Perhaps this exists already, or am under thinking it, but the data that I would use to train my NLU model exist in my database, in a traditional row/column format where each row is an utterance and one of the columns includes a label for it’s intent.
Is there a utility to take this tabular format and convert it to one of the two formats that NLU uses to train the models?
But for the storyline writing, you have to do it manually right ? As your database contains only classification data that is NLU.
Eventually I’d love to have everything database driven to feed and train both NLU and Core models, but let’s start with the first part of NLU. If I have a tabular dataset, I can’t feed that to NLU today, so I am looking to see if there are utilities that currently exist which would help transition from a data store to Rasa for training purposes.
I have checked Rasa Documentation, but find nothing related to load Data from Database.
But i can build one such component for you. Can you give me your table format and Database details (mysql, pgsql or mongodb)
Let’s keep it simple; two column sql database both are text/varchar