Advances in learning systems over the past two decades have enabled the development of technologies that help in the engagement of students. Although these systems may use behavioral procedures to improve reading skills, better outcomes for each student are obtained in the manual elaboration of a set of tasks by educational experts. However, the use of a manual process requires too much time, effort and subjectivity for the creation of tasks. Additionally, even with the aid of computational processes, the automatic generation may be impracticable due to the high search space for the possible combinations of tasks. This process could consider adapting the difficulty of a task to the student's knowledge, something little explored in educational work for children at the beginning of reading learning. The present thesis implements an approach to generate teaching tasks from the Matching-to-Sample procedure, adapting its difficulties through bio-inspired optimization meta-heuristics. This approach uses pre-test results applied to students and the configuration of teaching contents determined by educational tutors; these data allow the use of the algorithms to generate tasks and then the tasks can be presented in learning software. Experiments demonstrated a better convergence of the genetic algorithms for this domain, being able to generate tasks on a level of difficulty adapted to the students, and also according to pretests and configurations of attributes of the tasks defined by behavioral psychologists. As validation for this study, the tasks were applied to a group of students in the early stages of literacy achieving satisfactory effects in the individual learning process. In addition, two interactive learning software were implemented through a digital game and a web application, where the use of the digital game with playful features showed superior acceptance in the use of teaching tasks adapted for children in the initial phase of literacy.
|
