledge point j or the background content of the knowledge point i are overlapped, but i and j do not form the three relations, the knowledge point i and j form a reference relation; in the learning process, if a student grasps one of the knowledge point i or the knowledge point j, the student already has certain related background knowledge when learning the other knowledge point, and the function FR (i, j) is used for expressing the relationship; if knowledge points i, j can provide reference information to each other, FR (i, j) is 1, otherwise FR (i, j) is 0:
FR ij =FR(i,j)。
furthermore, a long-time and short-time memory network is introduced into the step four, a self-defined space propagation mechanism is fused into the long-time and long-time memory network, a learning analysis model based on long-time dependence among sequences is designed, and the learning condition of a student is modeled; analyzing knowledge point associated information for actual answer conditions and subject knowledge while updating the learning state of the student at each moment, and correspondingly updating the state of the related knowledge points, thereby modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level, comprising:
(1) the learning records of the students are coded and expressed, and the answering records of the students at the time t are (q) t ,r t ) The answer knowledge point is q t The answer result is r t And carrying out One-Hot coding processing on the answer pair:
the knowledge point sequence answered by each student from time 0 to t is:
q_seq s ={q 0 ,q 1 ,...,q t };
the embedding of the learning record from 0 to t time for each student is represented as:
INPUT s ={input 0 ,input 1 ,...,input t };
(2) the whole learning process of the LSTM modeling student specifically comprises the following steps:
2.1) processing a time in the time sequence in each step, wherein the input at the time t is input t Knowledge points and answers representing answers at time t, using an embedded matrix M x (M x ∈R 2C ) Performing embedded learning representation to obtain an embedded vector x t :
x t =input t M x ;
2.2) in the modeling process, use a state matrix A t Representing the learning state of the student at all knowledge points,time t, sequence of knowledge points answered by the student and state matrix A of the student at the previous time t-1 And obtaining the learning state of the student at the corresponding knowledge point at the last moment:
2.3) embedding vector x for learning record at t moment t State vector of student at corresponding knowledge point at t-1 momentAs input, an LSTM module is constructed to simulate the learning process of students, and a state matrix is updated;
2.3.2) constructing a forgetting gate:
2.3.3) construct output gate:
2.3.4) updating the learning state matrix of the student:
c t =f t *c t-1 +i t *g t ;
(3) analyzing knowledge point associated information Prob facing to actual answer situation transition And Prob cooccurrence When the student answers the question about the knowledge point q at the time t, the learning state on the knowledge point q changes, and then the learning state of the student on the knowledge point related to the knowledge point q changes correspondingly.
3.1) after the student answers the question about the knowledge point q, the learning state about the knowledge point q changes to a certain extent no matter the student is wrong, and the student usesRepresents the variation vector:
3.2) transition of probability matrix Prob from knowledge points transition Acquiring a knowledge point set having a transfer association with a knowledge point q:
3.3) calculating the influence of the change of state of the knowledge point q on the knowledge point with transition relation, wherein Emb q For the embedded representation of knowledge point q:
3.4) transition of probability matrix Prob from knowledge points cooccurrence Acquiring a knowledge point set having co-occurrence association with a knowledge point q:
3.5) calculating the influence of knowledge points having co-occurrence correlation due to the state change of the knowledge point q, wherein Emb q For the embedded representation of knowledge point q:
3.6) from the set of knowledge points that have a transfer relationship with knowledge point qSet of knowledge points having co-occurrence association with knowledge point qAnd corresponding influence vectorObtaining the actually generated learning state change, wherein f is a fully connected network layer:
(4) and analyzing knowledge point association information facing disciplinary knowledge, finding knowledge points which have organization association, support association, brother association and reference association with the knowledge point q according to knowledge association graphs CR, DR, BR and FR, and modeling the state change of the students on the knowledge points.
4.1) searching a knowledge point set related to the knowledge point q under the organization association:
4.2) retrieving a knowledge point set supporting the knowledge point q:
4.3) searching brother knowledge points of the knowledge point q:
4.4) finding a knowledge point with reference to the knowledge point q:
4.5) calculating the influence of the state change of the knowledge point q on the knowledge point associated with the state change:
4.6) obtaining the actually generated learning state change by correspondingly associating the knowledge point set and the influence vector:
4.7) updating the state matrix of the student according to the six types of learning state change vectors:
4.7.1) aggregate six variation vectors:
4.7.2) update student's state matrix:
(5) diagnosing knowledge cognition of the student from the updated state matrix and predicting future performance of the student:
r′ t =f predict (A t ,q t );
(6) defining a loss function according to the predicted performance and the real performance:
Loss=-∑ t (r t logr′ t +(1-r t )log(1-r′ t ));
(7) and updating all weight coefficients and bias coefficients in the model according to the loss function value and the gradient descent rule.
Another object of the present invention is to provide a learner knowledge structure and level modeling system applying the learner knowledge structure and level modeling method, the learner knowledge structure and level modeling system comprising:
the system comprises a construction module of answer pair sequences, a data processing module and a data processing module, wherein the construction module of answer pair sequences is used for collecting answer data of students on a certain subject from an online learning platform, carrying out preprocessing operation and generating time sequence knowledge point-answer condition answer pair sequences for each student;
the knowledge association diagram building module is used for building a knowledge association diagram facing to the actual answering situation through the time sequence answering pair sequence of the students;
the probability map building module is used for counting the frequency of successive occurrence of any two knowledge points and the frequency of successive correct answers to obtain a directed knowledge point transfer probability map and a undirected knowledge point co-occurrence probability map;
the knowledge association diagram building module is used for building a knowledge association diagram facing subject knowledge based on the subject content and the knowledge point meaning in the answer data;
the knowledge point association graph building module is used for dividing the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtaining the corresponding knowledge point association graph by a manual marking method;
the learning condition modeling module is used for introducing a long-time memory network, fusing a self-defined space propagation mechanism, designing a learning analysis model based on long-time dependence between sequences and modeling the learning condition of a student;
and the learning state modeling module is used for updating the learning state of the student at each moment, analyzing the knowledge point association information facing the actual answer condition and subject knowledge, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
It is a further object of the invention to provide a computer device comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to perform the steps of:
collecting answer data of student subjects, and constructing a knowledge association graph, a knowledge point transition probability graph and a knowledge point co-occurrence probability graph; constructing a knowledge association graph and associating and classifying knowledge points; a long-time and short-time memory network is introduced, a space propagation mechanism is fused, a learning analysis model based on sequence long-time dependence is designed, and the learning condition of a student is modeled; updating the learning state of the student, analyzing the associated information of the knowledge points, and updating the state of the knowledge points; and modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level.
It is another object of the present invention to provide a computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
collecting answer data of student subjects, and constructing a knowledge association graph, a knowledge point transition probability graph and a knowledge point co-occurrence probability graph; constructing a knowledge association graph and carrying out association classification on knowledge points; a long-time and short-time memory network is introduced, a space propagation mechanism is fused, a learning analysis model based on sequence long-time dependence is designed, and the learning condition of a student is modeled; updating the learning state of the student, analyzing the associated information of the knowledge points, and updating the state of the knowledge points; and modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level.
Another objective of the present invention is to provide an information data processing terminal for implementing the learner knowledge structure and level modeling system.
In combination with the technical solutions and the technical problems to be solved, please analyze the advantages and positive effects of the technical solutions to be protected in the present invention from the following aspects:
first, aiming at the technical problems existing in the prior art and the difficulty in solving the problems, the technical problems to be solved by the technical scheme of the present invention are closely combined with results, data and the like in the research and development process, and some creative technical effects are brought after the problems are solved. The specific description is as follows:
the invention provides a learner knowledge structure and level modeling method and system integrating a statistical chart and a subject knowledge chart. Counting the frequency of any two successive occurrences and successive correct answers of knowledge points to obtain a directed knowledge point transition probability map and a undirected knowledge point co-occurrence probability map, constructing a subject knowledge-oriented knowledge association map based on the meanings of the knowledge points, dividing knowledge point association into organization association, support association, brother association and reference association according to a domain knowledge base in an ICAI system, obtaining a corresponding knowledge point association map by a manual labeling method, introducing a long-term memory network and fusing a self-defined space propagation mechanism into the memory network, designing a learning analysis model capable of considering long-term dependence among sequences, modeling the learning condition of students, updating the learning state of the students at each moment, considering the actual answer condition and knowledge point association information facing subject knowledge, and correspondingly updating the state of the related knowledge points, therefore, the learning state of the student is accurately modeled, and the knowledge structure and the knowledge level of the student are diagnosed and estimated. The invention is beneficial to accurately and effectively modeling the knowledge structure and knowledge level of the student, thereby promoting the individual learning of the student and providing a new idea for the diagnosis and prediction of the knowledge structure and knowledge level of the student in the online learning platform.
According to the invention, the knowledge state of the student is more finely modeled, the state of the student on each knowledge point is individually characterized, and the knowledge state of the student can be more accurately calculated and updated; the invention fully utilizes subject background knowledge, fully excavates the knowledge map formed by the knowledge point set contained in the learning resources, and inputs the knowledge map into the model as prior information, thereby being beneficial to the improvement of model performance; the invention utilizes the long-time memory network module to model the time sequence learning process of students, and can effectively solve the long-term dependence problem in answer records.
The invention updates the knowledge state of the student after answering every time by using the self-defined updating mechanism, has higher interpretability compared with the method of completely depending on a neural network, and can help to trace the source of the progress or the step-back reason of the student on a certain knowledge point. The invention has higher accuracy for the diagnosis and prediction of the knowledge structure and level of the student, is superior to the traditional knowledge structure and level analysis method, and can provide more accurate guidance information for the student and more effective auxiliary information for teachers.
The indexes for carrying out the comparative experiment adopted by the invention comprise:
the invention compares the learner knowledge structure and level modeling method fusing the statistical chart and the subject knowledge chart with other knowledge structure and level analysis methods, and adopts the experimental indexes comprising: AUC (area Under dark), defined as the area enclosed by the ROC curve and the coordinate axis; and Accuracy, the Accuracy of the classification of the answer structure to the student. The present invention compares this approach to traditional knowledge structure and level analysis methods. All these methods are adjusted to the best state for fair comparison, the experimental comparison results of the learner knowledge structure and level modeling method and the conventional knowledge structure and level analysis method of the fusion statistical chart and discipline knowledge chart of the present invention are shown in table 1, and the training process of the model is shown in fig. 5 and 6.
TABLE 1 comparison of the results
Method | AUC | ACC |
Traditional knowledge structure and level analysis method | 0.7587 | 0.7430 |
Method for producing a composite material | 0.8357 | 0.7941 |
According to experimental results, compared with the traditional method, the learner knowledge structure and level modeling method fusing the statistical chart and the subject knowledge chart provided by the invention has the advantages that the AUC and the Accuracy are respectively improved by 7.7% and 5.11%, which shows that the method is more effective for modeling the knowledge state of students, and the introduced multi-dimensional knowledge structure definition method can also help to improve the performance of the knowledge structure and the level analysis method. In addition, the invention uses the LSTM to model the time sequence answering process more accurately, and provides more effective basis for the prediction of the future answering situation of the students. Experiments show that the prediction effect of the method is superior to that of the traditional method.
Secondly, considering the technical scheme as a whole or from the perspective of products, the technical effect and advantages of the technical scheme to be protected by the invention are specifically described as follows:
the invention effectively models the learning state of students on the knowledge point level, considers that the learning conditions of students on different knowledge points can be influenced mutually, models the association between the knowledge points, improves the accuracy of the model and is more close to the reality.
The invention adopts different angles to define the knowledge structure information, avoids completely relying on manual marking or completely relying on a neural network to define, fully utilizes the prior information of learning resources and improves the effectiveness of the model.
The learner knowledge structure and level modeling method fusing the statistical chart and the subject knowledge chart, provided by the invention, can be used for modeling the learning state of students on the knowledge point level and dynamically updating the knowledge state of the students on the basis of time sequence response data, so that the knowledge structure and level of the students are estimated and predicted, effective diagnosis information is provided for the students, the students are helped to adjust subsequent learning plans, the defects are missed and repaired, the learning efficiency is improved, and the students are promoted to develop personalized learning.
Third, as an inventive supplementary proof of the claims of the present invention, there are also presented several important aspects:
(1) the expected income and commercial value after the technical scheme of the invention is converted are as follows:
the invention supports personalized learning by big data and artificial intelligence, realizes the education concept of teaching according to the nature, can be widely applied to the fields of intelligent education systems, intelligent teaching guidance systems, intelligent teaching assistance, self-adaptive learning systems and the like, and has great commercial value.
(2) The technical scheme of the invention fills the technical blank in the industry at home and abroad:
the knowledge structure chart and the statistical knowledge structure chart in subject content are constructed, the knowledge structure charts with two different dimensions are fused, more comprehensive knowledge structure information can be obtained, and the blank that the knowledge structure chart is constructed only from a single angle and a multi-dimensional and comprehensive knowledge structure chart cannot be constructed in the industry technology is filled, so that the knowledge structure and the overall knowledge level of a learner are effectively modeled, and comprehensive and accurate guide information is provided for the personalized learning of the learner.
(3) The technical scheme of the invention solves the technical problem that people are eagerly to solve but can not be successfully solved all the time:
in the knowledge level diagnosis of learners, people are always eager for real-time and real-time mining methods oriented to dynamic knowledge structures. The learning activities performed by learners are essentially performed on their own knowledge structures, and the knowledge structures are continuously improved as the learners go deeper. The invention aims at the mining of dynamic knowledge structure and knowledge level and solves the technical problem of real-time knowledge structure diagnosis.
(4) The technical scheme of the invention overcomes the technical prejudice whether:
the traditional knowledge cognitive level mining method simply assumes that each knowledge point is independent, and the bias greatly hinders the application of intelligent knowledge level mining. The invention starts from the real disciplinary knowledge relationship, constructs a dynamic knowledge structure and a cognitive level diagnosis method, breaks the bias of the traditional technology, and greatly improves the practical value of the invention.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments of the present invention will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart of a learner knowledge structure and level modeling method provided in accordance with an embodiment of the present invention;
FIG. 2 is a schematic diagram of a learner knowledge structure and level modeling method according to an embodiment of the present invention;
FIG. 3 is a block diagram of a learner knowledge structure and level modeling system provided by an embodiment of the present invention;
FIG. 4 is a comparative illustration of experimental results provided by an embodiment of the present invention;
FIG. 5 and FIG. 6 are schematic diagrams of a model training process according to an embodiment of the present invention;
in the figure: 1. a response pair sequence construction module; 2. a knowledge association graph building module; 3. a probability map construction module; 4. a knowledge correlation diagram building module; 5. a knowledge point association graph building module; 6. a learning situation modeling module; 7. a learning state modeling module.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Aiming at the problems in the prior art, the invention provides a learner knowledge structure and level modeling method, a system, equipment and a terminal, and the invention is described in detail below by combining the attached drawings.
First, an embodiment is explained. This section is an explanatory embodiment expanding on the claims so as to fully understand how the present invention is embodied by those skilled in the art.
A description will first be made of some symbols appearing in the present invention, as shown in table 2.
TABLE 2 symbolic description
As shown in fig. 1, the method for modeling knowledge structure and level of a learner by fusing a statistical chart and a discipline knowledge chart according to an embodiment of the present invention includes the following steps:
s101, collecting answer data of students on a certain subject from an online learning platform, wherein the answer data comprises answering questions, answering errors, answering time, answering contents and corresponding subject knowledge points, carrying out data preprocessing operation, generating a time sequence knowledge point-answering condition answer pair sequence for each student, and if a certain question comprises a plurality of knowledge points, resolving the question into answer pairs corresponding to the number of the knowledge points;
s102, making an answer pair sequence according to the time sequence of students, and constructing a knowledge association diagram facing to the actual answer situation; counting the frequency number of any two knowledge points which occur one after another and the frequency number of any two knowledge points which are correctly answered one after another to obtain a directed knowledge point transition probability graph and an undirected knowledge point co-occurrence probability graph;
s103, constructing a knowledge association diagram facing subject knowledge based on question contents and knowledge point meanings in answer data, dividing the knowledge point association into organization association, support association, brother association and reference association according to knowledge point association in a domain knowledge base in an ICAI system, and obtaining a corresponding knowledge point association diagram by a manual labeling method;
s104, introducing a long-term memory network, fusing a self-defined space propagation mechanism into the long-term memory network, designing a learning analysis model capable of considering long-term dependence among sequences, and modeling the learning condition of students; and analyzing the knowledge point associated information facing the actual answer condition and subject knowledge while updating the learning state of the student at each moment, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
A schematic diagram of a learner knowledge structure and level modeling method provided by the embodiment of the invention is shown in FIG. 2.
As shown in fig. 3, the learner knowledge structure and level modeling system for fusing a statistical chart and a discipline knowledge chart according to an embodiment of the present invention includes:
the system comprises a response pair sequence construction module 1, a question-answer condition response pair sequence generation module and a question-answer condition response pair sequence generation module, wherein the response pair sequence construction module is used for collecting answer data of students on a certain subject from an online learning platform, carrying out preprocessing operation and generating a time sequence knowledge point-answer condition response pair sequence for each student;
the knowledge association diagram building module 2 is used for building a knowledge association diagram facing to the actual answering situation through the time sequence answering pair sequence of the students;
the probability map building module 3 is used for counting the frequency of the successive occurrence of any two knowledge points and the frequency of the successive correct answers to obtain a directed knowledge point transfer probability map and an undirected knowledge point co-occurrence probability map;
the knowledge association diagram building module 4 is used for building a knowledge association diagram facing subject knowledge based on the subject content and the knowledge point meaning in the answer data;
the knowledge point association graph building module 5 is used for dividing the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtaining a corresponding knowledge point association graph by a manual labeling method;
the learning condition modeling module 6 is used for introducing a long-time memory network and fusing a self-defined space propagation mechanism, designing a learning analysis model based on long-time dependence between sequences and modeling the learning condition of a student;
and the learning state modeling module 7 is used for updating the learning state of the student at each moment, analyzing the knowledge point association information facing the actual answer condition and subject knowledge, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
As a preferred embodiment, the present invention collects answer data of students on a certain subject from an online learning platform, including questions to be answered, correctness of answers, answer time, answer contents, and corresponding subject knowledge points, performs data preprocessing operation, generates a time-series "knowledge point-answer condition" answer pair sequence for each student, and if a certain question includes a plurality of knowledge points, disassembles the question into answer pairs corresponding to the number of knowledge points, and specifically includes:
(1.1) preprocessing the collected answer record data to generate a student set:
S={s 1 ,s 2 ,...s M }
topic set:
T={t 1 ,t 2 ,...,t N }
and (3) knowledge point set:
Q={q 1 ,q 2 ,...,q C }
(1.1.1) data information collected from a certain online learning platform is shown in table 3.
TABLE 3 data set information
Number of students | 4666 |
Knowledge points | 31 |
Number of questions | 226 |
Total number of replies recorded | 633970 |
Maximum number of student responses | 137 |
The shortest recorded number of student's answers | 21 |
(1.1.2) carrying out data preprocessing, deleting the response records with missing values, and sequencing the response records according to time;
(1.2) acquiring the answer sequence of each student according to the time sequence:
R s ={x 1 ,x 2 ,...,x t }
wherein x is t Questions to be answered by the student at time t and answer conditions, t t Topic answered at time t, t t ∈T,r t Indicating the condition of answering a t ∈{0,1},a t 0 means correct answer, r t 1 representsWrong answers:
x t =(t t ,r t )
(1.3) processing the answering record of each student into a time sequence answering pair of a knowledge point hierarchy according to the knowledge point corresponding to the question made by the student at each moment:
R s ={(q 1 ,r 1 ),(q 2 ,r 2 ),...,(q t ,r t )}
as a preferred embodiment, the invention constructs the knowledge association diagram facing the actual answering situation by making the answering pair sequence according to the time sequence of students. Counting the frequency of successive occurrence of any two knowledge points and the frequency of successive correct answers, and accordingly obtaining a directed knowledge point transition probability map and a undirected knowledge point co-occurrence probability map, which specifically comprises the following steps:
(2.1) traversing the answer records of each student, and counting the frequency that any two knowledge points appear in succession and are answered correctly in succession:
COUNT ij =count((q i ,r i ),(q j ,r j ))
(2.2) fixing the knowledge point i, counting the frequency of the correct answers of the knowledge point i and all the knowledge points in sequence:
(2.3) calculating the probability of transferring from the knowledge point i to the knowledge point j, thereby obtaining a knowledge point transfer probability graph Prob transition :
(2.4) undirected correlation, i.e. bi-directional correlation, takes into account both the frequency of transitions from knowledge point i to knowledge point j and the frequency of transitions from knowledge point j to knowledge point i. If the undirected correlation between i and j is stronger, COUNT ij And COUNT ji The smaller the absolute value of the difference therebetween, and COUNT ij And COUNT ji And alsoThe larger the correlation matrix Prob is, the more undirected the matrix Prob is calculated therefrom cooccurrence . In addition, to prevent the divisor from being zero, a bias value Δ is added to the divisor:
(2.5) because the value range of the calculation result of the formula is possibly large, the value is standardized:
as a preferred embodiment, the invention constructs a knowledge association diagram facing subject knowledge based on the subject content and knowledge point meaning in the answer data, divides the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtains the corresponding knowledge point association diagram by a manual marking method, wherein the method comprises the following steps:
in the (3.1) ICAI system, the association of the part of the knowledge points with the whole is called organization association, specifically, the complex knowledge points are all composed of several knowledge points, one complex knowledge point may contain other complex knowledge points or meta knowledge points, according to the division basis, the knowledge point set can be represented by one tree, here, the organization association is represented by a function CR (i, j), if the knowledge point j is a part of the knowledge point i, i.e., the knowledge points i, j are organized and associated, CR (i, j) is 1, otherwise, CR (i, j) is 0:
CR ij =CR(i,j)
(3.2) besides the artificial division of the organization association, the meaning expressed by the content of each knowledge point has an internal association, and the association between the knowledge points can be embodied from another aspect. If the knowledge point i must be mastered before learning the knowledge point j, the knowledge point i is called a supporting knowledge point of the knowledge point j, that is, the knowledge point i is a preliminary knowledge point of the knowledge point j. This association is represented by a function DR (i, j), where DR (i, j) ═ 1 indicates that the student needs to know the knowledge point i first if he wants to know the knowledge point j, otherwise, DR (i, j) ═ 0, so that:
DR ij =DR(i,j)
(3.3) Another association can be extended by the organizational association: the knowledge points composing the same composite knowledge point are in a brother relationship with each other and expressed by a function BR (i, j), namely: if it isCR (k, j) equals 1, BR (i, j) equals 1, otherwise BR (i, j) equals 0:
BR ij =BR(i,j)
(3.4) if the knowledge point i and the content related to the knowledge point j or the background content thereof overlap, but i and j do not form the three relationships, the knowledge points i and j form a reference relationship, and in the learning process, if a student grasps one of the knowledge point i or the knowledge point j, the student already has certain related background knowledge when learning the other knowledge point, the relationship is expressed by using a function FR (i, j), if the knowledge points i and j can provide reference information with each other, FR (i, j) is 1, otherwise, FR (i, j) is 0:
FR ij =FR(i,j)
as a preferred embodiment, the invention introduces a long-term and short-term memory network and integrates a self-defined spatial propagation mechanism into the long-term memory network, designs a learning analysis model which can take long-term dependence between sequences into consideration, models the learning condition of students, updates the learning state of the students at each moment, simultaneously takes the correlation information of knowledge points facing actual answer conditions and subject knowledge into consideration, and correspondingly updates the state of the related knowledge points, thereby accurately modeling the learning state of the students, and diagnosing and estimating the knowledge structure and level of the students, specifically comprising:
(4.1) the student's learning record is coded and expressed, and the student's answer record at the time t is (q) t ,r t ) The answer knowledge point is q t The answer result is r t And carrying out One-Hot coding processing on the answer pair:
the knowledge point sequence answered by each student from time 0 to t is:
q_seq s ={q 0 ,q 1 ,...,q t }
the embedding of the learning record from 0 to t time for each student is represented as:
INPUT s ={input 0 ,input 1 ,...,input t }
(4.2) in the online learning platform, the time range spanned by the student answer records may be large, and the Long-Term dependence problem in the recurrent neural network can be solved by a Long Short-Term Memory network (LSTM), so that the LSTM is adopted to model the whole learning process of the student.
(4.2.1) processing a time in the time sequence in each step, wherein the input at the time t is input t Knowledge points and answers representing answers at time t, using an embedded matrix M x (M x ∈R 2C ) Performing embedding learning representation to obtain an embedding vector x t :
x t =input t M x
(4.2.2) in modeling, a state matrix A is used t To represent the learning state of the student at all knowledge points,time t, sequence of knowledge points answered by the student and state matrix A of the student at the previous time t-1 The learning state of the student at the corresponding knowledge point at the last moment can be obtained:
(4.2.3) embedding vector x of learning record at time t t State vector of student at corresponding knowledge point at t-1 momentAs input, the LSTM module is constructed to simulate the learning process of the student and update its state matrix.
(4.2.3.2) construction of forgetting gate:
(4.2.3.3) construct the output gate:
(4.2.3.4) updating the learning state matrix of the student:
c t =f t *c t-1 +i t *g t
(4.3) consideration of knowledge Point Association information for practical answer situations, i.e. Prob transition And Prob cooccurrence When the student answers the question about the knowledge point q at the time t, the learning state of the student on the knowledge point q changes to a certain extent, and then the learning state of the student on the knowledge point related to the knowledge point q also changes correspondingly.
(4.3.1) the student answered something aboutAfter the question of the knowledge point q, the learning state of the knowledge point q will change to a certain extent regardless of the right or wrong, and the question is usedRepresents the change vector:
(4.3.2) transition of probability matrix Prob from knowledge points transition Acquiring a knowledge point set having transfer association with a knowledge point q:
(4.3.3) calculating the influence of the change of state of the knowledge point q on the knowledge point with transition relation, wherein Emb q For the embedded representation of knowledge point q:
(4.3.4) transition probability matrix Prob from knowledge points cooccurrence Acquiring a knowledge point set having co-occurrence association with a knowledge point q:
(4.3.5) calculating the influence of the change of state of the knowledge point q on the knowledge points associated with the co-occurrence thereof, where Emb q For the embedded representation of knowledge point q:
(4.3.6) from the set of knowledge points having a transfer relationship with knowledge point qSet of knowledge points having co-occurrence association with knowledge point qAnd corresponding influence vectorObtaining the actually generated learning state change, wherein f is a fully connected network layer:
and (4.4) considering knowledge point association information facing disciplinary knowledge, finding knowledge points which have organization association, support association, brother association and reference association with the knowledge point q according to the knowledge association graphs CR, DR, BR and FR, and modeling the state change of the students on the knowledge points.
(4.4.1) finding a knowledge point set related to the knowledge point q under the organization association:
(4.4.2) retrieving the knowledge point set supporting the knowledge point q:
(4.4.3) retrieving sibling knowledge points of knowledge point q:
(4.4.4) finding knowledge points that have reference to knowledge point q:
(4.4.5) calculating the influence of the state change of the knowledge point q on the knowledge point associated with the change:
(4.4.6) obtaining the actually generated learning state change according to the knowledge point set and the influence vector under the corresponding correlation:
(4.4.7) updating the state matrix of the student according to the above six types of learning state change vectors:
(4.4.7.1) aggregating six variation vectors:
(4.4.7.2) updating the student's state matrix:
(4.5) diagnosing knowledge awareness of the student from the updated state matrix and predicting the student's future performance:
r′ t =f predict (A t ,q t )
(4.6) defining a loss function based on the predicted and actual performances:
and (4.7) updating all weight coefficients and bias coefficients in the model according to the loss function values and the gradient descent rule. The experimental parameters of the present invention are shown in table 4.
Table 4 experimental parameter settings
Parameter(s) | Value of |
batch_size | 32 |
|
50 |
dropout | 0.3 |
learning rate | 0.001 |
|
50 |
The experimental development platform uses the mxnet framework. The experimental results of the method are compared with those of the traditional knowledge structure and level analysis method, the adopted indexes are AUC and ACCURACY, and the comparison result is shown in Table 5.
TABLE 5 results of the experiment
Method | AUC | ACC |
Traditional knowledge structure and level analysis method | 0.7587 | 0.7430 |
Method for producing a composite material | 0.8357 | 0.7941 |
FIG. 4 is a comparative illustration of experimental results provided by an embodiment of the present invention. According to experimental results, compared with a traditional knowledge structure and level analysis method, the AUC and ACC of the learner knowledge structure and level modeling method integrating the statistical chart and the subject knowledge chart are respectively improved by 7.7% and 5.11%, and the method is proved to be capable of better modeling the knowledge structure and level of students, higher in accuracy for student answer expression prediction and more effective than the traditional knowledge structure and level analysis method.
In conclusion, the learner knowledge structure and level modeling method and system integrating the statistical chart and the subject knowledge chart, provided by the invention, collect answer data of students on a certain subject from an online learning platform, make an answer pair sequence through the time sequence of the students and construct a knowledge association chart facing to the actual answer situation. Counting the frequency of any two successive knowledge points and the frequency of successive correct answers, obtaining a directed knowledge point transition probability graph and a non-directed knowledge point co-occurrence probability graph according to the frequency, constructing a knowledge association graph facing subject knowledge based on the subject content and the knowledge point meaning in answer data, dividing the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in a domain knowledge base in an ICAI system, obtaining the corresponding knowledge point association graph by a manual marking method, introducing a long-time memory network and fusing a self-defined space propagation mechanism into the long-time memory network, designing a learning analysis model capable of considering long-time dependence among sequences, modeling the learning condition of students, updating the learning state of the students at each moment, and simultaneously considering the actual answer condition and the knowledge point association information facing subject knowledge, and the state of the related knowledge points is correspondingly updated, so that the learning state of the student is accurately modeled, and the knowledge structure and the knowledge level of the student are diagnosed and estimated.
And II, application embodiment. In order to prove the creativity and the technical value of the technical scheme of the invention, the part is the application example of the technical scheme of the claims on specific products or related technologies.
The learner knowledge structure and level modeling method fusing the statistical chart and the subject knowledge chart, which is provided by the application embodiment of the invention, comprises the following steps:
(1) collecting answer data of students on a certain subject from an online learning platform, wherein the answer data comprises answering questions, answering errors, answering time, answer contents and corresponding subject knowledge points, carrying out data preprocessing operation, generating a time sequence knowledge point-answer condition answer pair sequence for each student, and if a certain question comprises a plurality of knowledge points, resolving the question into answer pairs corresponding to the number of the knowledge points;
(2) making an answer sequence according to the time sequence of students and constructing a knowledge association diagram facing to the actual answer condition; counting the frequency number of any two knowledge points which occur one after another and the frequency number of any two knowledge points which are correctly answered one after another to obtain a directed knowledge point transition probability graph and an undirected knowledge point co-occurrence probability graph;
(3) establishing a knowledge association diagram facing subject knowledge based on question content and knowledge point meanings in answer data, dividing the knowledge point association into organization association, support association, brother association and reference association according to knowledge point association in a domain knowledge base in an ICAI system, and obtaining a corresponding knowledge point association diagram by a manual labeling method;
(4) introducing a long-time memory network and fusing a self-defined space propagation mechanism into the long-time memory network, designing a learning analysis model capable of considering long-time dependence among sequences, and modeling the learning condition of students; and analyzing the knowledge point associated information facing the actual answer condition and subject knowledge while updating the learning state of the student at each moment, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
A schematic diagram of a learner knowledge structure and level modeling method provided by the application embodiment of the invention is shown in FIG. 2.
As an application embodiment of the present invention, the present invention collects answer data of students on a certain subject from an online learning platform, including questions to be answered, correctness of answers, answer time, answer content and corresponding subject knowledge points, performs data preprocessing operation, generates a time-series "knowledge point-answer condition" answer pair sequence for each student, and if a certain question contains a plurality of knowledge points, disassembles the question into answer pairs corresponding to the number of knowledge points, and specifically includes:
(1.1) preprocessing the collected answer record data to generate a student set:
S={s 1 ,s 2 ,...s M }
topic set:
T={t 1 ,t 2 ,...,t N }
and (3) knowledge point set:
Q={q 1 ,q 2 ,...,q C }
(1.1.1) data information collected from a certain online learning platform is shown in table 3.
Table 3 data set information
(1.1.2) carrying out data preprocessing, deleting the response records with missing values, and sequencing the response records according to time;
(1.2) acquiring the answer sequence of each student according to the time sequence:
R s ={x 1 ,x 2 ,...,x t }
wherein x is t Questions and answer conditions for students at time t, t t Topic answered at time t, t t ∈T,r t Indicating the condition of answering question, a t ∈{0,1},a t 0 means correct answer, r t 1 indicates an error in the answer:
x t =(t t ,r t )
(1.3) processing the answering record of each student into a time sequence answering pair of a knowledge point hierarchy according to the knowledge point corresponding to the question made by the student at each moment:
R s ={(q 1 ,r 1 ),(q 2 ,r 2 ),...,(q t ,r t )}
as an application embodiment of the invention, the invention constructs a knowledge association diagram facing to the actual answering situation by making the answering pair sequence according to the time sequence of students. Counting the frequency number of any two successive knowledge points and the frequency number of successive correct answers, and accordingly obtaining a directed knowledge point transition probability map and an undirected knowledge point co-occurrence probability map, wherein the method specifically comprises the following steps:
(2.1) traversing the answer records of each student, and counting the frequency that any two knowledge points appear in succession and are answered correctly in succession:
COUNT ij =count((q i ,r i ),(q j ,r j ))
(2.2) fixing the knowledge point i, counting the frequency of correct answers of the knowledge point i and all the knowledge points in sequence:
(2.3) calculating the probability of transferring from the knowledge point i to the knowledge point j, thereby obtaining a knowledge point transfer probability graph Prob transition :
(2.4) undirected correlation, i.e., bi-directional correlation, takes into account both the frequency of transitions from knowledge point i to knowledge point j and the frequency of transitions from knowledge point j to knowledge point i. If the undirected correlation between i and j is stronger, COUNT ij And COUNT ji The smaller the absolute value of the difference between, and COUNT ij And COUNT ji The larger the sum of (c) and (d), the more undirected correlation matrix Prob is calculated therefrom cooccurrence . In addition, to prevent the divisor from being zero, a bias value Δ is added to the divisor:
(2.5) because the value range of the calculation result of the formula is possibly large, the value is standardized:
as an application embodiment of the invention, the invention constructs a knowledge association diagram facing subject knowledge based on the subject content and knowledge point meaning in the answer data, divides the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtains the corresponding knowledge point association diagram by a manual marking method, wherein the method comprises the following steps:
in the (3.1) ICAI system, the association of the part of the knowledge points with the whole is called organization association, specifically, the complex knowledge points are all composed of several knowledge points, one complex knowledge point may contain other complex knowledge points or meta knowledge points, according to the division basis, the knowledge point set can be represented by one tree, here, the organization association is represented by a function CR (i, j), if the knowledge point j is a part of the knowledge point i, i.e., the knowledge points i, j are organized and associated, CR (i, j) is 1, otherwise, CR (i, j) is 0:
CR ij =CR(i,j)
(3.2) besides the artificial division of the organization association, the meaning expressed by the content of each knowledge point has an internal association, and the association between the knowledge points can be embodied from another aspect. If the knowledge point i must be mastered before learning the knowledge point j, the knowledge point i is called a supporting knowledge point of the knowledge point j, that is, the knowledge point i is a preliminary knowledge point of the knowledge point j. This association is represented by a function DR (i, j), where DR (i, j) ═ 1 indicates that if a student wants to know knowledge point j, it needs to know knowledge point i first, otherwise, DR (i, j) ═ 0, so that:
DR ij =DR(i,j)
(3.3) Another association can be extended by the organizational association: the knowledge points composing the same composite knowledge point are in a brother relationship with each other and expressed by a function BR (i, j), namely: if it isCR (k, j) equals 1, BR (i, j) equals 1, otherwise BR (i, j) equals 0:
BR ij =BR(i,j)
(3.4) if the knowledge point i and the knowledge point j relate to the content or the background content thereof, but i and j do not form the three relationships, the knowledge points i and j form a reference relationship, and if the student grasps one of the knowledge point i or the knowledge point j during the learning process, the student already has certain related background knowledge when learning the other knowledge point, and the relationship is expressed by a function FR (i, j), if the knowledge points i and j can provide reference information with each other, FR (i, j) is 1, otherwise, FR (i, j) is 0:
FR ij =FR(i,j)
as an application embodiment of the invention, the invention introduces a long-time and short-time memory network, fuses a self-defined space propagation mechanism into the long-time memory network, designs a learning analysis model capable of considering long-time dependence among sequences, models the learning condition of students, updates the learning state of the students at each moment, simultaneously considers the association information of knowledge points facing actual question answering conditions and subject knowledge, and correspondingly updates the state of the related knowledge points, thereby accurately modeling the learning state of the students and diagnosing and estimating the knowledge structure and level of the students, and specifically comprises the following steps:
(4.1) the learning records of the students are coded and expressed, and the answering records of the students at the time t are (q) t ,r t ) The answer knowledge point is q t The answer result is r t And carrying out One-Hot coding processing on the answer pair:
the knowledge point sequence answered by each student from time 0 to t is:
q_seq s ={q 0 ,q 1 ,...,q t }
the embedding of the learning record from 0 to t time for each student is represented as:
INPUT s ={input 0 ,input 1 ,...,input t }
(4.2) in the online learning platform, the answer records of the students may span a larger time range, and a Long-Term Memory network (LSTM) can solve the Long-Term dependence problem in the recurrent neural network, so that the LSTM is adopted to model the whole learning process of the students.
(4.2.1) processing a time in the time sequence in each step, wherein the input of the time t is input t Knowledge points and answer results representing answers at time t, using an embedded matrix M x (M x ∈R 2C ) Performing embedding learning representation to obtain an embedding vector x t :
x t =input t M x
(4.2.2) in the modeling process, a state matrix A is used t To represent the learning state of the student at all knowledge points,time t, sequence of knowledge points answered by the student and state matrix A of the student at the previous time t-1 The learning state of the student at the corresponding knowledge point at the last moment can be obtained:
(4.2.3) embedding vector x of learning record at t moment t State vector of student at corresponding knowledge point at t-1 momentAs input, the LSTM module is constructed to simulate the learning process of the student and update its state matrix.
(4.2.3.2) construction of forgetting gate:
(4.2.3.3) construct the output gate:
(4.2.3.4) updating the learning state matrix of the student:
c t =f t *c t-1 +i t *g t
(4.3) consideration of knowledge Point Association information for actual answer situations, i.e. Prob transition And Prob cooccurrence When the student answers the question about the knowledge point q at the time t, the learning state of the student on the knowledge point q changes to a certain extent, and then the learning state of the student on the knowledge point related to the knowledge point q also changes correspondingly.
(4.3.1) after the student answers the question about the knowledge point q, the learning state about the knowledge point q changes to a certain extent no matter the student is wrong, and the student uses the questionRepresenting the variation vector:
(4.3.2) transition of probability matrix Prob from knowledge points transition Acquiring a knowledge point set having a transfer association with a knowledge point q:
(4.3.3) calculate the effect on knowledge points with a transition relation to knowledge point q caused by the state change of knowledge point q, where Emb q For the embedded representation of knowledge point q:
(4.3.4) transition probability matrix Prob from knowledge points cooccurrence Acquiring a knowledge point set having co-occurrence association with the knowledge point q:
(4.3.5) calculating the influence of the change of state of the knowledge point q on the knowledge points associated with the co-occurrence thereof, where Emb q For the embedded representation of knowledge point q:
(4.3.6) from the set of knowledge points having a transfer relationship with knowledge point qKnowledge point set with co-occurrence association with knowledge point qAnd corresponding influence vectorObtaining the actually generated learning state change, wherein f is a fully connected network layer:
and (4.4) considering knowledge point association information facing discipline knowledge, finding knowledge points which have organization association, support association, brother association and reference association with the knowledge point q according to the knowledge association graphs CR, DR, BR and FR, and modeling the state change of the student on the knowledge points.
(4.4.1) finding a knowledge point set related to the knowledge point q under the organization association:
(4.4.2) retrieving the knowledge point set supporting the knowledge point q:
(4.4.3) retrieving sibling knowledge points of knowledge point q:
(4.4.4) finding knowledge points that have reference to knowledge point q:
(4.4.5) calculating the influence of the state change of the knowledge point q on the knowledge point associated with the change:
(4.4.6) obtaining the actually generated learning state change according to the knowledge point set and the influence vector under the corresponding correlation:
(4.4.7) updating the state matrix of the student according to the above six types of learning state change vectors:
(4.4.7.1) aggregating six variation vectors:
(4.4.7.2) updating the student's state matrix:
(4.5) diagnosing knowledge cognition of the student from the updated state matrix and predicting future performance of the student:
r′ t =f predict (A t ,q t )
(4.6) defining a loss function based on the predicted and actual performances:
and (4.7) updating all weight coefficients and bias coefficients in the model according to the loss function values and the gradient descent rule.
As shown in fig. 3, the learner knowledge structure and level modeling system for fusing a statistical chart and a discipline knowledge chart according to an embodiment of the present invention includes:
the system comprises a response pair sequence construction module 1, a question-answer condition response pair sequence generation module and a question-answer condition response pair sequence generation module, wherein the response pair sequence construction module is used for collecting answer data of students on a certain subject from an online learning platform, carrying out preprocessing operation and generating a time sequence knowledge point-answer condition response pair sequence for each student;
the knowledge association diagram building module 2 is used for building a knowledge association diagram facing to the actual answer situation through the time sequence answer pair sequence of the students;
the probability map building module 3 is used for counting the frequency of successive occurrence of any two knowledge points and the frequency of successive correct answers to obtain a directed knowledge point transfer probability map and a undirected knowledge point co-occurrence probability map;
the knowledge association diagram building module 4 is used for building a knowledge association diagram facing subject knowledge based on question contents and knowledge point meanings in answer data;
the knowledge point association graph building module 5 is used for dividing the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtaining the corresponding knowledge point association graph by a manual labeling method;
the learning condition modeling module 6 is used for introducing a long-term memory network, fusing a self-defined space propagation mechanism, designing a learning analysis model based on long-term dependence between sequences and modeling the learning condition of a student;
and the learning state modeling module 7 is used for updating the learning state of the student at each moment, analyzing the knowledge point association information facing the actual answer condition and subject knowledge, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
And thirdly, evidence of relevant effects of the embodiment. The embodiment of the invention achieves some positive effects in the process of research and development or use, and has great advantages compared with the prior art, and the following contents are described by combining data, diagrams and the like in the test process.
In addition, the method is compared with the experimental results of the traditional knowledge structure and level analysis method, the adopted indexes are AUC and ACCURACY, and the comparison result is shown in Table 5.
TABLE 5 results of the experiment
Method | AUC | ACC |
Traditional knowledge structure and level analysis method | 0.7587 | 0.7430 |
Method for producing a composite material | 0.8357 | 0.7941 |
FIG. 4 is a comparative illustration of experimental results provided by an embodiment of the present invention. According to experimental results, compared with a traditional knowledge structure and level analysis method, the AUC and ACC of the learner knowledge structure and level modeling method integrating the statistical chart and the subject knowledge chart are respectively improved by 7.7% and 5.11%, and the method is proved to be capable of better modeling the knowledge structure and level of students, higher in accuracy for student answer expression prediction and more effective than the traditional knowledge structure and level analysis method.
Experimental results show that the knowledge association diagram for actual question answering and the knowledge association diagram for subject knowledge constructed by the learner knowledge structure and level modeling method for fusing the statistical diagram and the subject knowledge diagram are effective for modeling the knowledge structure and level of the learner, can accurately estimate the whole knowledge mastering of the learner based on the knowledge structure diagram information with different dimensions, and in addition, introduces a long-time memory network and fuses a self-defined space propagation mechanism into the long-time memory network, designs a learning analysis model capable of considering long-time dependence among sequences, and models the learning condition of students to help to model the knowledge structure change process of the learner from two meanings of time and space; the learning state of the student is updated at each moment, meanwhile, the correlation information of knowledge points facing actual answering conditions and subject knowledge is considered, the state of the related knowledge points is correspondingly updated, the relationship of mutual influence among the knowledge points in an actual learning scene is effectively simulated, the learning state of the student can be accurately modeled by the relationship mode, and the knowledge structure and the knowledge level of the student can be more accurately diagnosed and estimated.
It should be noted that embodiments of the present invention can be realized in hardware, software, or a combination of software and hardware. The hardware portions may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or specially designed hardware. Those skilled in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such code being provided on a carrier medium such as a disk, CD-or DVD-ROM, programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier, for example. The apparatus and its modules of the present invention may be implemented by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of hardware circuits and software, e.g., firmware.
The above description is only for the purpose of illustrating the embodiments of the present invention, and the scope of the present invention should not be limited thereto, and any modifications, equivalents and improvements made by those skilled in the art within the technical scope of the present invention as disclosed in the present invention should be covered by the scope of the present invention.
Claims (10)
1. A learner knowledge structure and level modeling method, comprising:
collecting answer data of student subjects, and constructing a knowledge association graph, a knowledge point transition probability graph and a knowledge point co-occurrence probability graph; constructing a knowledge association graph and carrying out association classification on knowledge points; a long-time and short-time memory network is introduced, a space propagation mechanism is fused, a learning analysis model based on sequence long-time dependence is designed, and the learning condition of a student is modeled; updating the learning state of the student, analyzing the associated information of the knowledge points, and updating the state of the knowledge points; and modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level.
2. The learner knowledge structure and level modeling method of claim 1, wherein the learner knowledge structure and level modeling method comprises the steps of:
step one, collecting answer data of students on a certain subject from an online learning platform, wherein the answer data comprises answering questions, answering errors, answering time, answering content and corresponding subject knowledge points, performing data preprocessing operation, generating a time sequence knowledge point-answering condition answering pair sequence for each student, and if a certain question comprises a plurality of knowledge points, disassembling the questions into answering pairs corresponding to the number of the knowledge points;
step two, making an answer pair sequence through the time sequence of students, and constructing a knowledge association diagram facing to the actual answer condition; counting the frequency of successive occurrence of any two knowledge points and the frequency of successive correct answers to obtain a directed knowledge point transition probability graph and a undirected knowledge point co-occurrence probability graph;
thirdly, constructing a knowledge association diagram facing subject knowledge based on the question content and the knowledge point meanings in the answer data, dividing the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtaining a corresponding knowledge point association diagram by a manual labeling method;
introducing a long-time memory network, fusing a self-defined space propagation mechanism into the long-time memory network, designing a learning analysis model based on long-time dependence between sequences, and modeling the learning condition of students; and updating the learning state of the student at each moment, analyzing the knowledge point association information for actual answer conditions and subject knowledge, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
3. The learner knowledge structure and level modeling method according to claim 2, wherein said step one of collecting student answer data on a subject from the online learning platform and performing data preprocessing operation to generate a time series of knowledge point-answer conditions answer pair sequence for each student, and if a question includes a plurality of knowledge points, the parsing into answer pairs corresponding to the number of knowledge points comprises:
(1) preprocessing the collected answer record data to generate a student set:
S={s 1 ,s 2 ,...s M };
topic set:
T=*t 1 ,t 2 ,...,t N };
and (3) knowledge point set:
Q={q 1 ,q 2 ,...,q C };
(2) obtaining the answer sequence of each student according to the time sequence:
R s ={x 1 ,x 2 ,...,x t };
wherein x is t Questions to be answered by the student at time t and answer conditions, t t Topic answered at time t, t t ∈T,r t Indicating the condition of answering question, a t ∈{0,1},a t 0 means correct answer, r t 1 indicates an error in the answer;
x t =(t t ,r t );
(3) processing the answering record of each student into a time sequence answering pair of a knowledge point hierarchy according to a knowledge point corresponding to the question made by the student at each moment:
R s ={(q 1 ,r 1 ),(q 2 ,r 2 ),...,(q t ,r t )}。
4. the learner knowledge structure and level modeling method according to claim 2, wherein in said second step, a knowledge association graph for actual answer conditions is constructed through the student's time series answer pair sequence; the method comprises the following steps of counting the frequency number of any two successive knowledge points and the frequency number of successive correct answers to obtain a directed knowledge point transition probability map and an undirected knowledge point co-occurrence probability map, wherein the method comprises the following steps:
(1) and traversing the answer records of each student, and counting the frequency of the successive occurrence of any two knowledge points and the successive correct answer:
COUNT ij =count((q i ,r i ),(q j ,r j ));
(2) fixing knowledge points i, counting the frequency of correct answers of the knowledge points i and all the knowledge points:
(3) calculating the probability of transferring from the knowledge point i to the knowledge point j to obtain a knowledge point transfer probability graph Prob transition :
(4) The undirected association is bidirectional association, and the frequency number transferred from the knowledge point i to the knowledge point j and the frequency number transferred from the knowledge point j to the knowledge point i are analyzed; if the undirected correlation between i and j is stronger, COUNT ij And COUNT ji The smaller the absolute value of the difference therebetween, and COUNT ij And COUNT ji The larger the sum of (c) is, the more undirected relevance matrix Prob is calculated therefrom cooccurrence And adding an offset value Δ to the divisor:
(5) and (3) normalizing the calculated values:
5. the learner knowledge structure and level modeling method as set forth in claim 2, wherein the step three comprises constructing a knowledge association diagram facing subject knowledge based on question content and knowledge point meaning in answer data, dividing knowledge point association into organization association, support association, brother association and reference association according to knowledge point association in a domain knowledge base in an ICAI system, and obtaining a corresponding knowledge point association diagram by a manual labeling method comprises:
(1) in the ICAI system, the association of parts of knowledge with the whole is called organization association; the composite knowledge points are composed of a plurality of knowledge points, one composite knowledge point comprises other composite knowledge points or element knowledge points, and the knowledge point sets are represented by one tree according to the division basis; the organization association is expressed by a function CR (i, j), if knowledge point j is a part of knowledge point i and knowledge points i and j are associated with the organization, CR (i, j) is 1, otherwise CR (i, j) is 0:
CR ij =CR(i,j);
(2) besides artificially dividing organization association, the meaning expressed by the content of each knowledge point has internal association, and the association between the knowledge points is embodied from another aspect; before learning the knowledge point j, the knowledge point i must be mastered, and the knowledge point i is called as a supporting knowledge point of the knowledge point j, namely the knowledge point i is a preliminary knowledge point of the knowledge point j; the association is expressed by a function DR (i, j), where DR (i, j) is 1, which means that if a student wants to grasp a knowledge point j, the student needs to grasp the knowledge point i first, otherwise, DR (i, j) is 0, so that:
DR ij =DR(i,j);
(3) another association is extended by organizational associations: all knowledge points forming the same composite knowledge point mutually form a brother relationship and are represented by a function BR (i, j); if it isCR(k,j) 1, BR (i, j) is 1, otherwise BR (i, j) is 0:
BR ij =BR(i,j);
(4) if the knowledge point i and the related content of the knowledge point j or the background content of the knowledge point i are overlapped, but i and j do not form three associations, the knowledge point i and j form a reference relationship; in the learning process, if a student grasps one of the knowledge point i or the knowledge point j, the student already has certain related background knowledge when learning the other knowledge point, and the function FR (i, j) is used for expressing the relationship; if knowledge points i, j can provide reference information to each other, FR (i, j) is 1, otherwise FR (i, j) is 0:
FR ij =FR(i,j)。
6. the learner knowledge structure and level modeling method according to claim 2, wherein in step four, a long-time and short-time memory network is introduced and a self-defined spatial propagation mechanism is fused into the long-time and long-time memory network, a learning analysis model based on long-time dependency between sequences is designed, and the learning condition of a student is modeled; analyzing knowledge point associated information for actual answer conditions and subject knowledge while updating the learning state of the student at each moment, and correspondingly updating the state of the related knowledge points, thereby modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level, comprising:
(1) the learning records of the students are coded and expressed, and the answering records of the students at the time t are (q) t ,r t ) The answer knowledge point is q t In response to the result r t And carrying out One-Hot coding processing on the answer pair:
the knowledge point sequence answered by each student from time 0 to t is:
q_seq s ={q 0 ,q 1 ,...,q t };
the embedding of the learning record from 0 to t time for each student is represented as:
INPUT s ={input 0 ,input 1 ,...,input t };
(2) the whole learning process of the LSTM modeling student specifically comprises the following steps:
2.1) processing a time in the time sequence in each step, wherein the input at the time t is input t Knowledge points and answers representing answers at time t, using an embedded matrix M x (M x ∈R 2C ) Performing embedded learning representation to obtain an embedded vector x t :
x t =input t M x ;
2.2) in the modeling process, a state matrix A is used t Representing the learning state of the student at all knowledge points,time t, sequence of knowledge points answered by the student and state matrix A of the student at the previous time t-1 And obtaining the learning state of the student at the corresponding knowledge point at the last moment:
2.3) embedding vector x for learning record at t moment t State vector of student at corresponding knowledge point at t-1 momentAs input, an LSTM module is constructed to simulate the learning process of students, and a state matrix is updated;
2.3.2) constructing a forgetting gate:
2.3.3) construct output gate:
2.3.4) updating the learning state matrix of the student:
c t =f t *c t-1 +i t *g t ;
(3) analyzing knowledge point associated information Prob facing to actual answer situation transition And Prob cooccurrence When the student answers questions about the knowledge point q at the moment t, the learning state on the knowledge point q changes, and then the learning state of the student on the knowledge point related to the knowledge point q changes correspondingly;
3.1) after the student answers the question about the knowledge point q, the learning state about the knowledge point q changes to a certain extent no matter the student is wrong, and the student usesRepresents the change vector:
3.2) fromProbability matrix Prob of point-of-identity transition transition Acquiring a knowledge point set having a transfer association with a knowledge point q:
3.3) calculating the influence on the knowledge point with transition relation caused by the state change of the knowledge point q, wherein Emb q For the embedded representation of knowledge point q:
3.4) transition of probability matrix Prob from knowledge points cooccurrence Acquiring a knowledge point set having co-occurrence association with a knowledge point q:
3.5) calculating the influence of knowledge points having co-occurrence correlation due to the state change of the knowledge point q, wherein Emb q For the embedded representation of knowledge point q:
3.6) from the set of knowledge points that have a transfer relationship with knowledge point qSet of knowledge points having co-occurrence association with knowledge point qAnd corresponding influence vectorTo obtainActually generated learning state change, wherein f is a fully connected network layer:
(4) analyzing knowledge point association information facing disciplinary knowledge, finding knowledge points which have organization association, support association, brother association and reference association with the knowledge point q according to knowledge association graphs CR, DR, BR and FR, and modeling the state change of students on the knowledge points;
4.1) finding a knowledge point set related to the knowledge point q under the organization association:
4.2) retrieving a knowledge point set supporting the knowledge point q:
4.3) retrieving brother knowledge points of knowledge point q:
4.4) finding a knowledge point with reference to the knowledge point q:
4.5) calculating the influence of the state change of the knowledge point q on the knowledge point associated with the state change:
4.6) obtaining the actually generated learning state change by correspondingly associating the knowledge point set and the influence vector:
4.7) updating the state matrix of the student according to the six types of learning state change vectors:
4.7.1) aggregate six variation vectors:
4.7.2) update student's state matrix:
(5) diagnosing knowledge awareness of the student and predicting future performance of the student from the updated state matrix:
r′ t =f predict (A t ,q t );
(6) defining a loss function according to the predicted performance and the real performance:
Loss=-∑ t (r t logr′ t +(1-r t )log(1-r′ t ));
(7) and updating all weight coefficients and bias coefficients in the model according to the loss function value and the gradient descent rule.
7. A learner knowledge structure and level modeling system applying the learner knowledge structure and level modeling method as claimed in any one of claims 1 to 6, the learner knowledge structure and level modeling system comprising:
the system comprises a construction module of answer pair sequences, a data processing module and a data processing module, wherein the construction module of answer pair sequences is used for collecting answer data of students on a certain subject from an online learning platform, carrying out preprocessing operation and generating time sequence knowledge point-answer condition answer pair sequences for each student;
the knowledge association diagram building module is used for building a knowledge association diagram facing to the actual answer situation through the time sequence answer pair sequence of the students;
the probability map building module is used for counting the frequency of successive occurrence of any two knowledge points and the frequency of successive correct answers to obtain a directed knowledge point transfer probability map and a undirected knowledge point co-occurrence probability map;
the knowledge association diagram building module is used for building a knowledge association diagram facing subject knowledge based on the subject content and the knowledge point meaning in the answer data;
the knowledge point association graph building module is used for dividing the knowledge point association into organization association, support association, brother association and reference association according to the knowledge point association in the domain knowledge base in the ICAI system, and obtaining a corresponding knowledge point association graph by a manual labeling method;
the learning condition modeling module is used for introducing a long-time memory network, fusing a self-defined space propagation mechanism, designing a learning analysis model based on long-time dependence between sequences and modeling the learning condition of a student;
and the learning state modeling module is used for updating the learning state of the student at each moment, analyzing the knowledge point association information facing the actual answer condition and subject knowledge, and correspondingly updating the state of the related knowledge point, thereby modeling the learning state of the student and diagnosing and estimating the knowledge structure and level.
8. A computer arrangement comprising a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to carry out the steps of:
collecting answer data of student subjects, and constructing a knowledge association graph, a knowledge point transition probability graph and a knowledge point co-occurrence probability graph; constructing a knowledge association graph and associating and classifying knowledge points; a long-time and short-time memory network is introduced, a space propagation mechanism is fused, a learning analysis model based on sequence long-time dependence is designed, and the learning condition of a student is modeled; updating the learning state of the student, analyzing the association information of the knowledge points, and updating the state of the knowledge points; and modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level.
9. A computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the steps of:
collecting answer data of student subjects, and constructing a knowledge association graph, a knowledge point transition probability graph and a knowledge point co-occurrence probability graph; constructing a knowledge association graph and associating and classifying knowledge points; a long-time and short-time memory network is introduced, a space propagation mechanism is fused, a learning analysis model based on sequence long-time dependence is designed, and the learning condition of a student is modeled; updating the learning state of the student, analyzing the associated information of the knowledge points, and updating the state of the knowledge points; and modeling the learning state of the student, and diagnosing and estimating the knowledge structure and level.
10. An information data processing terminal for implementing the learner knowledge structure and level modeling system of claim 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210570381.1A CN114925610A (en) | 2022-05-24 | 2022-05-24 | Learner knowledge structure and level modeling method, system, equipment and terminal |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210570381.1A CN114925610A (en) | 2022-05-24 | 2022-05-24 | Learner knowledge structure and level modeling method, system, equipment and terminal |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114925610A true CN114925610A (en) | 2022-08-19 |
Family
ID=82809779
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210570381.1A Pending CN114925610A (en) | 2022-05-24 | 2022-05-24 | Learner knowledge structure and level modeling method, system, equipment and terminal |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114925610A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116976434A (en) * | 2023-07-05 | 2023-10-31 | 长江大学 | Knowledge point diffusion representation-based knowledge tracking method and storage medium |
-
2022
- 2022-05-24 CN CN202210570381.1A patent/CN114925610A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116976434A (en) * | 2023-07-05 | 2023-10-31 | 长江大学 | Knowledge point diffusion representation-based knowledge tracking method and storage medium |
CN116976434B (en) * | 2023-07-05 | 2024-02-20 | 长江大学 | Knowledge point diffusion representation-based knowledge tracking method and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Factors affecting the adoption of AI-based applications in higher education | |
CN111241243B (en) | Test question, knowledge and capability tensor construction and labeling method oriented to knowledge measurement | |
CN110941723A (en) | Method, system and storage medium for constructing knowledge graph | |
CN112508334B (en) | Personalized paper grouping method and system integrating cognition characteristics and test question text information | |
CN108257052B (en) | Online student knowledge assessment method and system | |
CN110991645A (en) | Self-adaptive learning method, system and storage medium based on knowledge model | |
CN112016767A (en) | Dynamic planning method and device for learning route | |
CN112257966B (en) | Model processing method and device, electronic equipment and storage medium | |
CN111126552B (en) | Intelligent learning content pushing method and system | |
CN109272160A (en) | Score on Prediction system and prediction technique | |
CN117252047B (en) | Teaching information processing method and system based on digital twinning | |
CN114254127A (en) | Student ability portrayal method and learning resource recommendation method and device | |
CN112069329B (en) | Text corpus processing method, device, equipment and storage medium | |
CN112687374A (en) | Psychological crisis early warning method based on text and image information joint calculation | |
CN114201684A (en) | Knowledge graph-based adaptive learning resource recommendation method and system | |
CN115455186A (en) | Learning situation analysis method based on multiple models | |
CN113408852A (en) | Meta-cognition ability evaluation model based on online learning behavior and deep neural network | |
CN114429212A (en) | Intelligent learning knowledge ability tracking method, electronic device and storage medium | |
CN116186250A (en) | Multi-mode learning level mining method, system and medium under small sample condition | |
CN115510286A (en) | Multi-relation cognitive diagnosis method based on graph convolution network | |
CN113283488B (en) | Learning behavior-based cognitive diagnosis method and system | |
CN114925610A (en) | Learner knowledge structure and level modeling method, system, equipment and terminal | |
Huizhong et al. | Research on the automation integration terminal of the education management platform based on big data analysis | |
CN116361744A (en) | Learner cognition tracking method and system for learning procedural evaluation | |
CN116521872A (en) | Combined recognition method and system for cognition and emotion and electronic equipment |
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 |