Investigating Deep Neural Structures and their Interpretability in the Domain of Voice Conversion