Automatic speech recognition and machine translation with deep neural networks for open educational resources, parliamentary contents and broadcast media