CN112508334B - Personalized paper grouping method and system integrating cognition characteristics and test question text information - Google Patents

Personalized paper grouping method and system integrating cognition characteristics and test question text information Download PDF

Info

Publication number
CN112508334B
CN112508334B CN202011233044.0A CN202011233044A CN112508334B CN 112508334 B CN112508334 B CN 112508334B CN 202011233044 A CN202011233044 A CN 202011233044A CN 112508334 B CN112508334 B CN 112508334B
Authority
CN
China
Prior art keywords
learner
test
test question
knowledge
text information
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.)
Active
Application number
CN202011233044.0A
Other languages
Chinese (zh)
Other versions
CN112508334A (en
Inventor
王志锋
余新国
左明章
叶俊民
张思
闵秋莎
罗恒
夏丹
姚璜
杨洋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central China Normal University
Original Assignee
Central China Normal University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Central China Normal University filed Critical Central China Normal University
Priority to CN202011233044.0A priority Critical patent/CN112508334B/en
Publication of CN112508334A publication Critical patent/CN112508334A/en
Application granted granted Critical
Publication of CN112508334B publication Critical patent/CN112508334B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
    • 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
    • G06N3/045—Combinations of 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
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/20—Education
    • G06Q50/205—Education administration or guidance
    • G06Q50/2057—Career enhancement or continuing education service
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Abstract

The invention belongs to the technical field of intelligent education, and discloses a personalized paper grouping method and a personalized paper grouping system for fusing cognitive characteristics and test question text information, wherein a cognitive diagnosis model is utilized to predict scores of learners on specific test questions based on cognitive levels; predicting the score of the learner on the specific test question based on the text information by using the cyclic neural network model; then constructing a probability matrix decomposition objective function based on the obtained learner based on the cognitive level and the predictive score based on the text information, and predicting the potential score of the learner on the specific test question; and finally, calculating KL divergence by using the estimated learner knowledge mastering vector and the learner incremental knowledge mastering vector, and selecting test papers which are suitable in difficulty and enable the learner knowledge mastering trend to be increased by combining potential scores of learners on the test questions to form personalized tests. The invention can self-define the paper-making result according to the test target and the test question difficulty, thereby greatly increasing the self-learning efficiency of the learner.

Description

Personalized paper grouping method and system integrating cognition characteristics and test question text information
Technical Field
The invention belongs to the technical field of intelligent education, and particularly relates to a personalized paper grouping method and system integrating cognitive characteristics and test question text information.
Background
At present, with the rapid development of the internet and the arrival of big data age, the traditional education industry has gradually begun to transform to digital education, and massive education resources are shared as information to an online education platform for learners to download and learn. The test questions of each subject are used by learners in large quantity as important resources in education to consolidate the learning knowledge of the learners in the class, however, the learners are difficult to directly screen out test questions really suitable for the learners from a large number of test questions, and more are trained by adopting the technical sea tactics. The personalized paper composing system can quickly master text information of the matched test questions according to the knowledge of the learner, gives out test questions which are suitable for the learning difficulty of the learner and aim to enhance the knowledge growth of the learner, so that the learner can learn and train better, and the learning efficiency of the learner is improved, and meanwhile, the learner can conduct targeted paper composing exercise on different learning conditions of each learner, so that the needs of the learner on personalized paper composing on the online learning platform are increasingly urgent.
The traditional paper grouping mode adopts a popular collaborative filtering recommendation idea in the field of electronic commerce, which is to analogize commodities to test questions, analogize users to learners, analogize the scoring of the commodities to the scoring of the learners to the test questions, so that the method is applied to collaborative filtering to obtain proper test questions, and a complete test paper is formed. However, the group paper test in this way is more prone to mining learner learning commonality, i.e. the test questions given to the target learner are more from the learner with high similarity, so that the simple questions and popular questions have larger recommended probability, the way is not given for the specific learning state of the learner, therefore, the group paper result cannot be obtained for the weak links mastered by the learner knowledge points, and the group paper result has no interpretation, and the learner more hopes to compose a set of questions suitable for the learner's cognitive characteristics through the group paper system, so as to conduct the targeted test question training on the weak knowledge points of the learner, therefore, the way cannot be well applied to the technical field of intelligent education. In the cognitive psychology, the cognitive diagnosis method is also used in a test paper of a learner in a later period, the cognitive diagnosis model can diagnose the learning state of the learner, the mastering condition of the learner on each knowledge point test question is analyzed according to the score of the learning model on the marked knowledge point test question, so that the correct answer probability of the learner on the given knowledge point test question is obtained, then the test question is screened, but the learning has certain concealment on the knowledge point mastering condition and cannot be directly calculated, in addition, the error of the result is possibly larger by using an estimation mode, and the commonality among the learners is ignored by using the paper strategy of the cognitive diagnosis model, so that a good test effect is difficult to obtain by using the method. Meanwhile, the traditional test paper assembly method often ignores text information of test questions, and subject keywords contained in the test question text often have a larger correlation with the probability of a learner answering the test questions correctly, so that analysis of the test question text is necessary to be added into a personalized test paper of the learner.
Through the above analysis, the problems and defects existing in the prior art are as follows:
(1) In the prior art, the traditional personalized paper composing method ignores the influence of test question text information on answering questions of learners.
(2) In the prior art, the method for performing test paper by using cognitive diagnosis ignores commonalities among learners, and the parameter estimation in a model of the method is sensitive to a data set and possibly generates larger errors;
(3) In the prior art, the method for combining test paper by collaborative filtering cannot fully consider knowledge point mastering conditions of learners, and can only combine paper according to learning commonalities of similar learners, neglecting learning characteristics of the learners in the learning process, and ensuring poor interpretation of the combined paper results.
The difficulty of solving the above problems is:
(1) How to integrate the test question text information into the cognition process of the learner, and connect the text with the score of the learner to obtain the mapping relation between the text and the score;
(2) How to combine the cognitive characteristics of the learner represented in the cognitive diagnosis result with the learning commonalities of the learner represented in the collaborative filtering result, namely comprehensively consider the learning commonalities and the characteristics, and make the prone scroll exercise aiming at the learning characteristics of the learner.
(3) How to measure or adjust the influence of learner learning characteristics, learning commonality and test question text information on the final paper result, and improve the application capability of the paper result under different expected conditions.
Through the analysis, the significance of solving the problems and the defects is as follows:
(1) In the invention, a cyclic neural network model RNN is used for extracting the text content of the test questions in the aspect of test question text analysis, and the mapping relation between the learning state of the learner and the test question text information is constructed through a full connection layer.
(2) The learning commonality and the characteristic and the test question text information are comprehensively considered by combining the learning characteristic of the learner represented in the cognitive diagnosis result and the learning commonality of the learner represented in the collaborative filtering result, and the test questions which are more difficult to test questions or are difficult to master to improve knowledge can be selected according to teaching targets.
(3) The invention takes the result of the model as prior information, can carry out personalized learning composition according to the learning conditions of different learners, gives out the learning analysis result with the learning state of the learners, the text information of test questions and the commonality among the learners, has strong interpretation of the composition result and realizes the teaching according to the material.
(4) Compared with the traditional test paper assembly algorithm, the test paper assembly method based on the cognitive diagnosis and the test paper text analysis has the advantages that compared with the traditional method, the test result shows that the test paper assembly method has larger performance improvement, and the defect of the traditional method in the aspect of test paper assembly is overcome. The method utilizes richer information, gives more accurate personalized scroll test to the learner, and improves the learning efficiency of the learner.
Disclosure of Invention
Aiming at the problems existing in the prior art, the invention provides a personalized paper grouping method and a personalized paper grouping system for fusing cognitive characteristics and test question text information. The invention aims to solve the problem that in the prior art, the traditional personalized paper composing method ignores the influence of test question text information on answering questions of learners; the method of testing the coil by using the cognitive diagnosis ignores the commonality among learners and the parameter estimation in the model is sensitive to the data set possibly generates larger error; the method for testing the scroll by utilizing collaborative filtering cannot practice the knowledge points of learners, and only can scroll according to the learning commonality of similar learners, so that the learning characteristics of the learners in the learning process are ignored, and the interpretation of the scroll result is poor.
The invention is realized in such a way that the personalized paper making method for fusing the cognitive characteristics of the learner and the text information of the test questions comprises the following steps:
step one, according to the actual answer situation of the learner and the distribution of the knowledge points of the test questions, estimating and calculating the knowledge mastering situation of the learner by using a cognitive diagnosis model, and predicting the score of the learner on the specific test questions based on the cognitive level;
extracting text content of the test questions by using a cyclic neural network model, constructing a mapping relation between a learning state of a learner and text information of the test questions through a full-connection layer, and predicting scores of the learner on specific test questions based on the text information;
thirdly, constructing a probability matrix decomposition objective function based on the obtained learner based on the cognitive level and the prediction score based on the text information, and predicting the potential score of the learner on a specific test question;
and step four, calculating KL divergence by using the estimated learner knowledge mastering vector and the learner incremental knowledge mastering vector, and selecting test papers which are provided with increased learner knowledge mastering trend and proper difficulty and are formed into personalized tests by combining potential scores of learners on the test questions.
Further, in the first step, according to the actual answer situation of the learner and the distribution of knowledge points of the test questions, estimating and calculating the knowledge mastering situation of the learner by using the cognitive diagnosis model, and predicting the score of the learner based on the cognitive level on the specific test questions includes:
(1.1) collecting the distribution of the test question knowledge points marked by the field expert and the response data of the learner; according to the distribution of the test question knowledge points marked by the field expert, calculating a Q matrix of the test question knowledge points required to be used in the cognitive diagnosis model;
(1.2) according to the learning prior knowledge point mastering mode eta, calculating the ideal answering situation of the learner:
wherein pi ij Represents the ideal answer condition of learner i on the jth test question, eta ik Representing the knowledge point k of learner i, q jk Indicating whether the known j-th test question examines the knowledge point k;
(1.3) estimating the probability s of a learner doing wrong test questions under the condition that the learner grasps all knowledge points of a test question and the probability g of the learner doing the test questions under the condition that the learner does not grasp all corresponding knowledge points according to a expectation maximization algorithm;
(1.4) calculating the probability of correct answer of the learner according to the estimated ideal answer situation of the learner, the probability s of the learner doing wrong test questions when the learner grasps some test questions to examine all knowledge points and the probability g of the learner doing the test questions when the learner does not grasp the corresponding knowledge points:
(1.5) obtaining a total likelihood function of the DINA model:
wherein l=2 K ；
(1.6) calculating knowledge mastery of the learner using maximum likelihood estimation based on the obtained total likelihood function:
(1.7) calculating the score condition of the learner on the new test question according to the knowledge mastery condition of the learner:
in the second step, the text content of the test question is extracted by using the cyclic neural network model, and a mapping relation between the learning state of the learner and the text information of the test question is constructed through the full-connection layer, and the predicting the score of the learner on the specific test question based on the text information comprises the following steps:
(2.1) performing text word segmentation, removal of stop words and other preprocessing on test question texts; processing test questions by using a continuous Word bag model of a Word vector model Word2vec, predicting Word vectors of target words according to Word vectors of a plurality of words in the context of the target words, vectorizing the test questions input after preprocessing, and obtaining embedded expression of test question Word levels;
(2.2) acquiring test question context characteristic representation by using a test question word vector as input and adopting a bidirectional long-short-time memory neural network, and acquiring test question sentence level representation by means of mean value pooling processing to acquire test question word embedded representation;
(2.3) constructing a full-connection depth neural network, performing feature fusion on the test question word embedding vector and the learner knowledge grasping vector to serve as input of the full-connection depth neural network, and outputting the score of a specific student on the test question;
(2.4) the result of the full connection layer (y 1 ,y 2 ,...,y n ) The processing is carried out by an output unit:
obtaining a value between 0 and 1, representing the probability of the student to answer the test question text correctly, comparing the probability with real score data, and carrying out weight correction in a network;
and (2.5) training the network model by using the training set data to obtain a trained neural network model, and predicting the answering results of the learner based on the text information by using the obtained trained neural network model.
Further, in step (2.1), the text word segmentation of the test question text includes: based on the mixed dictionary, a method combining a bidirectional maximum matching method and statistics is adopted to carry out mixed word segmentation on the test question text.
Further, in the step (2.1), the step of removing the stop word from the test question text includes: and adding words which are irrelevant to sentences and test question text topics, do not contribute to test question labeling tasks and have too low frequency in the test question text into the stop word stock, and deleting words of mixed word segmentation in the test question text, wherein the words appear in the stop word stock.
Further, in step (2.2), the acquiring the context feature representation of the test question by using the bidirectional long-short-time memory neural network includes:
the BiLSTM network adopts two LSTM to obtain the contextual characteristics of different test questions from opposite directions, and the formula is as follows:
Wherein a is 1 ,a 2 ,b 1 And b 2 G (·) is the hidden layer activation function for the weight coefficient,is the forward hidden layer output at the time t,outputting the backward hidden layer at the time t; and fusion of hidden layer output of two directions at each moment to construct final output h t ：
Wherein c 1 And c 2 And f (·) is the output activation function for the weight coefficient.
Further, in step (2.2), the obtaining the sentence-level representation of the test question through the mean value pooling process includes:
at the sentence-level representation layer of the test questions, the test question word embedded representation E is obtained through average pooling h ：
E h ＝mean(h 1 ,...,h t )
Where mean (·) is the average pooling operation, i.e., taking the average of the eigenvalues as output in the domain.
Further, in step (2.3), the fully connected deep neural network includes:
in the fully-connected deep neural network, the calculation method of the nth node value of the mth layer comprises the following steps:
where N is the number of units of the upper layer,representing the weight coefficients from the ith cell of the m-1 th layer to the nth cell of the m-th layer;
and obtaining a potential mapping relation between learner knowledge mastery, test question text information and scores by adopting a relu activation function.
In the third step, the method constructs a probability matrix decomposition objective function based on the obtained learner prediction score based on the cognitive level and the text information, and predicts the potential score of the learner on the specific test question, wherein the method comprises the following steps:
(3.1) marking the obtained score prediction based on cognitive level and score prediction learner based on text information as R 1 ,R 2 Using a probability matrix decomposition algorithm to construct a potential answer representation of the learner on the test question:
wherein U, V are respectively a learner characteristic matrix and a achievement characteristic matrix in the probability matrix decomposition, and alpha and beta are respectively adjusting parameters of learner learning conditions and test question text information;
(3.2) constructing a final objective function of the probability matrix decomposition:
and (3.3) optimizing an objective function by using a gradient descent method to obtain an optimal learner performance feature matrix U, V:
(3.4) predicting the performance of the learner by using the optimal feature matrix U and V obtained by training:
in the fourth step, the step of using the estimated learning knowledge mastering vector and the learning incremental knowledge mastering vector to calculate the KL divergence, and selecting the test paper for forming the personalized test by combining the potential scores of the learners on the test questions to increase the learning knowledge mastering trend and the test questions with proper difficulty comprises:
(4.1) obtaining learner knowledge mastering vector based on the analysis to obtain all incremental knowledge mastering vectors of the learner0≤d≤D；
(4.2) computing the learner knowledge mastery vector eta estimated by the cognitive diagnostic model i Incremental knowledge mastering with all learnersKL divergence measure of (2):
(4.3) selecting test paper which is used for personalized test and is composed of test questions with proper difficulty, wherein the learning knowledge mastering trend is increased:
another object of the present invention is to provide a personalized paper system for combining learner cognition characteristics and test question text information, for implementing the personalized paper method for combining learner cognition characteristics and test question text information, the personalized paper system for combining learner cognition characteristics and test question text information comprising:
the score prediction module based on the cognitive level is used for estimating and calculating the knowledge mastering condition of the learner by utilizing a cognitive diagnosis model according to the real answering condition of the learner and the knowledge point distribution of the test questions and predicting the score of the learner based on the cognitive level on the specific test questions;
the score prediction module based on the text information is used for extracting the text content of the test questions by using the cyclic neural network model, constructing the mapping relation between the learning state of the learner and the text information of the test questions through the full-connection layer, and predicting the score of the learner on the specific test questions based on the text information;
the topic selection strategy module is used for constructing a probability matrix decomposition objective function based on the obtained learner prediction scores based on the cognitive level and the text information and predicting potential scores of the learner on the specific test questions; and calculating KL divergence by using the estimated learner knowledge mastering vector and the learner incremental knowledge mastering vector, and selecting test papers which are provided with increased learner knowledge mastering trend and proper difficulty and are formed into personalized tests by combining potential scores of learners on the test questions.
Another object of the present invention is to provide a computer device, where the computer device includes a memory and a processor, where the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the personalized grouping method for fusing the learner cognitive characteristic and the test question text information.
Another object of the present invention is to provide a computer readable storage medium storing a computer program, where the computer program when executed by a processor causes the processor to execute the personalized grouping method for fusing learner cognitive characteristics and test question text information.
The invention further aims to provide an information data processing terminal which is used for realizing the personalized paper grouping method for fusing the cognitive characteristics of the learner and the text information of the test questions.
By combining all the technical schemes, the invention has the advantages and positive effects that: according to the invention, through fusion of the cognitive characteristics of the learner and the text information of the test questions, proper training test questions are provided for the learner, so that the problems of traditional thousand people, questions, sea tactics and the like are eliminated, the learner can perform more targeted training, and the learning efficiency is improved. The invention has wide application prospect in the fields of personalized learning, self-adaptive learning and intelligent education.
The invention provides a personalized paper composing method integrating learner cognition characteristics and test question text information, which combines a cognition diagnosis model, and performs learner test paper composing exercise based on collaborative filtering of the model and a test question text information extraction model. The invention analyzes knowledge point mastering conditions of learners so as to obtain learning states of the learners, and the used cognitive diagnosis model is a widely used cognitive diagnosis DINA model. In the invention, a cyclic neural network model RNN is used for extracting the text content of the test questions in the aspect of test question text analysis, and the mapping relation between the learning state of the learner and the test question text information is constructed through a full connection layer. The invention takes the result of the model as prior information and is used for training with other information by a collaborative filtering method based on probability matrix decomposition, so that the result of the group paper has the learning state of learners, text information of test questions and commonality among learners.
Compared with the traditional method, the personalized paper composing method integrating the cognitive characteristics and the test question text information is compared with the traditional paper composing algorithm, and experimental results show that the method has larger performance improvement than the traditional method, and the defect of the traditional method in the aspect of test question recommendation is overcome. The invention utilizes richer information, gives more accurate personalized test question recommendation to the learner, and improves the learning efficiency of the learner.
The invention realizes the personalized paper making method of combining the cognitive characteristics of the learner and the test question text information, can simultaneously combine the cognitive diagnosis, the test question text information and the common characteristic information of the learner to make test question recommendation for the target learner, and obtains a more accurate personalized paper making method of combining the cognitive characteristics and the test question text information, thereby greatly increasing the autonomous learning efficiency of the learner and helping the learner to make up short plates in knowledge. The invention can be applied to the fields of intelligent education technology, education data mining and the like, can also provide effective support for subsequent education resource recommendation and the like, and helps an online education platform and a digital education platform to better predict the achievement of a learner, thereby efficiently finding out weak links of the knowledge points of the learner and providing accurate remedial measures.
The invention compares the personalized paper grouping method integrating the cognitive characteristics of the learner and the test question text information with other test question knowledge estimation methods, and the accuracy, recall rate and F1 value comparison results of the personalized paper grouping method integrating the cognitive characteristics of the learner and the test question text information with other methods under the same data set are shown in the table 1.
Table 1 comparison of experimental results
From the experimental results, it can be seen that: the personalized paper grouping method for combining the cognitive characteristics of the learner and the text information of the test questions combines the cognitive characteristics of the learner, the learning commonality of the learner and the text information of the test questions. The accuracy of the winding result is obviously superior to other comparison experiments. Therefore, experiments show that the personalized paper grouping method for fusing the cognitive characteristics of the learner and the text information of the test questions is more effective than other methods in terms of accuracy, recall ratio, F1 value and the like.
Meanwhile, the analysis shows that the assembly method and the probability matrix decomposition method based on the DINA are slightly unstable, and the accuracy rate of the assembly method and the probability matrix decomposition method can be reduced along with the increase of the number of the assembly test questions. Conventional probability matrix decomposition is easy to implement, but the potential information in the extracted data set is insufficient, which results in generally low accuracy of probability matrix decomposition, especially in the face of large amounts of training data. In a word, the personalized test paper combining method for combining the cognitive characteristics of the learner and the text information of the test questions has the best experimental effect, and can combine the test questions with different difficulty levels according to the test difficulty and the examination target requirement and promote the cognitive growth of the learner.
According to the method, the test question text information is introduced to serve as an important measurement index for teaching test question recommendation, so that key information in the test question text can be utilized by the method provided by the invention.
According to the invention, test question text information and a cognitive diagnosis technology are fused, a collaborative filtering method is adopted, the learning condition of a learner obtained from the test question text information and the cognitive diagnosis is fused into an objective function with collaborative filtering optimization, and the relation between the three is introduced for adjusting parameters, so that an optimal group paper model matched with the current data set is obtained.
In summary, the personalized paper making method integrating the cognitive characteristics of the learner and the text information of the test questions realizes more accurate learner test question recommendation and personalized paper making, and combines three aspects of cognitive diagnosis, test question text information and learner learning commonality information to make test question recommendation for a target learner. The method can be applied to the fields of education resource recommendation and evaluation, education data mining and the like, thereby providing effective support for subsequent education resource recommendation and the like, helping an online education platform, better predicting the achievement of a learner by a digital education platform and providing a personalized grouping scheme, and efficiently diagnosing weak links of the knowledge points of the learner and providing accurate remedial measures.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments of the present application will be briefly described below, and it is obvious that the drawings described below are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of a personalized grouping method for fusing learner cognitive characteristics and test question text information according to an embodiment of the present application.
Fig. 2 is a flowchart of a personalized grouping method for merging learning cognitive characteristics and test question text information provided by an embodiment of the application.
FIG. 3 is a schematic diagram of a personalized group paper system for fusing learner cognitive characteristics and test question text information provided by the embodiment of the application;
in the figure: 1. a score prediction module based on cognitive level; 2. a score prediction module based on the text information; 3. and the topic selection strategy module.
Detailed Description
The present application will be described in further detail with reference to the following examples in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
Aiming at the problems existing in the prior art, the invention provides a personalized paper grouping method for fusing the cognitive characteristics of learners and the text information of test questions, and the invention is described in detail below with reference to the accompanying drawings.
The symbols related to the invention are as follows:
as shown in fig. 1-2, the personalized paper grouping method for fusing the cognitive characteristics of learners and the text information of test questions provided by the embodiment of the invention comprises the following steps:
s101, estimating and calculating the knowledge mastering condition of a learner by using a cognitive diagnosis model according to the real answering condition of the learner and the distribution of the knowledge points of the test questions, and predicting the score of the learner on a specific test question based on the cognitive level;
s102, extracting text contents of test questions by using a cyclic neural network model, constructing a mapping relation between a learning state of a learner and text information of the test questions through a full-connection layer, and predicting scores of the learner on specific test questions based on the text information;
s103, constructing a probability matrix decomposition objective function based on the obtained learner based on the cognitive level and the prediction score based on the text information, and predicting the potential score of the learner on the specific test question;
s104, calculating KL divergence by using the estimated learner knowledge mastering vector and the learner increment knowledge mastering vector, and selecting test papers which are provided with increased learner knowledge mastering trend and proper difficulty and are formed into personalized tests by combining potential scores of learners on the test questions.
In step S101, according to the actual answer situation of the learner and the distribution of the test question knowledge points provided by the embodiment of the present invention, using the cognitive diagnosis model to estimate and calculate the knowledge grasping situation of the learner, and predicting the score of the learner based on the cognitive level on the specific test question includes:
(1.1) collecting the distribution of the test question knowledge points marked by the field expert and the response data of the learner; according to the distribution of the test question knowledge points marked by the field expert, calculating a Q matrix of the test question knowledge points required to be used in the cognitive diagnosis model;
(1.2) according to the learning prior knowledge point mastering mode eta, calculating the ideal answering situation of the learner:
wherein pi ij Represents the ideal answer condition of learner i on the jth test question, eta ik Representing the knowledge point k of learner i, q jk Indicating whether the known j-th test question examines the knowledge point k;
(1.3) estimating the probability s of a learner doing wrong test questions under the condition that the learner grasps all knowledge points of a test question and the probability g of the learner doing the test questions under the condition that the learner does not grasp all corresponding knowledge points according to a expectation maximization algorithm;
(1.4) calculating the probability of correct answer of the learner according to the estimated ideal answer situation of the learner, the probability s of the learner doing wrong test questions when the learner grasps some test questions to examine all knowledge points and the probability g of the learner doing the test questions when the learner does not grasp the corresponding knowledge points:
(1.5) obtaining a total likelihood function of the DINA model:
wherein l=2 K ；
(1.6) calculating knowledge mastery of the learner using maximum likelihood estimation based on the obtained total likelihood function:
(1.7) calculating the score condition of the learner on the new test question according to the knowledge mastery condition of the learner:
in step S102, the method for extracting text content of a test question by using the recurrent neural network model provided by the embodiment of the present invention, and constructing a mapping relationship between a learning state of a learner and text information of the test question through a full connection layer, and predicting a score of the learner on a specific test question based on the text information includes:
(2.1) performing text word segmentation, removal of stop words and other preprocessing on test question texts; processing test questions by using a continuous Word bag model of a Word vector model Word2vec, predicting Word vectors of target words according to Word vectors of a plurality of words in the context of the target words, vectorizing the test questions input after preprocessing, and obtaining embedded expression of test question Word levels;
(2.2) acquiring test question context characteristic representation by using a test question word vector as input and adopting a bidirectional long-short-time memory neural network, and acquiring test question sentence level representation by means of mean value pooling processing to acquire test question word embedded representation;
(2.3) constructing a full-connection depth neural network, performing feature fusion on the test question word embedding vector and the learner knowledge grasping vector to serve as input of the full-connection depth neural network, and outputting the score of a specific student on the test question;
(2.4) the result of the full connection layer (y 1 ,y 2 ,...,y n ) The processing is carried out by an output unit:
obtaining a value between 0 and 1, representing the probability of the student to answer the test question text correctly, comparing the probability with real score data, and carrying out weight correction in a network;
and (2.5) training the network model by using the training set data to obtain a trained neural network model, and predicting the answering results of the learner based on the text information by using the obtained trained neural network model.
In step (2.1), the text word segmentation of the test question text provided by the embodiment of the invention comprises the following steps: based on the mixed dictionary, a method combining a bidirectional maximum matching method and statistics is adopted to carry out mixed word segmentation on the test question text.
In step (2.1), the method for removing the stop words from the test question text provided by the embodiment of the invention comprises the following steps: and adding words which are irrelevant to sentences and test question text topics, do not contribute to test question labeling tasks and have too low frequency in the test question text into the stop word stock, and deleting words of mixed word segmentation in the test question text, wherein the words appear in the stop word stock.
In step (2.2), the method for acquiring the context characteristic representation of the test question by adopting the two-way long and short-term memory neural network provided by the embodiment of the invention comprises the following steps:
the BiLSTM network adopts two LSTM to obtain the contextual characteristics of different test questions from opposite directions, and the formula is as follows:
wherein a is 1 ,a 2 ,b 1 And b 2 G (·) is the hidden layer activation function for the weight coefficient,is the forward hidden layer output at the time t,outputting the backward hidden layer at the time t; and fusion of hidden layer output of two directions at each moment to construct final output h t ：
Wherein c 1 And c 2 And f (·) is the output activation function for the weight coefficient.
In step (2.2), the obtaining the sentence-level representation of the test question through the mean value pooling processing provided by the embodiment of the invention comprises the following steps:
at the sentence-level representation layer of the test questions, the test question word embedded representation E is obtained through average pooling h ：
E h ＝mean(h 1 ,...,h t )
Where mean (·) is the average pooling operation, i.e., taking the average of the eigenvalues as output in the domain.
In step (2.3), the fully-connected deep neural network provided by the embodiment of the invention comprises:
in the fully-connected deep neural network, the calculation method of the nth node value of the mth layer comprises the following steps:
where N is the number of units of the upper layer,representing the weight coefficients from the ith cell of the m-1 th layer to the nth cell of the m-th layer;
And obtaining a potential mapping relation between learner knowledge mastery, test question text information and scores by adopting a relu activation function.
In step S103, the method according to the embodiment of the present invention constructs a probability matrix decomposition objective function based on the obtained learner prediction score based on the cognitive level and the text information, and predicts the potential score of the learner on the specific test question, where the prediction score includes:
(3.1) marking the obtained score prediction based on cognitive level and score prediction learner based on text information as R 1 ,R 2 Using a probability matrix decomposition algorithm to construct a potential answer representation of the learner on the test question:
wherein U, V are respectively a learner characteristic matrix and a achievement characteristic matrix in the probability matrix decomposition, and alpha and beta are respectively adjusting parameters of learner learning conditions and test question text information;
(3.2) constructing a final objective function of the probability matrix decomposition:
and (3.3) optimizing an objective function by using a gradient descent method to obtain an optimal learner performance feature matrix U, V:
(3.4) predicting the performance of the learner by using the optimal feature matrix U and V obtained by training:
in step S104, the method for selecting test paper for forming personalized test by using the estimated learning knowledge mastering vector and the learning incremental knowledge mastering vector, calculating KL divergence, combining potential scores of learners on test questions, and selecting test paper with increased learning knowledge mastering trend and proper difficulty comprises:
(4.1) obtaining learner knowledge mastering vector based on the analysis to obtain all incremental knowledge mastering vectors of the learner0≤d≤D；
(4.2) computing the learner knowledge mastery vector eta estimated by the cognitive diagnostic model i Incremental knowledge mastering with all learnersKL divergence measure of (2):
(4.3) selecting test paper which is used for personalized test and is composed of test questions with proper difficulty, wherein the learning knowledge mastering trend is increased:
as shown in fig. 3, the personalized group paper system for fusing cognitive characteristics of learners and text information of test questions provided by the embodiment of the invention includes:
the score prediction module 1 based on the cognitive level is used for estimating and calculating the knowledge mastering condition of the learner by utilizing a cognitive diagnosis model according to the real answering condition of the learner and the knowledge point distribution of the test questions and predicting the score of the learner based on the cognitive level on the specific test questions;
the score prediction module 2 based on text information is used for extracting text contents of test questions by using a cyclic neural network model, constructing a mapping relation between a learning state of a learner and the text information of the test questions through a full-connection layer, and predicting scores of the learner on specific test questions based on the text information;
the topic selection strategy module 3 is used for constructing a probability matrix decomposition objective function based on the obtained learner prediction scores based on the cognitive level and the text information and predicting potential scores of the learner on specific topics; and calculating KL divergence by using the estimated learner knowledge mastering vector and the learner incremental knowledge mastering vector, and selecting test papers which are provided with increased learner knowledge mastering trend and proper difficulty and are formed into personalized tests by combining potential scores of learners on the test questions.
The technical effects of the present invention will be further described with reference to specific examples.
Example 1:
the invention discloses a personalized paper grouping method integrating learner cognition characteristics and test question text information, which comprises the following steps:
step one, according to the actual answering situation of the learner and the distribution of the test question knowledge points, learning states such as knowledge mastering situations of the learner are mined by utilizing a cognitive diagnosis model.
And secondly, extracting text content of the test questions by using a cyclic neural network model, and constructing a mapping relation between a learning state of a learner and text information of the test questions through a full connection layer.
And thirdly, constructing a probability matrix decomposition objective function by integrating the learning state of the learner, the text information of the test questions and the cognitive commonality of the learner, and mining potential scores of the learner on the specific test questions.
And step four, calculating KL divergence by using the estimated learner knowledge mastering vector and the learner increment knowledge mastering vector, and selecting test papers which are used for personalized test and are formed by combining potential scores of learners on test questions and test questions with proper difficulty, wherein the learning knowledge mastering trend is increased.
Further, the first step includes:
step a): collecting the distribution of the test question knowledge points marked by the field expert and the answering data of the learner;
Step b): according to the distribution of the test question knowledge points marked by the field expert, calculating a test question knowledge point representation Q matrix required to be used in the cognitive diagnosis model;
step c): according to the learning person prior knowledge point mastering mode eta, calculating the ideal answering condition of the learning person:
π ij represents the ideal answer condition of learner i on the jth test question, eta ik Representing the knowledge point k of learner i, q jk Indicating whether the known j-th test question examines the knowledge point k;
step d): estimating the probability s of a learner doing wrong test questions under the condition that the learner grasps a certain test question and examines all knowledge points according to an expectation maximization algorithm, and the probability g of the learner doing wrong test questions under the condition that the learner does not grasp the corresponding knowledge points;
step e): according to the estimated ideal answer situation of the learner, the probability s of the learner doing wrong test questions when the learner grasps some test questions to examine all knowledge points, and the probability g of the learner doing wrong test questions when the learner does not grasp the corresponding knowledge points, the probability of the learner doing correct answer is calculated:
step f): the overall likelihood function of the DINA model is thus obtained:
wherein l=2 K Since the formula contains the hidden variable eta l Maximum likelihood estimation cannot be directly performed, so the method of expectation maximization is adopted to solve:
Step g): e, utilizing the s obtained in the previous round j And g is equal to j Estimate the calculation matrix P (r|η) = [ P (R) i |η l )] I×L And calculates a matrix P (η|r) = [ P (η) using P (r|α) l |R i )] L×I Is a value of (2).
Step h): m steps: respectively orderAnd->The method can obtain: />Wherein->Indicating the desire of the number of people lacking at least one knowledge point required for the j-th question among the learners belonging to the first knowledge point mastering mode,/for the learner>Representation->The number of people who answer the correct j-th question expects, < +.>And->Meaning of->And->Similarly, the difference is +.>And->Is a desire in the case where the learner grasps knowledge points required for all j-th questions. Therefore, from the estimate obtained in step E, the +.>And->And thus a new s j And g is equal to j And (5) estimating.
Step i): using the total likelihood function, the learner's knowledge mastery is calculated using maximum likelihood estimates:
step j): according to the knowledge mastering condition of the learner, calculating the scoring condition of the learner on the new test question:
further, the second step comprises:
step 1): preprocessing test question texts, wherein the preprocessing mainly comprises text word segmentation and stop removal;
step 2): and carrying out text word segmentation on the test questions. Based on a mixed dictionary, adopting a method combining a bidirectional maximum matching method and statistics to perform mixed word segmentation on test question texts;
Step 3): and stopping words aiming at the mixed word segmentation result. Words which do not contribute to the task of labeling the test questions are removed irrespective of sentences and the text subjects of the test questions, and words with too low frequency do not contribute to the task of labeling the test questions, so that the words with too low frequency are treated as stop words. According to the two rules, a stop word stock is established, words appearing in the stop word stock are deleted, and words with too low frequency are deleted;
step 4): and processing the test questions by using a continuous Word bag (Continuous Bag Of Words, CBOW) model of a Word vector model Word2vec, vectorizing the test questions input after preprocessing, and obtaining the embedded representation of the test question Word level. The CBOW model predicts word vectors of the target words according to word vectors of a plurality of words in the context of the target words, so that the test questions are vectorized;
step 5): the method comprises the steps of using a test question word vector as input, firstly obtaining a test question context characteristic representation by using a bidirectional long short-time Memory neural network (Bidirectional Long Short-Term Memory, biLSTM), and then obtaining a test question sentence level representation by introducing a mean value pooling operation, so as to obtain a test question word embedded representation.
Step 6): the BiLSTM network uses two LSTMs to obtain the contextual characteristics of different questions from opposite directions, whose calculations are defined as:
Wherein a is 1 ,a 2 ,b 1 And b 2 G (·) is the hidden layer activation function for the weight coefficient,is the forward hidden layer output at the time t,for the backward hidden layer output at the time t, finally fusing the hidden layer outputs of the two directions at each time to construct a final output h t ：/>
Wherein c 1 And c 2 F (·) is the weight coefficient, f (·) is the output activation function;
step 7): at the sentence-level representation layer of the test questions, the test question word embedded representation E is obtained through average pooling h ：
E h ＝mean(h 1 ,...,h t )
The mean (·) is an average pooling operation, that is, the average of the feature values is taken as output in the field, so that the representative information in the whole window information can be obtained, and the feature dimension of the test question text and the number of model network parameters are reduced.
Step 8): and constructing a fully-connected deep neural network, carrying out feature fusion on the test question word embedded vector and the learner knowledge mastering vector to serve as input of the network, and outputting the score of a specific student on the test question.
Step 9): in the fully-connected deep neural network, the calculation method of the nth node value of the mth layer comprises the following steps:
where N is the number of units of the upper layer,representing the weight coefficients from the ith cell of the m-1 th layer to the nth cell of the m-th layer;
obtaining potential mapping relations between learner knowledge mastery, test question text information and scores by adopting a relu activation function;
Step 10): results of the full connection layer (y 1 ,y 2 ,...,y n ) The processing is carried out by an output unit:
obtaining a value between 0 and 1, representing the probability of the student to answer the test question text correctly, and comparing the probability with real score data, thereby realizing weight correction in a network;
step 11): after training the training set data, a trained neural network model can be obtained, so that the answering result of the learner based on the text information can be predicted.
Further, the third step includes:
step A): the learning person based on the score prediction of the cognitive level and the score prediction based on the text information obtained by the calculation in the second step and the third step are respectively marked as R 1 ,R 2 Using a probability matrix decomposition algorithm to construct a potential answer representation of the learner on the test question:
wherein U, V are respectively a learner characteristic matrix and a achievement characteristic matrix in the probability matrix decomposition, and alpha and beta are respectively adjusting parameters of learner learning conditions and test question text information;
step B): deducing a final objective function of probability matrix decomposition according to a score calculation formula integrated with the learning condition of the learner, the text information of the test questions and the commonality of the learner:
step C): optimizing an objective function by using a gradient descent method, and firstly, respectively solving partial derivatives of two feature matrixes U and V according to the objective function:
/>
And (3) respectively enabling the partial derivative to be 0, obtaining a recursive formula iterative calculation of the method until the result converges or reaches the maximum iterative times, and finally obtaining the optimal learner achievement feature matrix U, V:
step D): finally, obtaining a result prediction result of the learner by using the optimal feature matrix U and V obtained by training:
step E): and adjusting super parameters in the experiment and adjusting parameters alpha and beta of learner learning conditions and test question text information to obtain parameters most suitable for the data set, thereby obtaining a final training model.
Further, the fourth step includes:
step I): the learner knowledge mastery vector obtained by analysis is marked as eta i From eta i Obtaining all increment knowledge mastering vectors of learners0≤d≤D；
Step II): learner knowledge mastering vector eta estimated by calculating cognitive diagnosis model i Incremental knowledge mastering with all learnersKL divergence measure of (2):
step III): therefore, the test paper for personalized test is formed by selecting the test questions with proper difficulty, which increases the knowledge mastering trend of the learner, so that a recommendation result which has the learning condition of the learner, the text information of the test questions and the commonality among the learners is obtained.
The invention compares the personalized grouping method integrating the cognitive characteristics of learners and the text information of test questions with other test question knowledge estimation methods, and the calculation method comprises the following steps:
/>
Wherein L (i) represents personalized learning test questions formulated for the ith learner, M (i) represents test questions matched with the learner in the question bank, and L (i). AndM (i) represents an intersection of the two. The accuracy precision@k indicates the probability of recommending correctness in the recommendation result, the recall ratio recall@k is also called recall ratio, the degree of correct recommendation in the recommendation result matching question bank is indicated, and the accuracy and the recall ratio are in a certain contradiction amount, namely the recall ratio is lower when the accuracy is higher. In order to conveniently display experimental results, the traditional cognitive diagnosis method is recorded as DINA, and the traditional collaborative filtering method is recorded as PMF.
The accuracy, recall rate and F1 value of the personalized paper grouping method integrating the cognitive characteristics of the learner and the text information of the test questions under the same data set are shown in the table 1.
Table 1 comparison of experimental results
From the experimental results, it can be seen that: the personalized paper grouping method for combining the cognitive characteristics of the learner and the text information of the test questions combines the cognitive characteristics of the learner, the learning commonality of the learner and the text information of the test questions. The accuracy of the winding result is obviously superior to other comparison experiments. Therefore, experiments show that the personalized paper grouping method for fusing the cognitive characteristics of the learner and the text information of the test questions is more effective than other methods in terms of accuracy, recall ratio, F1 value and the like.
Meanwhile, the analysis shows that the group paper model and the probability matrix decomposition method based on the DINA are slightly unstable, and the accuracy rate of the group paper model and the probability matrix decomposition method can be reduced along with the increase of the number of the group paper test questions. Conventional probability matrix decomposition is easy to implement, but the potential information in the extracted data set is insufficient, which results in generally low accuracy of probability matrix decomposition, especially in the face of large amounts of training data. In a word, the personalized test paper combining method for combining the cognitive characteristics of the learner and the text information of the test questions has the best experimental effect, and can combine the test questions with different difficulty levels according to the test difficulty and the examination target requirement and promote the cognitive growth of the learner.
The invention introduces test question text information as an important measurement index of the personalized group paper, so that key information in the test question text can be utilized by the method provided by the invention.
According to the invention, test question text information and a cognitive diagnosis technology are fused, a collaborative filtering method is adopted, the learning condition of a learner obtained from the test question text information and the cognitive diagnosis is fused into an objective function with collaborative filtering optimization, and the relation between the three is introduced for adjusting parameters, so that an optimal group paper model matched with the current data set is obtained.
In summary, the personalized paper making method integrating the cognitive characteristics of the learner and the test question text information realizes a more accurate paper making method, and the method combines three aspects of cognitive diagnosis and test question text information, and learner learning commonality information to make paper making strategies for target learners, so that the paper making results can be customized according to test targets and test question difficulty, the self-learning efficiency of the learners is greatly improved, and the learner is helped to make up short plates on the knowledge of the learner in a class more quickly. The method can be applied to the fields of education resource recommendation and evaluation, education data mining and the like, thereby providing effective support for subsequent education resource recommendation and the like, helping an online education platform, better predicting the achievement of a learner by a digital education platform and providing a personalized grouping scheme, and efficiently diagnosing weak links of the knowledge points of the learner and providing accurate remedial measures.
The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention will be apparent to those skilled in the art within the scope of the present invention.

Claims (9)

1. The personalized paper system for combining the cognitive characteristics of the learner and the text information of the test questions is characterized by comprising an application information data processing terminal, wherein the personalized paper system for combining the cognitive characteristics of the learner and the text information of the test questions comprises the following components:
the score prediction module based on the cognitive level is used for estimating and calculating the knowledge mastering condition of the learner by utilizing a cognitive diagnosis model according to the real answering condition of the learner and the knowledge point distribution of the test questions and predicting the score of the learner based on the cognitive level on the test questions;
the score prediction module based on the text information is used for extracting the text content of the test questions by using the cyclic neural network model, constructing the mapping relation between the learning state of the learner and the text information of the test questions through the full-connection layer, and predicting the score of the learner on the test questions based on the text information;
the topic selection strategy module is used for constructing a probability matrix decomposition objective function based on the obtained learner prediction scores based on the cognitive level and the text information and predicting potential scores of the learner on the test questions; calculating KL divergence by using the estimated learner knowledge mastering vector and the learner incremental knowledge mastering vector, and selecting test questions with increased learner knowledge mastering trend to form a test paper for personalized test by combining potential scores of the learner on the test questions;
The method comprises the steps of constructing a probability matrix decomposition objective function based on the obtained learner prediction scores based on the cognitive level and the text information, and predicting potential scores of the learner on the test questions comprises:
(3.1) recording the obtained score prediction based on the cognitive level and the score prediction based on the text information asUsing a probability matrix decomposition algorithm to construct a potential answer representation of the learner on the test question:
wherein U is i ，V j Respectively a learner characteristic matrix and a achievement characteristic matrix in probability matrix decomposition, wherein alpha and beta are respectively adjusting parameters of learner learning conditions and test question text information;
(3.2) constructing a final objective function of the probability matrix decomposition:
(3.3) optimizing the objective function by using the gradient descent method to obtain the learner characteristic matrix U i Sum score feature matrix V j ：
(3.4) utilizing learner characteristic matrix U i Sum score feature matrix V j Predicting the achievement of learners:
the method for selecting test papers for personalized test by combining potential scores of learners on test questions by using estimated learner knowledge mastering vectors and learner increment knowledge mastering vectors comprises the following steps:
(4.1) obtaining learner knowledge mastering vector based on the analysis to obtain all incremental knowledge mastering vectors of the learner
(4.2) computing the learner knowledge mastery vector eta estimated by the cognitive diagnostic model i Grasping vectors with all incremental knowledgeKL divergence measure of (2):
(4.3) selecting test questions with increased knowledge mastering trend of the learner to form test papers for personalized test:
2. the personalized paper grouping method for fusing the cognitive characteristics of the learner and the text information of the test questions is characterized by being applied to an information data processing terminal, and comprises the following steps of:
according to the real answering situation of the learner and the distribution of the knowledge points of the test questions, estimating and calculating the knowledge grasping situation of the learner by using a cognitive diagnosis model, and predicting the score of the learner on the test questions based on the cognitive level;
extracting text content of the test questions by using a cyclic neural network model, constructing a mapping relation between a learning state of a learner and text information of the test questions by using a full-connection layer, and predicting scores of the learner on the test questions based on the text information;
constructing a probability matrix decomposition objective function based on the obtained learner based on the cognitive level and the prediction score based on the text information, and predicting the potential score of the learner on the test question;
calculating KL divergence by using the estimated learner knowledge mastering vector and the learner incremental knowledge mastering vector, and selecting test questions with increased learner knowledge mastering trend to form a test paper for personalized test by combining potential scores of the learner on the test questions;
The method comprises the steps of constructing a probability matrix decomposition objective function based on the obtained learner prediction scores based on the cognitive level and the text information, and predicting potential scores of the learner on the test questions comprises:
(3.1) recording the obtained score prediction based on the cognitive level and the score prediction learner based on the text information asUsing a probability matrix decomposition algorithm to construct a potential answer representation of the learner on the test question:
wherein U is i ，V j Learner characteristic matrix and achievement characteristic matrix in probability matrix decomposition respectively, and alpha and beta are learning respectivelyThe user learns the condition and the adjustment parameters of the test question text information;
(3.2) constructing a final objective function of the probability matrix decomposition:
(3.3) optimizing the objective function by using the gradient descent method to obtain the learner characteristic matrix U i Sum score feature matrix V j ：
(3.4) utilizing learner characteristic matrix U i Sum score feature matrix V j Predicting the achievement of learners:
the method for selecting test papers for personalized test by combining potential scores of learners on test questions by using estimated learner knowledge mastering vectors and learner increment knowledge mastering vectors comprises the following steps:
(4.1) obtaining learner knowledge mastering vector based on the analysis to obtain all incremental knowledge mastering vectors of the learner
(4.2) computing the learner knowledge mastery vector eta estimated by the cognitive diagnostic model i Grasping vectors with all incremental knowledgeKL divergence measure of (2):
(4.3) selecting test questions with increased knowledge mastering trend of the learner to form test papers for personalized test:
3. the personalized paper grouping method for merging learning cognitive characteristics and test question text information according to claim 2, wherein the estimating and calculating the learning knowledge of the learner by using the cognitive diagnosis model according to the real answer situation of the learner and the test question knowledge point distribution, and predicting the learning level-based score of the learner on the test question comprises:
(1.1) collecting the distribution of the test question knowledge points marked by the field expert and the response data of the learner; according to the distribution of the test question knowledge points marked by the field expert, calculating a Q matrix of the test question knowledge points required to be used in the cognitive diagnosis model;
(1.2) according to the learning prior knowledge point mastering mode eta, calculating the ideal answering situation of the learner:
wherein pi ij Represents the ideal answer condition of learner i on the jth test question, eta ik Representing the knowledge point k of learner i, q jk Indicating whether the known j-th test question examines the knowledge point k;
(1.3) estimating probability s of a learner doing wrong test questions under the condition of grasping all knowledge points of a test question according to a expectation maximization algorithm j And the probability g of the learner doing the test questions without fully mastering the corresponding knowledge points j ；
(1.4) according to the estimated ideal answer situation of the learner and the probability s of the learner doing wrong test questions under the condition of grasping all knowledge points of the test questions j And the probability g of the learner doing the test questions without fully mastering the corresponding knowledge points j The probability of correct response of the learner is calculated:
(1.5) obtaining a total likelihood function of the DINA model:
wherein l=2 K ；
(1.6) calculating knowledge mastery of the learner using maximum likelihood estimation based on the obtained total likelihood function:
(1.7) calculating the score condition of the learner on the new test question according to the knowledge mastery condition of the learner:
4. the personalized paper grouping method for merging cognitive characteristics of a learner and text information of a test question according to claim 2, wherein the text content of the test question is extracted by using a cyclic neural network model, a mapping relation between a learning state of the learner and the text information of the test question is constructed through a full connection layer, and the predicting of the score of the learner based on the text information on the test question comprises:
(2.1) performing text word segmentation, removal of stop words and other preprocessing on test question texts; processing test questions by using a continuous Word bag model of a Word vector model Word2vec, predicting Word vectors of target words according to Word vectors of a plurality of words in the context of the target words, vectorizing the test questions input after preprocessing, and obtaining embedded expression of test question Word levels;
(2.2) acquiring test question context characteristic representation by using a test question word vector as input and adopting a bidirectional long-short-time memory neural network, and acquiring test question sentence level representation by means of mean value pooling processing to acquire test question word embedded representation;
(2.3) constructing a full-connection depth neural network, performing feature fusion on the test question word embedding vector and the learner knowledge grasping vector to serve as input of the full-connection depth neural network, and outputting the score of the student on the test question;
(2.4) the result of the full connection layer (y 1 ,y 2 ,...,y n ) The processing is carried out by an output unit:
obtaining a value between 0 and 1, representing the probability of the student to answer the test question text correctly, comparing the probability with real score data, and carrying out weight correction in a network;
and (2.5) training the network model by using the training set data to obtain a trained neural network model, and predicting the answering results of the learner based on the text information by using the obtained trained neural network model.
5. The personalized composition method for merging learning cognitive characteristics and test question text information as set forth in claim 4, wherein in step (2.1), the text segmentation of the test question text comprises: based on a mixed dictionary, adopting a method combining a bidirectional maximum matching method and statistics to perform mixed word segmentation on test question texts;
in the step (2.1), the step of removing the stop words from the test question text comprises the following steps: adding words which are irrelevant to sentences and test question text topics, do not contribute to test question labeling tasks and have too low frequency in the test question text into a stop word stock, and deleting words which appear in the stop word stock in mixed word segmentation in the test question text;
in the step (2.2), the step of acquiring the context characteristic representation of the test question by adopting the bidirectional long-short-time memory neural network comprises the following steps:
the BiLSTM network adopts two LSTM to obtain the contextual characteristics of different test questions from opposite directions, and the formula is as follows:
wherein a is 1 ,a 2 ,b 1 And b 2 G (·) is the hidden layer activation function for the weight coefficient,for the forward hidden layer output at time t, < ->Outputting the backward hidden layer at the time t; and fusion of hidden layer output of two directions at each moment to construct final output h t ：
Wherein c 1 And c 2 And f (·) is the output activation function for the weight coefficient.
6. The personalized composition method for merging learner cognitive characteristics and test question text information according to claim 4, wherein in the step (2.2), the obtaining the test question sentence level representation through the mean pooling process comprises:
at the sentence-level representation layer of the test questions, the test question word embedded representation E is obtained through average pooling h ：
E h ＝mean(h 1 ,...,h t )；
Wherein mean (·) is an average pooling operation, i.e. taking the average of the eigenvalues in the domain as output;
in step (2.3), the fully connected deep neural network comprises:
in the fully-connected deep neural network, the calculation method of the nth node value of the mth layer comprises the following steps:
where N is the number of units of the upper layer,representing the weight coefficients from the ith cell of the m-1 th layer to the nth cell of the m-th layer;
and obtaining a potential mapping relation between learner knowledge mastery, test question text information and scores by adopting a relu activation function.
7. A computer device, characterized in that the computer device comprises a memory and a processor, the memory stores a computer program, and the computer program when executed by the processor causes the processor to execute the personalized paper grouping method for fusing learner cognitive characteristics and test question text information according to any one of claims 2-6.
8. A computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to perform the personalized composition method of merging learner cognitive characteristics and test question text information according to any one of claims 2 to 6.
9. An information data processing terminal, characterized in that the information data processing terminal is used for realizing the personalized grouping method for fusing the cognitive characteristics of learners and the text information of test questions according to any one of claims 2-6.
CN202011233044.0A 2020-11-06 2020-11-06 Personalized paper grouping method and system integrating cognition characteristics and test question text information Active CN112508334B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011233044.0A CN112508334B (en) 2020-11-06 2020-11-06 Personalized paper grouping method and system integrating cognition characteristics and test question text information

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011233044.0A CN112508334B (en) 2020-11-06 2020-11-06 Personalized paper grouping method and system integrating cognition characteristics and test question text information

Publications (2)

Publication Number Publication Date
CN112508334A CN112508334A (en) 2021-03-16
CN112508334B true CN112508334B (en) 2023-09-01

Family

ID=74955388

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011233044.0A Active CN112508334B (en) 2020-11-06 2020-11-06 Personalized paper grouping method and system integrating cognition characteristics and test question text information

Country Status (1)

Country Link
CN (1) CN112508334B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113127734B (en) * 2021-03-30 2022-11-25 江苏乐易智慧科技有限公司 Personalized test paper pushing method based on genetic algorithm
CN113221007B (en) * 2021-05-21 2022-09-23 合肥工业大学 Method for recommending answering behavior
CN113377942A (en) * 2021-07-12 2021-09-10 北京乐学帮网络技术有限公司 Test paper generation method and device, computer equipment and storage medium
CN113569870B (en) * 2021-07-31 2022-06-07 西北工业大学 Cross-modal problem Q matrix automatic construction method based on heterogeneous graph neural network
CN113743083B (en) * 2021-09-06 2024-03-12 东北师范大学 Test question difficulty prediction method and system based on deep semantic characterization
CN114418443A (en) * 2022-01-28 2022-04-29 华南师范大学 Test paper quality detection method, system, device and storage medium
CN114238613B (en) * 2022-02-22 2022-05-27 北京一起航帆科技有限公司 Method and device for determining mastery degree of knowledge points and electronic equipment
CN116431800B (en) * 2023-03-03 2023-11-03 广州秒可科技有限公司 Examination interface generation method, device and readable storage medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2896037A1 (en) * 2014-07-03 2016-01-03 Mentorum Solutions Inc. Adaptive e-learning system and method
CN108711125A (en) * 2018-07-24 2018-10-26 电子科技大学 A kind of blocks of knowledge centrad and difficulty quantization method
CN110264091A (en) * 2019-06-24 2019-09-20 中国科学技术大学 Student's cognitive diagnosis method
CN110502636A (en) * 2019-08-27 2019-11-26 华中师范大学 A kind of joint modeling and method for digging and system towards subjective and objective examination question
CN110516116A (en) * 2019-08-27 2019-11-29 华中师范大学 A kind of the learner's human-subject test method for digging and system of multistep layering
CN111241243A (en) * 2020-01-13 2020-06-05 华中师范大学 Knowledge measurement-oriented test question, knowledge and capability tensor construction and labeling method
CN111445153A (en) * 2020-03-31 2020-07-24 华中师范大学 Objective test question attribute mode estimation and correction method and system for education measurement

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102233255B1 (en) * 2018-04-11 2021-03-29 한국전자통신연구원 Mehtod and apparatuse for learner diagnosis using reliability of cognitive diagnosis medel

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CA2896037A1 (en) * 2014-07-03 2016-01-03 Mentorum Solutions Inc. Adaptive e-learning system and method
CN108711125A (en) * 2018-07-24 2018-10-26 电子科技大学 A kind of blocks of knowledge centrad and difficulty quantization method
CN110264091A (en) * 2019-06-24 2019-09-20 中国科学技术大学 Student's cognitive diagnosis method
CN110502636A (en) * 2019-08-27 2019-11-26 华中师范大学 A kind of joint modeling and method for digging and system towards subjective and objective examination question
CN110516116A (en) * 2019-08-27 2019-11-29 华中师范大学 A kind of the learner's human-subject test method for digging and system of multistep layering
CN111241243A (en) * 2020-01-13 2020-06-05 华中师范大学 Knowledge measurement-oriented test question, knowledge and capability tensor construction and labeling method
CN111445153A (en) * 2020-03-31 2020-07-24 华中师范大学 Objective test question attribute mode estimation and correction method and system for education measurement

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于协同过滤和认知诊断的试题推荐方法;齐斌；邹红霞；王宇；李冀兴;《计算机软件及计算机应用》;235-240 *

Also Published As

Publication number Publication date
CN112508334A (en) 2021-03-16

Similar Documents

Publication Publication Date Title
CN112508334B (en) Personalized paper grouping method and system integrating cognition characteristics and test question text information
CN109919810B (en) Student modeling and personalized course recommendation method in online learning system
CN110413741B (en) Subjective question-oriented intelligent paper marking method
CN107230174B (en) Online interactive learning system and method based on network
CN111241243B (en) Test question, knowledge and capability tensor construction and labeling method oriented to knowledge measurement
CN107590127B (en) Automatic marking method and system for question bank knowledge points
CN110941723A (en) Method, system and storage medium for constructing knowledge graph
CN112257966B (en) Model processing method and device, electronic equipment and storage medium
CN109949637B (en) Automatic answering method and device for objective questions
CN114254208A (en) Identification method of weak knowledge points and planning method and device of learning path
CN109272160A (en) Score on Prediction system and prediction technique
CN111460101A (en) Knowledge point type identification method and device and processor
CN115455186A (en) Learning situation analysis method based on multiple models
CN114254127A (en) Student ability portrayal method and learning resource recommendation method and device
CN114429212A (en) Intelligent learning knowledge ability tracking method, electronic device and storage medium
CN114780723B (en) Portrayal generation method, system and medium based on guide network text classification
Bagaria et al. An intelligent system for evaluation of descriptive answers
Wang et al. Utilizing artificial intelligence to support analyzing self-regulated learning: A preliminary mixed-methods evaluation from a human-centered perspective
Gaheen et al. Automated students arabic essay scoring using trained neural network by e-jaya optimization to support personalized system of instruction
CN113283488A (en) Learning behavior-based cognitive diagnosis method and system
Arifin et al. Automatic essay scoring for Indonesian short answers using siamese Manhattan long short-term memory
CN115687657A (en) Knowledge graph-assisted test question recommendation method
CN114117033B (en) Knowledge tracking method and system
CN112785039B (en) Prediction method and related device for answer score rate of test questions
CN114925610A (en) Learner knowledge structure and level modeling method, system, equipment and terminal

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
GR01 Patent grant
GR01 Patent grant