A Review of Graphical Model Architectures for Speech Recognition