KR101993299B1 - System and method for diagnosing mass attributes of leaner - Google Patents

Abstract

The present invention relates to a system and method for diagnosing a learner's learning ability by dividing it into a plurality of concepts.
The learner's multiple conceptual diagnosis method according to the present invention includes a step of selecting a plurality of concepts to be diagnosed, a step of providing a problem with a plurality of selected concepts, a step of constructing a relationship between a problem and a concept, A step of constructing a matrix of correct answers to each question of the student, a step of constructing item parameter of the question, a step of constructing dictionary probability information about the selected concept, The method comprising the steps of:

Description

[0001] SYSTEM AND METHOD FOR DIAGNOSING MASS ATTRIBUTES OF LEANER [0002]

The present invention relates to a system and method for diagnosing a learner's learning ability by dividing it into a plurality of concepts.

The MLE (Maximum Likelihood Estimation) method of the DINA (deterministic input, noisy "and" gate) model according to the prior art finds a probability using a uniform distribution. In the case of a large number of concepts, (The total number of patterns of the concept: ^2K ), which makes it difficult to implement the DB.
(Patent Document 1) KR10-2013-0037541 A

The present invention has been proposed in order to solve the above-mentioned problem. The present invention calculates only the marginal probability for each concept without storing all the number of patterns of the concept, calculates the dictionary probability necessary for learner diagnosis using K stores , And to propose a system and method capable of diagnosing a learner's learning ability by dividing it into a plurality of concepts.

The learner's multiple conceptual diagnosis method according to the present invention includes a step of selecting a plurality of concepts to be diagnosed, a step of providing a problem with a plurality of selected concepts, a step of constructing a relationship between a problem and a concept, A step of constructing a matrix of correct answers to each question of the student, a step of constructing item parameter of the question, a step of constructing dictionary probability information about the selected concept, The method comprising the steps of:

Since the number of concepts that can be estimated at a time is limited, the DINA model requires a plurality of concept estimation methods in order to utilize in the actual education field. According to the embodiment of the present invention, It is possible to estimate the degree of understanding of a number of concepts quickly by using the model repeatedly and to improve the estimation accuracy by storing the dictionary information.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the following description.

1 is a diagram illustrating a configuration of a learner's multiple conceptual diagnosis system according to an embodiment of the present invention.
2 is a diagram illustrating a preparatory process for diagnosing a learner's multiple conceptual diagnosis system according to an embodiment of the present invention.
FIG. 3 is a diagram illustrating a process of estimating the degree of understanding of a plurality of concepts of individual learners in a multiple concept diagnosis system of a learner according to an embodiment of the present invention.
FIG. 4 is a flowchart illustrating a preparatory process for learner diagnosis of a learner's multi-concept diagnosis method according to an embodiment of the present invention.
5 is a flowchart illustrating a learner diagnostic process of a learner's multiple concept diagnosis method according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, advantages and features of the present invention and methods of achieving them will be apparent from the following detailed description of embodiments thereof taken in conjunction with the accompanying drawings.

The present invention may, however, be embodied in many different forms and should not be construed as being limited to the exemplary embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, And advantages of the present invention are defined by the description of the claims.

It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In the present specification, the singular form includes plural forms unless otherwise specified in the specification. &Quot; comprises "and / or" comprising ", as used herein, unless the recited component, step, operation, and / Or added.

Before describing the embodiments of the present invention, the DINA model will be described first to help those skilled in the art understand it.

The DINA (deterministic input, noisy "and" gate) model is a representative cognitive diagnostic model that estimates whether learners master each concept from group test information.

In other words, the DINA model estimates each student's conceptual understanding from the group test information. When the number of students who have taken the test is I, the number of problems is J, and the number of concepts is K, There are two main types.

The first is a Q-matrix, a J × K binary matrix indicating the concepts associated with the concepts used in the group trials, and the second is an R-matrix, in which the students are informed of the correct answer It is an I x J binary matrix.

Based on these two pieces of information, it is estimated whether or not each concept is understood by each student, and the result is output as an I × K binary matrix.

The Q-matrix, which is the mapping between problem and concept, is defined in advance by educational experts. The concepts necessary to solve the problem are defined in detail and mapped to the detailed concept for each problem.

The i-th student's understanding vector for the k-th concept, ie, α _i, which represents the understanding of the concept of learner i, is a binary vector. A value of 1 for each element means that the concept is understood. .

The "AND" gate in the DINA model assumes that the student needs to know both the problem and the mapped concept in order to get the correct answer.

The potential response vector η _ij is modeled as [Equation 1] as an element for determining whether the i th student has the ability to solve the j th problem.

[a] is the maximum integer not exceeding a, η _ij is expressed as 1 if the learner i has the ability to match the problem j, 0 if the learner i is not capable of the problem, 1 "," (2) "

Where q _jk is an element of the Q-matrix and indicates whether knowledge of concept k is needed to solve problem j, 1 indicates that the concept is needed, and 0 indicates that the concept is not needed.

k means the total number of concepts in the model. α _ik means whether learner i understands concept k, and α _i is a vector having α _ik as an element.

The "NOISE" of the DINA model is a noise parameter that indicates the nature of the problem. The noise parameter, s _j , which is the probability that a potential response will be mistakenly missed despite the 1, and the probability that a potential response will be met _j , and is expressed by Equation (3) and Equation (4) below.

These problem parameters are problem-specific parameters that vary in value depending on the problem and do not vary from student to student.

From this assumption, given the learner's concept vector α, the probability density function of the i-th learner response X _ij is given by [Equation 5] and [Equation 6], where X _ij is the learner i solving the problem j If the answer is positive, it means correct answer. If it is 0, it means bad answer.

This can be summarized as a single equation.

The concept of the DINA model can be estimated by the maximum likelihood estimation method.

First, assuming that X _ij are conditionally independent and have the same probability distribution with respect to a given concept vector α, the following is summarized as follows.

를 구하면 아래 [수학식 10]과 같다.From this, the Bayesian method estimates the learner concept vector

(10). ≪ EMI ID = 10.0 >

&Quot; (10) "

과 문제 파라미터 s_j, g_j를 동시에 추정할 수 있다.In this case, the above equation is implemented by the Expectation maximization algorithm,

And problem parameters s _j and g _j can be estimated at the same time.

을 저장하는 메모리가 급증하여 한 번에 추정 가능한 개념 수는 제한적이다(컴퓨터 사양에 따라 다르지만, 보통 추정 가능한 개념 수는 약 12개 정도이다).However, when the above model is actually implemented, the number of concept patterns increases exponentially as the number of concepts increases (the number of concept patterns is 2 ^K when the number of concepts is K), so that the dictionary probability P (α) percentage

, The number of concepts that can be estimated at one time is limited (usually about 12 concepts are available, depending on the computer specification).

Also, as the number of concepts increases, the computation time of the EM algorithm exponentially increases.

Therefore, when estimating the understanding of a large number of concepts with the above model, it is inevitable to divide the concepts into appropriate numbers and estimate them several times.

In particular, the dictionary probability is difficult to store and utilize in the existing method because the entire pattern is different according to the concept to be estimated.

1 is a diagram illustrating a configuration of a learner's multiple conceptual diagnosis system according to an embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of a learner's multiple concept diagnosis system according to an embodiment of the present invention. FIG. 3 is a diagram illustrating a process of estimating the degree of understanding of a plurality of concepts of individual learners of a multiple concept diagnosis system of a learner according to an embodiment of the present invention. Referring to FIG.

In order to diagnose a number of concepts, a learner's multiple concept diagnosis system is constructed as shown in Fig.

The learning behavior DB 200 according to the embodiment of the present invention stores the R-matrix (X = {X _ij }), and stores all the correct answer information about the problem solved by the learners.

The content information DB 300 stores the entire concept and a Q-matrix (Q = {Q _jk }) for the problem. All the problems provided to the learner must be matched with a predefined concept, the information is stored in the contents information DB 300, and the item parameters s _j , g _j , which are estimated together in the diagnosis process, are also stored.

The learner's understanding DB 400 is stored with the diagnosis time of the learner's conceptual understanding diagnosis result and the whole conceptual understanding degree of the whole learner is expressed on the time axis.

The content DB 100 may be configured together with the learner's multiple conceptual diagnosis system according to an embodiment of the present invention or may be configured with an external DB of another Learning Management System (LMS). The contents DB 100 stores learning contents actually provided to the learner.

Hereinafter, a preparation process for diagnosis according to an embodiment of the present invention will be described with reference to FIG.

When the learner solves the problem, the learner response data (question ID, concept ID, learner ID, R matrix elements) are accumulated from the contents DB 100 to the learner terminal 500 by providing the test questions for the preparation process.

 If the learner response information for each problem is sufficiently accumulated in the learning activity DB 200, the item parameters of the respective items (probability of matching with a shot, probability of being mistakenly mistaken) are estimated from the DINA model.

The learning behavior DB 200 accumulates information in the R matrix for all learners and responses, and is a DINA model input configuration for item parameter estimation.

In this case, because the number of concepts is large, the entire Q-matrix and the entire R-matrix can not be used as they are, so the Q-matrix and the R-matrix are divided into subsets and estimated several times.

In an embodiment of the present invention, the divided Q-matrix is referred to as a Q'-matrix and the divided R-matrix as an R'-matrix, and each subset may be of arbitrary size.

At this time, the problem is divided into groups based on the problem, and the number of concepts related to the problem is not more than eight and the groups do not overlap with each other.

From the R'-matrices of the learning behavior DB 200 obtained therefrom and the Q'-matrices of the contents information DB 300, an item parameter for the problem is estimated by the EM algorithm using the existing DINA model, And stores the result in the content information DB 300.

When the estimation of all the item parameters is completed through the preliminary preparation process, as shown in FIG. 3, the understanding degree of a plurality of concepts of individual learners is estimated.

If a learner or a learning manager selects a plurality of concepts to be diagnosed, the learner terminal 500 provides problems regarding the concept by referring to the concept-problem link in the contents information DB 300. [

The response information to the learner's problem is stored in the learning activity DB 200 and then composed of R'-matrices (one learner, no problem is unlimited), and the Q'- (300).

The posterior probabilities of the learners' understanding according to the learner concept can be obtained from the R'-matrices and the Q'-matrices and the item parameter information by the maximum likelihood estimation (MLE) method of Equation (10) There is an advantage that the operation speed is very fast and can be implemented in real time.

According to the embodiment of the present invention, a method of utilizing the dictionary probability information to derive a more accurate result is proposed.

In the conventional MLE method, the dictionary probability P (α) is obtained by using a uniform distribution.

According to the embodiment of the present invention, the learners' understanding degree DB 400 can be accumulated to constitute a dictionary probability of understanding of concepts of all learners and utilize them as dictionary probabilities.

However, when the number of concepts is large, the memory for storing the dictionary probability increases exponentially, which makes it difficult to construct the DB in practice.

Thus, in accordance with an embodiment of the present invention does not store all of the entire pattern can be 2 ^K pieces of the concept to solve these problems, marginal probability for each concept _{(P (α 1 = 1)} , P (α 2 = 1 ), ..., P (? _K = 1)), and uses K repositories.

According to the embodiment of the present invention, it is possible to calculate the dictionary probability required for diagnosing a learner, assuming that the probabilities for each concept have independent characteristics. For example, 3 is selected, the information of P (? ₁ = 1) and P (? ₃ = 1) is retrieved from the learner's understanding DB 400 and the information of the individual learners Estimate the degree of comprehension.

&Quot; (11) "

FIG. 4 is a flowchart illustrating a preparatory process for learner diagnosis of a learner's multiple concept diagnosis method according to an embodiment of the present invention. FIG. 5 is a flowchart illustrating a learner diagnostic process of a learner's multiple concept diagnosis method according to an embodiment of the present invention. Fig.

Referring to FIG. 4, the preliminary preparation process for learner diagnosis according to an embodiment of the present invention constructs an R-matrix by obtaining learner problem response information after constructing a concept definition and Q-matrix (S110) S120).

In step S130, it is determined whether the learner response information is sufficiently accumulated in the learning activity DB. If the learner response information is not sufficiently accumulated, step S120 is repeatedly performed. If the accumulated Q- '- matrix and an R'-matrix (S140).

 Subsequently, an item parameter for the problem is estimated from the R'-matrices of the learning behavior DB 200 and the Q'-matrix of the contents information DB 300 (S150).

5, a learner diagnostic process according to an embodiment of the present invention includes a step S210 in which a plurality of concepts requiring diagnosis are selected from a learner or a learning manager, a problem associated with a selected concept, '- matrix from the contents information DB 300 (S220).

Then, the learner problem reaction information and R'-matrix are constructed (S230), and item parameter information of the question is configured (S240).

Step S250 of constructing the dictionary probability information of the selected concept constructs a dictionary probability for the selected concept as shown in Equation (11) according to the assumption that the probabilities for the respective concepts described above have independent characteristics.

Estimates the understanding degree of each learner according to the concept, stores the result (S260), terminates the learning according to whether the instruction is terminated (S270), or returns to the step S210.

The embodiments of the present invention have been described above. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

100: contents DB 200: learning activity DB
300: contents information DB 400: learner understanding DB

Claims (9)

(a) receiving, by the content information DB, selection information on a plurality of concepts to be diagnosed from a learner or a learning manager;
(b) providing the problem of the selected plurality of concepts to the learner terminal by referring to the concept-problem link in the contents information DB, and constructing a relationship between the problem and the concept;
(c) a matrix of the learning behavior DB about the response information to the question and whether the student answered each question correctly, and an R matrix for storing information on whether or not the students answered correctly, Constructing the Q matrix by dividing each subset;
(d) the contents information DB configuring the item parameter of the question;
(e) Learner understanding The DB accumulates learner understanding data on the selected concept to construct a dictionary probability of the learner's conceptual understanding. Using the marginal probability for each concept, Calculating the dictionary probability necessary for the learner's diagnosis by taking into consideration; And
(f) The learner understanding DB estimates the posteriori probability of the learner's understanding according to the concept and stores the result, and estimates the understanding degree of the plural learner's concepts
A number of conceptual diagnostic methods for learners.
The method according to claim 1,
Wherein the learning behavior DB confirms whether the learner response information is accumulated in the learning behavior database before the step (a), and a matrix obtained by dividing the Q matrix representing a relation between the problem and concept used in the group test, Constructing a matrix obtained by dividing the R matrix representing the correct answer to each question, and estimating the item parameter
A number of conceptual diagnostic methods for learners.
3. The method of claim 2,
The matrix is divided so that the number of concepts associated with the problem is less than or equal to the predefined number
A number of conceptual diagnostic methods for learners.
delete delete A database in which the learner's multiple conceptual diagnostic programs are stored; And
And a processor for executing the program,
The processor receives the learner response data for the test problem for the preliminary preparation process, and generates a matrix by dividing the Q matrix representing the relationship between the problem and the concept used in the group test, and R Constructs a matrix that divides the matrix, estimates the item parameters,
The learner 's understanding data is accumulated to construct the learner' s conceptual understanding probability. The marginal probability for each concept is used and the learner ' Estimating understanding of many concepts of learners
Multiple conceptual diagnosis system of learners.
The method according to claim 6,
The processor divides the matrix so that the number of concepts associated with the problem is less than a predetermined number based on the problem
Multiple conceptual diagnosis system of learners.
delete delete

Images