Listen, Attend And Spell: A Neural Network For Large Vocabulary Conversational Speech Recognition | Manumap