Train your bot based on actual user messages

Once you have a bot configured you want to train your NLP models with actual user messages so that your bot becomes smarter over time and support different kinds of expressions. For this you can use the NLP Train section.

  • Let us first create some actual training data by talking to our bot on Facebook Messenger. If you haven't activated your bot on Facebook Messenger yet, follow the steps in the Channel wizard in the platform.

All conversations from your channels will be loaded into the Train module. Conversations from the emulator are ignored.

  • Look at the train NLP section. As you can see the user messages has been labelled by the NLP model so that we can evaluate this message. It seems that our model has detected the correct intent and entity values.

  • Click on '+' to add this candidate expression as validated expression.

  • Let's do a second test by sending a new user message on Facebook Messenger.

  • It seems that the entities are not recognised correctly.

  • Adjust the expression by adding those entities, train and publish your model and test in the NLP test section. If needed add similar expressions to this intent.