How to train rasa with huge data?

If I need to train Rasa with csv file how can I do that so that it can learn quickly?

what exactly you mean by csv file. Rasa models are supervised models. So they need labelled data (for rasaNLU data should be labelled with intents and for core it should be story format). If you have data something similar to a csv of user conversions then you have to clean it first in nlu and story format then only you can train them.

there are some useful tools like https://www.npmjs.com/package/rasa-nlu-trainer and Chatito DSL - Generate dataset for chatbots can help to clean the data.

1 Like

I got it, thanks for the suggestion. If model is predicting wrong how to change and train the model in rasa?

two possiblities

  1. If your nlu model is predicting wrong intent or entity then you need to improve your training data (nlu.md). You can check it by the command rasa shell nlu for any test utterance check whether the intent and entity are predicting correctly.
  2. If NLU is predicting correct intent then check for the rasa core (story.md) use the interactive shell to test it. command: rasa interactive refer Interactive Learning for more.
1 Like