Every intent has a Training Material section consisting of the Main User Input and a number of Alternative User Inputs.
Using Natural Language Classification, your short tail classifier (be that Humley's Octavo or a third party provider) will use this data to match an intent to a user utterance, in order to provide a response.
To create this mapping, the classifier needs to be 'trained'.
When to Train
Each time you create or edit intents, you may see a change in the counter next to the Train option in the header (top right), which indicates the number of intents that have been added or changed since the last time the project's knowledge base was trained.
You need to Train your project when you are ready to make these changes available to your end users.
How to Train
Click the Train button in the page header (top right).
If there have been any changes or additions since the project was last trained, you will have the option to commence training.
Click the Train Project button.
The counter next to the Train option in the header will change to a spinning circle, indicating that training has started.
Training can take between a few minutes to an hour, depending on the size of your knowledge base. Irrespective of the number of changes, the whole knowledge base is subject to the training process.
N.B. Your Small Talk intents are treated as a separate collection to your Standard and Flow intents. This means that if you are training changes to your Standard intents, and no Small Talk intents have been added or edited since the last training, the Small Talk Intents will be not be included in the training process and vice versa.
Once training is completed, the counter next to the Train option in the header will change to a 'checkmark' or 'tick'.