This algorithm arises from the need to classify questions in educational multimedia intelligently.
Use the Item Response Theory (IRT) to determine the probability that a question or item will be answered correctly according to the difficulty, degree of discrimination and probability of divination.
development-of-an-algorithm-for-its-use-in-educational-softwareThen, with this probability that the answer is correct, a Bayesian network is drawn up that will determine the sequence of contents or subjects to be studied a posteriori according to the expected skill level. Materials on IRT, Bayesian networks, and classical item response theory were used.
Key words: Item response theory, Bayesian networks, multimedia.
INTRODUCTION
This algorithm comes from the Computerized Adaptive Tests (TAI) which are much more general. It is the result of having simplified and specified it in the direction of educational multimedia. There are subsequent works with only IRT and with only Bayesian networks applying classical evaluation theory. Eva Millán (Millán, et al., 2001) has developed a very interesting work in this branch of uniting IRT and Bayesian networks, but sometimes she uses IRT where she violates the assumptions posed by this theory, for example she includes questions that are related with two or more contents violating the Unidimensionality assumptionof the item. This determines that the theory is not used correctly and if the assumptions do not fit the hypothesis, the thesis is not valid. So we can conclude that the algorithm as Millán conceives it has certain problems with the IRT in the assumptions that are corrected in this algorithm.
The objective of this work is to obtain an efficient algorithm and a relative ease of implementation to apply to educational software, mainly for the multimedia created by our software houses to put it in the hands of students at all levels of education. Because it is not just an algorithm that shows the ability of an examinee, it also guides you and shows you the correct answer in case the answer is wrong. Of course, this may be optional depending on what type of software the algorithm will be used, if it is a test where it does not matter that the answers are known, although it is not the main objective, this step of showing the correct answer in case of incorrectness.The main difference of this algorithm is that it adapts the TAIs to the idea of smart tutors.
- DEVELOPING
1.1 Artificial Intelligence
For a long time, man has endeavored to create a similar, not only in the external aspect, but in reasoning and feelings. The monster of Mary Shelley's play, Frankenstein is an expression of this longing. The rise of computers opened the door to what was for a long time off limits to the fiction books: the creation of an intelligent entity, artificial intelligence. This branch of science soon gained popularity among scientific circles. Some of them gave different definitions like:
"I try to reproduce (model) the way people identify, structure, and solve difficult problems" (Miller, et al., 1982).
“The interesting task of getting computers to think (Haugeland, 1985).
"The art of creating machines with the ability to perform functions that when done by people require intelligence." (Kurzweil, 1990)
"A field of study that focuses on the explanation and emulation of intelligent behavior based on computational processes." (Schalkoff, 1990)
"Study of the intelligence mechanisms and the technologies that support it." (Newell, et al., 1988)
"The study of how to get computers to do tasks that, for the moment, humans do better." (McGraw Hill. Rich, et al., 1991) It is essentially the way in which a real problem - tedious for man - must be solved by a computational model.
1.2 Expert System
Any computer-aided system that with human intervention automatically executes knowledge tasks in a cumbersome domain of application and where skills, experiences, judgments previously acquired by experts, and heuristic knowledge derived from historical data accumulated by man in observation are used. of the world that surrounds it and not from bibliographic sources, it is called the Expert System (SE).
An SE is mainly made up of the Inference Engine (MI), whose function is to test the hypothesis and is generally implemented as a Software that performs search operations, using algorithms that operate on the Knowledge Base (BC) of the application domain. The BC relies on the BC for the search for solutions; therefore, the magnitude and precision of the information that the BC stores on the problem domain have a decisive weight in the certainty of the answers. (Lago Aguilar, 2004) .
1.3 Expert systems based on probabilities.
A SEBP is considered a type of SE whose BC of knowledge is made up of elements of a probabilistic space or abstract knowledge and the MI is based on the Bayesian Inference Framework through the strategy known as probabilistic reasoning or probabilistic inference.
In a SEBP probability is assumed as an intuitive measure of uncertainty, in which the joint probability distribution (CPD) of the variables is used to describe the dependency relationships between them and conclusions are drawn with well-known formulas from the Theory of
Odds.
Traits and classes are used to characterize the population elements of any objects. Understand by class each possible solution to the problem and by trait, each characteristic presented by the classes. Membership of an object to a certain class is given by the characteristics that characterize the class and it can belong to one and only one of them. When selecting an object from the population, it is also determined to which class it belongs and what are the characteristics that characterize it. The frequencies of appearance of said objects for the possible values of the traits and classes to which they belong, is precisely what makes up the BC of the SEBP. The relationships between the features of the objects in the population are described by their joint probability function and that is whythat this function is part of what is called knowledge. (Lago Aguilar, 2004)
1.4 Bayes' theorem.
If the probability of A occurring is PA, we call the probability of events A and B (both) PAB. PA - B to the probability that A occurs when we know that B has occurred (it is called conditional probability). It has been determined that the probability of occurrence of A and B is equal to the product of multiplying the probability of B by that of A once B has occurred:
Suppose we divide A into three subsets (A1, A2, A3) as shown in Figure 1:
Figure 1
These events are mutually exclusive and they constitute all possible situations, note that PA 1 PA 2 PA 3 1, because event A will always occur. Event B can occur in any of the three situations. If we remember the previous equation:
is the dichotomous value of item j being answered (1 correct, 0 incorrect). P j is the probability that item j is answered correctly.
Set of questions or items.
It is a standardized level that depends on the level of teaching and subject of the test.
See 2.7 TAI Computerized Adaptive Test
See 2.7 TAI Computerized Adaptive Test.
Idem.
Perhaps the model is a series of numbers that only makes sense to the program.
See 2.7 TAI Computerized Adaptive Test.
How the next item will be calculated is still under study.
Download the original file
