When working with natural language from our users, it is sometimes necessary to dive further into their statement to understand what their true intent is. In addition, humans have a tendency to communicate in ways which might include multiple intents in the same statement. In both of these cases, we can use disambiguation to clarify our user's intents and ensure that their needs are being addressed appropriately.

Watch the video


Step 11: Demonstration of need

This step serves to demonstrate a use case for when we might need to employ disambiguation.

  1. From the Conversational AI section of the Conversational Cloud, navigate to Intent Builder and select the Getting Started Domain.

    The Getting Started Domain should contain both an Order Status and a Billing Question intent.

  2. Use the debugger to test out some utterances. In the Test User Input panel on the right, enter the text “I want to know about my bill and recent order”. Make sure that the Search in domain toggle is enabled, and click Test.

  3. Note that this input results in a FAIR PLUS score for both of our intents. In this case, the user would be directed to our Fallback dialog despite the fact that there is some confidence that their utterance matches multiple intents. This example clearly demonstrates the need to drill deeper with the user and provide a pathway to find resolution.

Step 12: Implement a Disambiguation dialog

In recognizing the need to provide this level of consideration to our users, Conversation Builder comes equipped with a special type of dialog called a Disambiguation dialog.

  1. Navigate back to Conversation Builder, and select your Getting Started bot.
  2. Click Add Dialog in the lower-left corner, and fill the resulting Add Dialog form with the following:
    • Dialog Name: Disambiguation
    • Dialog Type: Disambiguation Dialog
    • Match Threshold: Fair Plus
    • Disambiguate only selected domains: Getting Started Domain

    The Disambiguation dialog is a special type of dialog that serves the single purpose of providing disambiguation for a user’s utterance. Like the Fallback dialog, Disambiguation dialogs are limited to one per bot.

    Click Save.

    The resulting Disambiguation dialog provides a new interaction type specifically for handling disambiguation. This is displayed much like a multiple choice interaction, with the intents that are being disambiguated representing the button options.

  3. Click into the interaction text and change the display text to read “Which can I assist you with?”

  4. In the upper-right corner of the interaction, click the Settings icon. This takes you to the Interaction Settings window. Select the Advanced tab to further customize the intents that will be displayed. Here, you can select either two or three intents to display and an optional, static label that can be useful for a “Main Menu” or escalation message. For this demonstration, leave all the default values.

  5. Click Preview to test this disambiguation dialog using the same phrase that you used earlier in Intent Builder. Reset the session, and type “I want to know about my bill and recent order”. This should result in a multiple choice display of the two intents that are being disambiguated.

  6. Selecting either option should result in the appropriate dialog being started for the user.

    While made up of a single interaction, the Disambiguation dialog goes a long way in ensuring that users are being directed to the correct solution. Every bot solution should include a way to drill deeper into the user's intent so that you are always pushing them towards a solution.

What's next?

Continue on to the next tutorial in the series.