Is it better to have lots of stories? - general discussion

Hey Everyone,

I have a bot with around 110 stories and I was just wondering if there is such thing as “too many stories”. What do you think, e.g. what if it is going to be thousands of stories, is that still fine, does it help the model to train better? Wouldn’t that make it terribly slow?

Please share your thoughts on this topic and whatever relating ideas!

Hello @lauraperge. That’s a great question. I think it’s important to approach it as any other ML problem - the more data you have, the better your results should be. However, it’s very important to make sure that you have good data to train your models. In Rasa’s case, your model will get better if your training data will contain stories with different dialogue turns which will enable your assistant to learn various deviations from the happy path and learn how to generalise on unseen dialogues. Another thing that you should keep in mind having lots of training stories is that the more data you will have, the more computationally expensieve the training will be. So, the short answer is yes - more training stories will improve your dialogue models, but it’s important to have good quality data.


Thanks for the answer!