How to design a big amount of QNAs with RASA

Hi All, I have around 3 million records of QNAs about agriculture field, and I want to use rasa to develop my chatbot. I don’t know how to design the data to fit in nlu.md and stories.md files.

  • How to think about the Intents?
  • How to extract the entities?
  • How to design the stories?

here’s a sample of the data

Hello @wafaa you can categories intents based on nature of query or crop type based om season etc. For storied you have to simply think of conversation flow. Just imagine how will your user will come tp chatbot and ask questions and relative responses (uttarances and custom action).

Since you have huge database you can try to integrate the sheet itself to answer questions from users.

For ex :

User : Tell me about Wheat? Rasa Bot will query relative information from sheet and integrate it into response like this : - Bot : wheat is {cell number} crop. The favorable conditions for its growth are {cell number}.

Thanks @athenasaurav for your replay, Do you mean by integrate the sheet is to make some custom action to get the answer form the sheet based on the user inputs like (question itself, crop, season)?

for what I did, I used the “QueryType” as an Intent, but for one “QueryType e.g. Plant Protection” there are a thousands of questions for a hundreds of Crops !! and tried it, the bot couldn’t even detect the Intent

So, I thought that the intents are wrong form the start.