Hello, I am new to Rasa. I am trying to convert my Chatterbot project to Rasa. In Chatterbot, I had a large amount of training data, and would like the bot to be trained with the same amount of data. My question is how should I structure my Rasa project as the training data scales? Wouldn’t the domain file keep getting longer with more utter actions? Is there a better approach to training with a large amount of data?
Thanks.
Hi Braden,
What does your Chatterbot do? More training data shouldn’t necessarily mean longer domain files so long as the new stories, regardless what path they go down, only contain actions/utterances that are already in the domain. How many different actions are you expecting your bot to manage?
My chatterbot was built as a Q&A bot for a local company, meaning the user can ask questions about the company’s business or structure and the bot answers appropriately. I had about 120 unique questions available for the user to ask. Do I need an action defined for each Q&A?
I guess that depends how it was handled in your Chatterbot. How did that bot answer – did it have pre-written answers for each question, or did it connect to some sort of FAQ page based on the question and pull answers from there, or something else?
Our demobot currently has intents for each FAQ and an utterance to respond to it, but she has way fewer than yours does.
Yes, the responses were pre-written. I found some answers in a different post here: Mapping FAQ with RASA for large dataset (2000+). I might have to rethink my implementation. Thank you.
How did you solve this problem in rasa ? i have same issue