Avaliação Automática da Fluência em Leitura para Crianças em Fase de Alfabetização

Resumo

In Brazil, large-scale learning assessment is fundamental for public bodies responsible for education. Through these evaluations, it is possible to plan public policies aimed at improving education. When it comes to the first years of elementary school, an important aspect to evaluate is the ability of the student to use their mother tongue in the oral mode. Nevertheless, the process of evaluating orality on a large scale is still a very costly and time-consuming task. This paper proposes and evaluates the use of automatic speech recognition (ASR) for the automation of this evaluation process. Experiments were performed on a real base of audios and it was demonstrated that the automatic evaluation closely reflects the quality of the analyzed readings.