Um sistema baseado em PLN e AG para apoiar a mediação pedagógica em fóruns de discussão.

With the growth of e-learning discussion forums have been widely used to promote interaction and collaboration between students and teachers asynchronously. Despite the benefits to the teaching and learning process the use of the forums in the e-learning can mean overload for the teachers/tutors given the large amount of information generated in the debates. Taking into consideration the importance of the teaching performance in the follow-up of the discussions and consequent help to the students of the e-learning, the teacher/tutor overload is considered a problem. Therefore, this study presents a system that integrates Natural Language Processing and Genetic Algorithm to classify postings from discussion forums such as Doubt, Neutral comment or Answer. As a way of evaluating the performance of the proposed system, experiments were conducted in three databases, with a total of 1769 posts, arriving at the average F-measures of 0.981, 0.997 and 0.988 in the classification of posts as Doubt, Neutral comment and Answer, respectively. After the experiments the results were analyzed by means of the z statistical test with a confidence interval of 95%. The results show the potential of the proposed system to identify post genres, from discussion forums, and the teacher/tutor can use it as a support to direct their efforts, for example, to students with any doubts in the current discussion.