Train your Virtual Agent to recognize Intents

Model user Intents and train your Virtual Agent to understand them!

To help your end customers achieve their goals, your Virtual Agent needs some training to match some well-defined user intents with the bandwidth of potential ways to express those intents. Learn more about Machine Learning Intents in the NLU section of our documentation.

#1 Create a Flow and navigate to the NLU Engine

Start with creating a new Flow and name it e.g. "IT Service Request". Then select "NLU" from the top menu.


#2 Create an Intent

Click Create Intent on the left-hand side. Give your Intent a name that describes its purpose and serves as an identifier e.g. "createTicket". Enter an example sentence a user could use to express their intent e.g.: "I need to create ticket". In the lines below, add more examples like "Open a ticket", "I have a problem". Save your Intent.

#3 Create another Intent

Create a second Intent. Use e.g. "resetPassword" with the sentences "Forgot my password", "I lost my password" and "Cannot login" and save the Intent.


#4 Train Intents

Click Build Model on top of the screen to start the building process of the machine learning model based on your example sentences. You can always check the progress in task menu of the top bar. Once the model is built, you will receive a green success banner.


#5 Try it out!

Open the Interaction Panel, switch to "Settings" and activate "Expert mode". Go back to the chat and type "file ticket". The Virtual Agent won't respond (yet) but you see meta data indicating that the Virtual Agent clearly understood the Intent "createTicket".

#6 Look behind the scenes

Switch to the "Info" tab in the Interaction Panel. You'll see a JSON object in the sub-tab "Input" giving you insights into the process: "createTicket" was recognized as an Intent and there is an IntentScore representing the confidence level.


#DONE! Next, let the Virtual Agent act based on the the user intention!



Please sign in to leave a comment.

Was this article helpful?
2 out of 2 found this helpful