Skip to main content
The Yes/No Intents (Beta) provides a machine learning-powered feature to detect yes and no responses similarly to traditional Intent matching, enabling a more natural and fluent conversational experience. Training a model with a small set of example sentences recognizes various yes and no phrases, enhancing an AI Agent’s ability to understand user confirmations or denials. The Question, Optional Question and Code Nodes support the Yes/No Intents feature.

Key Benefits

  • Improved Natural Language Understanding. Unlike the earlier Confirmation Words feature, which required manual synonym definitions, Yes/No Intents use a preinstalled vocabulary that can be expanded. This approach makes it easier to detect diverse expressions.
  • Enhanced User Experience. Yes/No Intents enable AI Agents to respond accurately to a broader range of affirmative or negative responses, reducing misunderstandings and improving conversations.

Restrictions

  • Yes/No Intents don’t support the Intent Trainer.
  • Yes/No Intents don’t provide training traffic lights feedback, which may limit visibility into model performance.

How to Use

To set up this feature, do the following:
Each Locale has its own Yes/No Intents model, requiring separate training. The Universal Locale includes English examples, but additional examples should be added for better accuracy in other languages.To configure a locale, do the following:
  • GUI
  • API
  1. To see the Project’s configured locales, click Manage > Localization.
  2. Select the locale and click ellipsis, then Edit Yes/No Intents.
  3. In the Edit Yes/No Intents pane, you can configure Intents for the current locale:
    • Yes Intent — by default, the setting is toggled on. When the setting is toggled off, the Intent is excluded from the build.
    • No Intent — by default, the setting is toggled on. When the setting is toggled off, the Intent is excluded from the build.
    • Reject Intent — by default, the setting is toggled on. You need to add examples manually. When the setting is toggled off, the Intent is excluded from the build.
  4. Additionally, for Yes and No Intents, you can add rules.
  5. Click Save & Build to build a model. The Yes/No model is always built when you edit and save settings.
Select how you want to apply the Yes/No Intents feature: to the entire Project or to the selected Flow.
  • Project-Level Settings
  • Flow-Level Settings
To use Yes/No Intents for all Flows, do the following:
  1. In the left-side menu of the Project, go to Manage > Settings.
  2. On the Settings page, click NLU Settings.
  3. In the General Flow Logic section, select one of the following options from the Yes/No Logic list:
    • Yes/No Intents with extended rules — evaluates the standard Confirmation words logic first. If the standard Confirmation Words logic fails to determine the input type, the Yes/No Intents model is evaluated.
    • Yes/No Intents — evaluates the Yes/No Intents logic for every user input. The standard Confirmation Words logic is only used as a fallback when no Yes, No, or Reject Intents from the Yes/No Intents model are triggered.
  4. If you selected Yes/No Intents – with extended rules or Yes/No Intents, configure Yes/No threshold. The confidence threshold slider sets the minimum score required for Yes/No Intents to trigger. A Yes, No, or Reject Intent will only be triggered if the Intent score is greater than or equal to this threshold. This value applies to Yes/No Intents models for all locales. You can change this value for a specific Flow.
  5. Click Save to apply settings and build a model. The Yes/No model is always built when you edit and save settings.
If you have set settings at the project level but want to change them for certain Flows, you should go to the Flow-level settings.

How to Test

Check if the Yes/No Intent feature works as expected via the Interaction Panel:
  1. In the existing Flow, open the Interaction Panel by clicking interaction-panel Chat with your Agent in the upper-right corner of the page.
  2. Run a Flow.
  3. On the INFO tab, select Input to view JSON. In the yesNoIntentResults object, you will see the result of Yes/No Intents model evaluation.
  • Yes Intent
  • No Intent
  • Reject Intent
  • Fallback Reject Intent
"nlu": {
  "yesNoIntentResults": {
      "finalIntentName": "yesIntent",
      "finalIntentScore": 0.8887347172756052,
      "scores": [
            {
            "id": "a396391d-3cab-43ab-8b8a-cefd0135041b",
            "name": "yesIntent",
            "score": 0.8887347172756052
            },
            {
            "id": "fa2ab1d3-b5cd-4021-91c6-8b642ec51dd8",
            "name": "noIntent",
            "score": 0.32411091251643426
            }
      ]
    }
},
"mode": "TextOnly",
"type": "pAnswer"
Q1: What happens if the Yes/No logic is set to Yes/No Intents, but both yes and no Intents have been individually disabled in the Locale settings?A1: The standard Confirmation Words logic is used.Q2: Are Yes/No Intents independent of standard intent scoring?A2: Yes, the Yes/No Intents are completely independent of standard intent scoring (input.intentScore or input.nlu.intentMapperResults). They specifically find the types nAnswer and yAnswer within the Flow. These types are used for the Yes/No type in Question Nodes and for confirming Intents, similar to confirmation words.Q3: How are Yes/No Intents evaluated when the Execute Flow node is used with and without the Parse Intents toggled on?A3: Yes/No Intent scoring behavior differs only if the Yes/No Intent settings are different in the two Flows.For example:Prerequisites:
  • Flow A has Yes/No Intents and only one node, which is an Execute Flow node that executes Flow B.
  • Flow B has Yes/No Intents - with extended rules and one Say node.
Scenario 1:
  1. The Parse Intents setting is disabled in the Execute Flow node of Flow A.
  2. When the user sends yes in Flow A, Yes Intent is triggered in Flow B.
  3. Flow B inherits the Yes/No Logic from Flow A.
Scenario 2:
  1. The Parse Intents setting is enabled in the Execute Flow node of Flow A.
  2. When the user sends yes in Flow A, Yes Intent is not triggered in Flow B.
  3. Flow B uses its own Yes/No Logic.

Example

Imagine a user interacting with an AI Agent in a Star Wars quiz: Without Yes/No Intents:
User: Hello! 
AI Agent: Do you want to play the Star Wars quiz?
User: Why not
AI Agent: Not sure I understood this correctly. Try again.
With Yes/No Intents:
User: Hello! 
AI Agent: Do you want to play the Star Wars quiz?
User: Why not
AI Agent: Let's start!

More Information

I