Enviar Registro por Correo electrónico: Exploring convolutional, recurrent, and hybrid deep neural networks for speech and music detection in a large audio dataset