Detecting Multiple Speech Disfluencies using a Deep Residual Network with Bidirectional Long Short-Term Memory