When building conversational assistants with Rasa, what kind of dialogue designer do you need?
I read some documentation on it, amongst them @juste_petr blog on designing training stories
So I understand when using Rasa, dialogue design involves two phases:
- Initial, the basic to get started by creating some happy paths
- Evaluating and using real dialogues to further develop the conversational agent
As far as the initial phase concerns, to me dialog design seems to be much like creating a bot with only rules:
- You have to decide which happy paths are to be designed
- You have to write the dialogues / stories: which turns, intents, actions
And, off-course, you’ll also have to decide on which Rasa technique to use (rules and/or stories). To me, this doesn’t need a converstaional designer.
But when it comes to using real conversations, you’ll need:
- An communication and subject matter expert to create stories. Either by interaction with customers (=users of the bot) or by interactive learning.
- An expert on analysing the dialogues and creating the training data out of them.
- And off-course datascientist (building pipelines, data-analysing) and developer (custom actions, forms, interfaces)
So am I right when I state, there’s little need for a conversational designer who mostly worked on rule-based bots and has little or no experience with Rasa?