What are intents?

Dialogflow is a platform with complicated machine learning and artificial intelligence capabilities that matches what the user said to the intent. You can think of intents as questions that users will be asking about U.S. Presidents. When you first start with the FaqChatBot agent in Dialogflow, you will see two default intents: the Welcome Intent and the Fallback Intent, as shown in the following flowchart. The Welcome Intent gets triggered when the user first starts to interact with the FaqChatBot. Then, the user will start to ask questions and those questions will be matched against the intents, and if no matching intents are found then the Fallback Intent will be triggered and the FaqChatBot will respond with, I didn't get that. Can you say it again?.

The following flowchart shows the process flow of how intents are processed in Dialogflow:

 

The following screenshot shows the default intents that are created by Dialogflow:

ML classification sets the level of confidence of the matching algorithm identifying the user voice input to the intents. For example, if the user asks the FaqChatBot agent, Who is the first president of the United States?, the agent will attempt to match the voice against any defined intents in the system. When Dialogflow matches the voice against the existing intent in the system, a confidence threshold of 30% will be used. If the confidence level is above 30%, the FaqChatBot agent will process the user's request using the matched intent, however, if the confidence level is below 30%, the agent will reroute the request to the Default Fallback Intent, where all unidentifiable requests are processed. Another way to look at this is when your friend asks you a question over the phone and you cannot hear well, so you cannot understand what your friend said. But, sometimes, based on a partial understanding of the voice coming through the phone, you can predict or guess what your friend said. If you want to increase the accuracy of the matching algorithm, you can increase the default value from 0.3 to a much higher value in the agent settings. But for the purpose of this book, we will keep this set as the default.

The following screenshot shows the FaqChatBot app's ML Settings: