Title:	SCORING ALGORITHMS FOR EVALUATION TESTS IN ADVANCED E-LEARNING SYSTEMS
Author(s):	Roberto Barchino Plata , José María Gutiérrez Martínez , José Javier Martínez Herráiz
ISBN:	972-98947-0-1
Editors:	António Palma dos Reis and Pedro Isaías
Year:	2003
Edition:	2
Type:	Poster/Demonstration
First Page:	1026
Last Page:	1026
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	The proposal of this paper is to show ideas around the automatic correction of machine-generated tests in the context of e-learning systems. We show works realized around IMS norms in its QTI box. These works, which are orientated to implement the rules in a real system and its use to get a measure of its posibilities, have led to the necessity to bear in mind more complex scenaries than the scenaries posed by IMS and the need to develop adapted algorithms to these scenaries. The last scenary studied, includes the realization of proofs in several stages where the questions will have associated a level and they could only be selecionated for the same or higher level of that stage. A student will access to higher level questions only if he has reached a certain degree in the previous stage. The last version of our system includes a new evaluation algorithm which extends QTI standard following its specification. In this version we storage questions using XML like the own QTI structure. The new algorithm was developed to get the maximun similarity to the behaviour of the teacher evaluating a test. The reason of developing this new algorithm is to look for a higher precision in pupils’ knowlegment evaluation. We must take care of several items which will be able to give us the start in the development of the new learning evaluation algorithms. These items are: the exams have to contain the essential containts of the subject, the questions of all the exams must be clear without ambiguity, and also it must be known by the system users the evaluation criteria or the algorithm, and at last the quality of the test exam. With all this, we developed an algorithm which has 5 steps in the test generating phase. Each step contains four questions and depends directly on the previous step and the student’s answer to those questions. This behaviour makes the tests extremely flexible and variable, and needs an adapted method for evaluating the results of the test. Also, it is very important the exams preparation phase, in which, the teachers, create the questions for the test and evaluate the importance and difficulty of each question. The development of standards for components of e-learning systems are very important. Even more important, is to work next to the borders of these standards because it confirms the work done with them. The shown scenary and the applied scoring algorithm, work in this zone of knowledgment to open new ways for the development of actual standards. The results we have got till now and the shown ideas appear clearly like a way to be followed by the new versions of the standards and by our own systems.
	Go Back