Manage Continuous Stream Transcription Profiles

Required permissions: Automation & AI > Launch Automation & AI Studio Script View, Create/Edit

Continuous Stream Transcription provides transcription of conversation audio. It sends the transcription as a continuous stream throughout the interaction.

Add a Continuous Stream Transcription Profile

A transcription profile defines the transcription options you want to use, including the language to transcribe.

  1. In CXone Mpower, click the app selector and select Other > Automation & AI > Transcription Hub > Continuous Stream.
  2. Click Add Transcription Profile.
  3. Enter a Transcription Profile Name. To differentiate this profile from your other profiles, use a name that highlights what makes this profile different. For example, you could name the profile for its intended use, the provider, or the language of the profile. If you use more than one transcription provider, make sure the profile name includes the provider.
  4. Under Select Provider, click the transcription provider you want to use. You can choose from CXone Mpower Transcription, CXone Mpower Transcription for Copilot, Google Transcription, or Microsoft Transcription. Some of these options must be enabled in your CXone Mpower system by your Account Representative.
  5. Click Next.

  6. On the Configuration page, complete the fields to define the transcription options for this profile.

    Field Details
    Language Code Select the language that you want this profile to transcribe. Currently the only available language is English.
    Interim Results Click to enable this option if you want the transcript to show temporary results that are refined over time. The default is for this option to be disabled. Do not enable this option for Salesforce Assist or Copilot for Agents. They require final utterance results.
    Field Details
    API Key

    Enter the API key for Transcription for Copilot. Your Account Representative provides this key for you.

    Language Code Select the language that you want this profile to transcribe.
    Model

    Select the machine learning model that you want the transcription service to use when transcribing the audio. These models have been trained to recognize speech from specific audio types and sources. The available models depend on the Language Code you selected.

    You can learn more about the available modelsAn icon of a square with an arrow pointing from the center out to the upper right corner..
    Smart Format

    Enable this option to have transcripts automatically formatted for improved readability. This option applies formatting to the transcripts. This includes formatting such as punctuation, paragraph breaks, and using numerals in place of the text versions of the number, for example using 2 instead of two.

    Punctuate This automatically enables when you enable Smart Format. It cannot be disabled.
    Interim Results Select this option if you want the transcript to show temporary results that are refined over time. The default is for this option to be disabled. Do not enable this option. Transcription for Copilot requires the final utterance results.
    Find and Replace

    Use this section to define words or phrases that you want replaced with different text in the final transcripts. For example, you could replace examples of foul language with a different word, such as bleep.

    To add a find and replace configuration: 

    1. Click Add Find and Replace.
    2. In the Name field, enter the word as it appears in the transcript.
    3. In the Value field, enter the text you want to replace the word specified in Name.
    Keywords

    Use keywords to boost or suppress certain words or phrases:

    • Boosting assigns more weight to the word or phrase. The transcription service is more likely to choose that word or phrase when the audio match is close.
    • Suppressing assigns less weight to the word or phrase. The transcription service is less likely to choose the word or phrase.

    Note: Suppressing keywords is only available for base Models.

    To add keyword boosting or suppression: 

    1. Click Add Keywords.
    2. In the Name field, enter a word or phrase.
    3. In the Value field, enter a positive integer from 1 to 10 to boost the specified word or phrase, or a negative integer from -1 to -10 to suppress it.
    Field Details
    Transcription Type

    Choose the Google Transcription account to use with this profile: 

    • Select CXone to use the account owned by CXone Mpower.
    • Select Your Account to use your organization's account.

    JSON Service Account

    The service account key for the Google Transcription account you're using:

    • If you're using the CXone Mpower account, use the key provided for you by your Account Representative.
    • If you're using your own account, use the key file you created and downloaded from your Google Cloud Console.

    Copy and paste the entire contents of the key file into this field, including the curly braces surrounding the key.

    The service account key is a JSON object similar to the following example:

    {

    "type": "service_account",
    "project_id": "ic-demo2-aai-gcp",
    "private_key_id": "sdv80sdf08ssd9sdf0sdf8sdfsdf90sdf8",
    "private_key": "-----BEGIN PRIVATE KEY-----
    nMIIEvgIBADANBgkqhkiG9w0BAQEFAASCBKgwggSkAgEAAoIBAQCzgrS8vp
    <private key has been truncated for this example>
    ----END PRIVATE KEY-----\n",
    "client_email": "your@client_email-gcp.iam.gserviceaccount.com",
    "client_id": "123456789",
    "auth_uri": "https://accounts.google.com/o/oauth2/auth",
    "token_uri": "https://oauth2.googleapis.com/token",
    "auth_provider_x509_cert_url": "https://www.googleapis.com/oauth2/v1/certs",
    "client_x509_cert_url": "https://www.googleapis.com/robot/v1/metadata/x509/
    dialogflow-dthjop%40ic-demo2-aai-gcp.iam.gserviceaccount.com"
    }
    Language Code Select the language that you want this profile to transcribe.
    Model

    Select the machine learning model that you want the transcription service to use when transcribing the audio. These models have been trained to recognize speech from specific audio types and sources. Not all languages support every model. Refer to the Google documentation for information about language support.

    Available options are: 

    • Command and Search: Use when transcribing shorter audio clips, such as commands or voice search. Consider using Latest Short instead.
    • Default: Use when the audio doesn't fit the other models. It can transcribe any type of audio, but results may be lower quality for some types.
    • Latest Long: Use with long-form content such as spontaneous speech or conversations.
    • Latest Short: Use when the utterances are a few seconds long, such as commands. Consider using this model instead of Command and Search.
    • Phone Call: Use for transcribing the audio from a phone call.
    • Telephony: Use for higher quality audio in a phone call.
    • Telephony Short: Use for short conversations (less than ten seconds long) in a phone call.

    Sentiment Required

    Click this option if you're using sentiment next best actions in Salesforce Assist. When enabled, complete the following fields that appear:

    • Conversation Provide ID: Enter the ID of the conversation profile from Google Contact Center AI. You must create a conversation profile while setting up Google Contact Center AI with Google. It determines the suggestions made to agents during interactions.
    • Region: Enter the region where your Google Cloud Project is hosted.
    Use Enhanced

    Enable this option to use the enhanced version of your selected Model. This is only an option if you are using the Phone Callmodel. This can help increase the accuracy of transcription.

    The default is for this option to be disabled. Enabling this option increases the cost of transcription.

    Interim Results Enable this option if you want the transcript to show temporary results that are refined over time. The default is for this option to be disabled. Do not enable this option for Salesforce Assist or Copilot for Agents. They require final utterance results.
    Enable Automatic Punctuation Enable this option to use the Google Transcription automatic punctuation option. When enabled, the transcription service automatically infers where punctuation such as commas, periods, and question marks should go and adds them to the transcript.
    Class Token

    Add Google class tokens to improve the accuracy of transcription:

    1. Under Class Token, enter the token name as the Value.

    2. In the Boost field, enter a number between 1 and 20 to assign a weighted value to the specific token. Higher numbers provide more of a boost.

    3. Click Add Class Token to add another token and boost value.
    Field Details
    Subscription Key

    Paste in the subscription key for your Azure AI Speech Service. This key allows you to authentication with your Speech Service subscription.

    To find your subscription key, log in to the Azure portal and search for your speech service resource. The subscription key is located on the Keys and Endpoint tab.

    This field is required.
    Service Region

    Enter the region of the Azure data center that hosts your speech service resource. For example, westus2. The region ensures your transcription requests are routed to the correct location. It's also used for authentication along with the subscription key.

    You can find the region in the Azure portal.

    This field is required.

    Endpoint ID

    Complete this field if you're using a custom speech model created using Azure Speech Studio. Paste the endpoint of your custom model in this field. Your custom model must be deployed in Azure Speech Studio before you add it to a transcription profile in CXone Mpower.

    If this field is left blank, Microsoft Transcription uses the default baseline speech recognition model.

    Language Code

    Select the language that this transcription profile will transcribe from the drop-down list.

    This field is required.
    Interim Results Enable this option to have the transcript show temporary results that are refined over time. The default is for this option to be disabled. Do not enable this option for Salesforce Assist or Copilot for Agents. They require final utterance results.
    Enable Automatic Punctuation Enable this option to have Microsoft Transcription automatically insert punctuation marks such as periods, commas, and question marks into the transcription. This helps improve readability and clarity.  The default is for this option to be disabled.
  7. Click Add.

Assign a Transcription Profile to an Action

Permissions Required:  Studio Scripts View, Create/Edit

If you're using Continuous Stream Transcription with a virtual agent, you don't need to follow these steps. The virtual agent integration process prompts you to select a transcription profile. You don't need to include the Cloud Transcribe action in scripts for virtual agent integrations.

To use Continuous Stream Transcription for non-virtual agent applications, you need to add the Cloud Transcribe action to your Studio script. You must assign a transcription profile to it. If you are transcribing multiple languages in the same script, you need to add one Cloud Transcribe action for each language. Assign a transcription profile to each action.

  1. In Studio, open the script you're working with.
  2. If the Cloud Transcribe action isn't already in the script, add it to the correct location and ensure that it's properly configured.
  3. Double-click the Cloud Transcribe action to open the Transcription Profiles page.
  4. Locate the transcription profile you want to use or create a new one.
  5. Click the Selection icon next to the profile name.
  6. Click Yes.

Add Class Tokens to a Google Transcription Profile

Class tokens allow you to enhance transcription for improved accuracy. It's only an option if you use Google Transcription service. Follow these steps to add class tokens to an existing profile. If you are creating a new profile, follow the steps on this page to add a Continuous Stream Transcription profile.

  1. In CXone Mpower, click the app selector and select Other > Automation & AI > Transcription Hub > Continuous Stream.
  2. Click the transcription profile you want to modify.

    1. Under Class Token, enter the token name as the Value.

    2. In the Boost field, enter a number between 1 and 20 to assign a weighted value to the specific token. Higher numbers provide more of a boost.

    3. Click Add Class Token to add another token and boost value.

  3. Click Save.

Test the Connection to Google Transcription

If you use Google Transcription Service, you can test the connection between it and CXone Mpower. This allows you to verify that the transcription profile is correctly set up. You can also use it during troubleshooting. It can help you when your transcription provider and CXone Mpower aren't connecting. Testing is not available for other transcription providers.

  1. Click the app selector icon of app selector and select Other > Automation & AI and then click Transcription Hub.
  2. Click Continuous Stream Transcription and select the transcription profile you want to test.
  3. Click Test Connection at the top of the page. A toastClosed Small, temporary pop-up messages on the bottom of the screen indicating things like errors and successes. message displays the results of the test. If the test fails, check your account details and repeat the test.