Skip to main content

Previewing your applications

You often want to preview the changes in your application to validate that they had the desired effect. This can be done using Speechly Dashboard or Speechly CLI.

Speechly Dashboard​

To preview an application:

  1. Log in to Speechly Dashboard
  2. Select your Application
  3. Go to the Preview tab
  4. Press the microphone button, say something and you will receive a transcript in real-time.

If your application uses any NLU features, they will be visible here, too. Please note that the Preview tab only supports Speechly models at the moment, to preview other models use Speechly CLI.

Dashboard preview

Previewing changes​

After you have made changes, deploy the application, wait for the training to finish and then preview it. Please note that you can't preview undeployed changes.

If you want to inspect and review the changes before deploying them to your production application, consider having separate production and development applications. This will insure a clean separation between development and production versions of your application. This creates a bit more work as you need to maintain two application.

Viewing speech segments​

In the Preview tab, you can open the browser developer console to see the JSON response that's emitted for each speech segment. This might be come in handy when debugging.

{
  "id": 0,
  "contextId": "1dce01bf-3380-4690-9847-6cd7ae75cf04",
  "isFinal": true,
  "words": [
    {
      "value": "schedule",
      "index": 2
    },
    {
      "value": "for",
      "index": 3
    },
    {
      "value": "october",
      "index": 4
    },
    {
      "value": "tenth",
      "index": 5
    }
  ],
  "entities": [
    {
      "type": "date",
      "value": "2022-10-10",
      "startPosition": 4,
      "endPosition": 6,
      "isFinal": true
    }
  ],
  "intent": {
    "intent": "schedule",
    "isFinal": true
  }
}

Speechly CLI​

Speechly CLI doesn't have a microphone based preview like Speechly Dashboard, but there are similar commands available:

  • transcribe - Use this command to transcribe audio file(s).
  • evaluate - Use this command to evaluate model ASR or NLU accuracy against a given test set.
Info

See Evaluate ASR accuracy and Evaluate NLU accuracy to learn more about evaluation.