LESSON: Different types of NLG

Text-to-speech considerations. We developed a system to generate responses for a dialogue system, where the texts were meant to be read out loud by a Text-to-Speech (TTS) component. We designed, implemented, and tested narratives with written text in mind. Later on, when we started listening to the narratives, we realized that they sounded robotic, so we had to go back to the design phase, which meant undoing or patching a lot of the work we had already done.