What is it for?

The training page allows you to train the voice agent to better understand what the callers said. All the sentences said by the callers that have not been understood will be displayed here, so that you can classify them in the right intent. This is what is called supervised training.

Overview

The training page is divided into 2 parts:

• The list on the left:
  • it allows you to navigate between the questions of the voice agent, where there have been not understood utterances
  • it allows you to visualize those utterances with the number of occurrences. The list is sorted so that you can see the utterances that are the most not understood first.
  • you can also choose the period and the number of occurrence you want
• The detailed page on the middle:
  • the calls where the conversation error occurred
• The detailed page on the right:
  • It allows you to actually do the training of the utterance and choice the right intent for this new utterance

How to classify an utterance into an intent?

One of the first things you will want to do is to classify one of the not understood utterances into the voice agent’s intents. Intents of your voice agent are created in the

To do that, you have to click on the utterance you want to classify. It will appear in a text input, where you can select the specific part of the sentence you want to add to the intent. Indeed, sometimes all the sentence is not worth being classified, and you will want to only keep a part of it.

Once you have selected the part of the sentence you are interested in, you can choose to add it to intent or an entity. In our case, we want to add it to an intent.

You will be able to add the selected sentence to any intent :

• Expected intents: this is the intent that is expected at this specific question
• Fallback intents: those are the intents used in the Fallback script

Once you have selected the right intent, a tag will be displayed: