WO2018079968A1

WO2018079968A1 - Method and apparatus for providing personalized educational content, and computer program

Info

Publication number: WO2018079968A1
Application number: PCT/KR2017/005916
Authority: WO
Inventors: 허재위; 차영민; 서창호; 이강욱
Original assignee: 주식회사 뤼이드; 한국과학기술원
Priority date: 2016-10-26
Filing date: 2017-06-07
Publication date: 2018-05-03
Also published as: KR101836206B1

Abstract

The present invention relates to a method by which an electronic device provides personalized educational content, comprising the steps of: a) collecting a user's question solving result data; b) estimating a comprehension level L of at least one concept for each user and a composition level R of the concept for each question by using a part of the result data; and c) estimating the user's comprehension level X of the questions, and estimating a probability P that the user solves the questions by using the comprehension level.

Description

Methods, devices, and computer programs that provide personalized educational content

The present invention relates to a method for analyzing learning data and providing user-tailored educational content. More specifically, the present invention is a method for estimating the probability of solving a problem that the user has not solved by analyzing the result data of the user's problem solving.

Until now, educational content has generally been packaged. For example, a collection of questions on paper contains at least 700 questions per book, and online or offline lessons are also sold at once, with a minimum of one to two months of study.

However, for educated students, there is a need for personalized content rather than packages because both individual vulnerable units and vulnerable problem types are different. It is because learning to pick only the types of vulnerable problems of vulnerable units is much more efficient than solving the entire 700 problems.

However, it is very difficult for students as trainees to identify their own weaknesses. Furthermore, in the traditional education industry such as academy and publishing house, it is not easy to provide optimized problems for individual students because they analyze students and problems based on subjective experience and intuition.

As such, in the conventional education environment, it is not easy for the trainee to provide personalized content that can produce the most efficient learning results, and students have a problem that they quickly lose their sense of achievement and interest in the packaged education content.

An object of the present invention is to solve the above problems. More specifically, the present invention provides a method and apparatus for analyzing a learning data, estimating a conceptual diagram of a problem, a user's understanding of each concept, and calculating a probability of a correct answer for a problem not solved by the user to provide a user-specific problem. It aims to provide.

According to an embodiment of the present disclosure, a method of providing personalized education content by an electronic device may include: a collecting the problem solving result data of a user; Estimating an understanding degree L of at least one concept for each user and a configuration diagram R of the concept for each problem by using a part of the result data; And estimating an understanding degree X of the user's understanding of the problem and estimating a probability P of the user correcting the problem using the understanding.

According to the present invention, the correlation between the user and the problem is analyzed by a coefficient matrix having a low correlation between the user and the problem, without having to define a separate concept for problem and / or analysis of the user, and estimated with a high confidence probability of the correct answer of the individual problem for the individual user. It is effective.

1 is a flowchart illustrating a method for recommending a user-customized problem according to an embodiment of the present invention.

2 is a graph showing the results of experimenting with the relationship between problem comprehension and problem correct probability with sufficiently large data.

3 is a view for explaining a method of estimating a user's conceptual understanding L, a conceptual configuration diagram R of a problem according to an embodiment of the present invention.

It is apparent that the present invention is not limited to the description of the embodiments described below, and various modifications may be made without departing from the technical gist of the present invention. In describing the embodiments, descriptions of technical contents which are widely known in the technical field to which the present invention belongs and are not directly related to the technical gist of the present invention will be omitted.

On the other hand, the same components in the accompanying drawings are represented by the same reference numerals. In the accompanying drawings, some components may be exaggerated, omitted, or schematically illustrated. This is to clarify the gist of the present invention by omitting unnecessary description that is not related to the gist of the present invention.

In order to provide a user-customized problem, conventionally, a method of evaluating a student's level and recommending a problem that is judged to be appropriate for the level of learning has been taken in consideration of whether the student has corrected or wrong a specific problem.

For example, if a student frequently misses the issue of "verb tense" in English, the student is vulnerable to the "verb tense", suggesting another problem that is the subject of the verb tense. .

This method requires manual definition of the subject concepts by experts to analyze problems and users.

According to the conventional method, the concepts constituting the subject are individually defined in advance by the expert, and the expert judges and tags the concept of each concept for the subject. After that, each user solves the tagged problems for a particular concept and the learner's ability is analyzed based on the information.

For example, if a user wrongs problem 1 and the preposition and the same name are tagged in the first problem, the user analyzes that the understanding of the preposition and the same name is poor.

However, the above method has a problem that the tag information depends on the subjectivity of the person. The reliability of the result data is not high because tag information generated mathematically without human involvement is not mathematically assigned to the problem.

In order to solve the problem, it is necessary to mathematically model the relationship between the user and the problem through analysis of the learning result data.

In order to accomplish this, the present invention mathematically links the correlation between the user and the problem through a low coefficient matrix, analyzes the learning result data, updates the correlation, and calculates the probability of a correct answer for the problem that the user has not yet solved. An algorithm is proposed.

1 is a flowchart illustrating a method of recommending a user-customized problem by a data analysis system according to an exemplary embodiment of the present invention.

In operation 110, the service server may collect multiple-choice problem solving result data of the user.

The problem solving result data may be collected by a service server constructing a database of various problems on the market and collecting a result of the user solving the problems. The problem database includes listening assessment questions and may be in the form of text, images, audio, and / or video.

The service server according to the embodiment of the present invention may configure the user's multiple-choice problem solving result data set Y in the form of a list of users, problems, and results. For example, Y (i, j) means the result of solving the problem j by the user i, according to an embodiment of the present invention can be given a value of 1 for the correct answer, and 0 for the wrong answer.

According to an embodiment of the present invention, if Y is a sufficient size, the concept understanding L and the problem configuration R of the user may be estimated from Y without defining the concept separately. (Step 120)

L (i, j) may mean an understanding of concept j of user i, and R (i, j) may mean the degree to which problem i includes concept j. At this time, according to an embodiment of the present invention, each element of L has a value between 0 and 1, and the sum of the elements of each row of R may be 1.

For example, if the value of the first row of L is [0, 0, 1, 0.5, 1], this means that user 1 does not understand the first and second concepts at all, and the third and fifth concepts are fully understood. And the fourth concept would be interpreted as half understood.

For example, if the value of the first row of R is [0, 0.2, 0.5, 0.3, 0], problem 1 contains about 20% of concept 2, concept 3 contains about 50%, and question 4 The concept will be interpreted as containing about 30%.

According to an embodiment of the present invention, a more detailed method of estimating the conceptual understanding degree L of the user from the result data and the conceptual inclusion degree R of the problem will be described later with reference to FIG. 3.

On the other hand, the data analysis system according to the embodiment of the present invention, if L and R are estimated with sufficient reliability, can mathematically connect the correlation between the user and the problem through a low coefficient matrix.

For example, if the total number of users to be analyzed is n, and the total number of questions to be analyzed is m, the service server defines the conceptual understanding matrix L of the user as an n by r matrix, and the matrix R for the degree of inclusion of each problem in the concept. Can be defined by the m by r matrix. In this case, if L is connected to R's transpose matrix R ^T , we can analyze the relationship between the user and the problem without defining the concept r separately.

In other words, the problem understanding matrix X for each user is the product of the transpose matrix of L and R (

Can be expressed as (Step 130)

With this applied, in the example above where the value of the first row of L is [0, 0, 1, 0.5, 1], and the value of the first row of R is [0, 0.2, 0.5, 0.3, 0], user 1 Understanding X (1,1) for this problem 1

It can be calculated as That is, user 1 can be estimated to understand problem 1 65 percent.

However, the user's understanding of a particular problem and the probability of solving a particular problem may not be the same. In the example above, if user 1 understands problem 1, 65 percent, what is the probability that the answer is correct when actually solving problem 1?

To this end, the present invention introduces the methodology of Item Response Theory used in psychology, cognitive science, pedagogy and the like. Item response theory is a test theory for measuring the subject's characteristics, such as cognitive ability, physical ability, skills, knowledge, attitudes, personality traits, or difficulty of an assessment item, using responses to assessment items.

According to the present invention, a multidimensional two-parameter logistic (M2PL) latent trait model (M2PL) devised by Reckase and McKinely was applied to analyze the relationship between X and P.

As a result of experimenting with sufficiently large data by applying the above theory, the probability P of matching the problem with X is not linear, and the result as shown in FIG. 2 was observed.

FIG. 2 is a two-dimensional graph of a result of experimenting with problem understanding degree X and probability P to solve problem using sufficiently large data, where X axis is understanding and Y axis is the probability of correct answer.

Using the graph above, a function for estimating the probability P for the user to solve the problem

Can be derived as in Equation 2 below. In other words, the problem correct probability probability P function

The problem understanding can be calculated by applying X.

In the example above, where User 1 understands Problem 1, 65 percent, the probability that User 1 solves Problem 1 is

Calculated as: 86%. In other words, User 1 does not understand Concepts 2 and 4 at all, fully understands Concept 3, and Problem 1 is 20% of Concept 2, 50% of Concept 3, and Concept 4 of 30%. According to this, if user 1 solves problem 1, there is an 86% probability that the answer will be correct.

In operation 150, the data analysis system may recommend a user-specific problem using a probability of correct answer to the user's problem.

For example, a service server can create a list of suggested problems for a user and provide them, sorting the remaining problems in order of low probability P for a particular user, except for problems that the user has already solved in a set of problem databases. .

In other words, the algorithm recommending the problem to user i sorts the i-th row of the matrix P, starting with the lowest value, sequentially listing the corresponding problem numbers, and suggesting the solution in the order except the previously solved problem. It can work in such a way.

According to another embodiment of the present invention, the service server may include the problem in the recommendation problem list when the user's answer probability P is less than or equal to the threshold even if the user has already solved the problem.

According to this, when a specific user solves a specific problem, the probability that the answer is correct varies from user to user, and because the problem varies from problem to problem, a more optimized problem can be recommended to the user.

For example, if the value of row 1 of the matrix P is [0.3, 0.4, 0.1, 0.9, 0.7], the probability that User 1 has corrected the problems 1, 2, 3, 4, 5 is 30%, 40%, 10, respectively. It will be interpreted as%, 90%, 70%. The service server may provide the user with a problem recommendation list prioritized in the order of problems 3, 1, 2, 5, and 4. At this time, if user 1 has already solved problem 5, the service server may prepare a problem recommendation list except problem 5.

3 is a diagram for describing a method of estimating a conceptual understanding level L of a user and a conceptual configuration diagram R of a problem according to an exemplary embodiment of the present invention. FIG. 3 illustrates a detailed description of step 120 of FIG. 1, and illustrates a process of estimating a user's conceptual understanding L and a concept inclusion R of a problem based on a stochastic gradient algorithm.

In operation 310, the data analysis system according to an exemplary embodiment of the present disclosure may assign L and R to initial values according to conditions.

Furthermore, the problem solving result data can be generated with a list that sets val to the corresponding user, 1 for correct answers, and 0 for incorrect answers. For example, a problem-solving result data set for each user can be converted into a list having the correct answer (i, j, 1) when the user i solves the problem j, and (i, j, 0) if the answer is incorrect. .

After that, the data of the list can be loaded randomly one by one. (Step 320)

In operation 330, the data analysis system may calculate the temporary value tmp of the probability P for solving the problem by applying L and R at the corresponding point in time. This is to reflect the difference between L and R between the probability P calculated by applying L and R at that time and val, which is the result of the actual problem solving.

If the selected data is (i, j, val), the data analysis system uses L _i , R _j at that point in time.

Can be calculated.

For example, if the selected data is (1, 1, 0), it means that user 1 solves problem 1 and is wrong. However, if we calculate using L ₁ , R ₁ at that time, and user 1 solves problem 1, there is a 86 percent probability, that is, if P = 0.86, L ₁ , R ₁ is the data ( It needs to be updated to reflect 1, 1, 0).

To this end, the data analysis system according to the exemplary embodiment of the present invention may calculate an approximate gradient grad for the selected data (i, j, val) using the difference between the temporary value tmp and the current value val. (Step 340) This can be calculated using the formula below.

Furthermore, L _i and R _j can be updated according to the following formula by reflecting the number of problems solved by the user and the number of solved problems with grad. (Step 350)

In this case, the mu is a parameter that determines the performance of the formula is selected by the experiment, the step may be selected by the experiment as a parameter for the rate at which the formula accepts new data.

According to an embodiment of the present invention, if the above process is repeated for the entire data, and if all the data is reflected, the list is randomly mixed again, and if the process is sufficiently repeated, L and R can be reliably estimated. (Step 360)

The embodiments of the present invention disclosed in the specification and the drawings are only specific examples to easily explain the technical contents of the present invention and aid the understanding of the present invention, and are not intended to limit the scope of the present invention. It is apparent to those skilled in the art that other modifications based on the technical idea of the present invention can be carried out in addition to the embodiments disclosed herein.

Claims

What is claimed is: 1. A method for providing personalized educational content by an electronic device.

A step of collecting result data of the user's problem solving;

Estimating an understanding degree L of at least one concept for each user and a configuration diagram R of the concept for each problem by using a part of the result data;

And estimating an understanding degree X of the problem of the user and estimating a probability P of the user correcting the problem using the understanding degree.
The method of claim 1, wherein the b step,

Create a list in which the initial values of L and R are given as arbitrary values, and the result value of solving the problem j of the user i is given a current value val of 1 for correct answers and 0 for incorrect answers. Loading one by one;

Calculating a temporary value tmp of a probability P for solving the problem by applying L and R at a specific time point;

An approximate gradient grad is calculated for the loaded data (i, j, val) using the difference between the current value val and the temporary value tmp,
Characterized in that;

And updating the concept understanding L i of the user and the concept diagram R J of the problem by using v j as the number of problems u i and the number of problems j solved by v j . To provide customized educational content.
The method of claim 2,

Step b,

If the number of users constituting the result data set is n and the number of problems is m, the number of concepts is set to r, the L is defined by an n by r matrix, and the R is m by r matrix. Including the steps defined by

Wherein each element of L has a value between 0 and 1, and the sum of element values of each row of R is 1;
The method of claim 3, wherein

X is a matrix X = LR T expressed as the product of the transpose of L and R,

P is a value obtained by applying a function to each element of X,

The function
Method for providing a customized education content, characterized in that the following formula.
The method of claim 4, wherein

And excluding the problems already solved by the user, and sorting the remaining problems in order of low probability P to the user.
The method of claim 5,

The updating step,

Updating the user's conceptual understanding L i and the conceptual configuration diagram R J of the problem according to the following formula,

Mu is selected by an experiment as a parameter that determines the performance of the formula, and step is selected by the experiment as a parameter for the rate at which the formula accepts new data.