Add NLU training data
Let's just teach our NLU model to recognize simple things like "Hi", "Thanks", "Bye". We'll do more advanced stuff later.
Botfront comes with pre-trained intents for general conversation (Chit Chat). We'll just use those for now. The video shows how to:
- Import Chit Chat intents to your model
- train and test your model
Create a Bot response
Now that our bot can understand a few things, let's see how we can get it to respond. The following video shows how to apply create a bot response and to assign it to an intent.
Botfront adds a special behaviour to intents prefixed with
This example is minimal, but you can do more advanced assignments such as combinations of intent and entities.
How is this different from the Rasa
MappingPolicylets you map one intent to one action. The Botfront behaviour lets you map any combination of intent and entities to a bot response.
- Adding/changing questions doesn't require training, as all Q&A are handled with a single story.
- A nice corollar is that you can use this single story to handle all chitchat or Q&A inside your contextual stories.
Monitor and improve
You can follow the conversations from the
conversations menu item, and monitor NLU from the
Activity tab in your NLU model.
Make sure to check Log utterances to activity in your
Model > Settings > Pipeline
- How to setup Botfront on your machine
- How to add data and train your first NLU mode,
- How to create a simple Q&A bot without coding.
There's a lot you can do with this already, but there's way more. You could read the NLU guide and build a more advanced NLU model