Intents

Intent represents the action requested by the user. The intent expressed in the input sentence is identified by the NLU through a classifier trained with sentences provided by the developer in the agent’s configuration.

So that the system can perform training without any issues, each intent should have at least 5 example sentences. However, for better results, we recommend a minimum of 20 example sentences for each intent.

According to the context of the developed application, for the system to learn the patterns of sentences to be identified, more diverse examples may be necessary.

In defining intents, the developer should strive for a balance in the quantity of example sentences among the defined intents, meaning, ideally each intent should have approximately the same number of example sentences.

Below, we provide an example of a JSON configuration for an intent defined in the NLU agent:

{
  "name": "Depositar",
  "description": "Depositar",
  "sentence_examples": [
    {
      "text": "quero colocar dinheiro na minha conta corrente"
    },
    {
      "text": "vou fazer um depósito na poupança"
    },
    {
      "text": "tenho um depósito para fazer"
    },
    {
      "text": "depositar dinheiro na minha conta de poupança"
    },
    {
      "text": "gostaria de depositar R$10,00 em minha conta corrente"
    }
  ]
}