Emotion Recognition Through Voice Acoustics and Semantics with RNNs