An Iterative Dual Pathway Structure for Speech-to-Text Transcription