Continuous Stream Transcription provides real-time, self-service speech-to-text
Also called STT, this process converts spoken language to text. transcription. It's for use when you need audio transcription sent in a continuous stream throughout a conversation. It's ideal for use during interactions with a live agent. It can also be used during interactions with a virtual agent A software application that handles customer interactions in place of a live human agent..
Continuous Stream Transcription transcribes a conversation as it happens. It sends the transcript in real time to the destination application with only a minimal delay. This allows the transcript to be available for use almost instantly. For example, if the agent application displays the transcript, agents can use it to confirm details with the contact The person interacting with an agent, IVR, or bot in your contact center., such as an address or account number. The transcript includes both the agent and contact sides of the conversation.
CXone Mpower offers two native options for Continuous Stream Transcription, CXone Mpower Transcription and CXone Mpower Transcription for Copilot. CXone Mpower also supports third-party transcription services.
CXone Mpower supports another type of transcription. Turn-by-Turn Transcription provides real-time transcription that's sent in segments based on turns in the conversation.
Use Cases for Continuous Stream Transcription
You can use Continuous Stream Transcription to:
- Provide real-time transcripts for live agents, such as with Salesforce Assist.
- Provide summaries of interactions for live agents, such as with CXone Mpower AutoSummary.
- Detect the contact's sentiment Overall mood or result of the interaction as determined by analysis of words, phrases, and context of the transcript. and suggest the next best action the agent can take with Salesforce Assist.
- Provide interaction transcripts for agents in CXone Mpower Agent Suite applications.
- Export transcripts outside of the CXone Mpower platform.
- Provide transcripts to supervisors in CXone Mpower Supervisor.
- Transcribe voicemail messages. This option can be set up with custom scripting.
- Transcribe contact utterances State that allows an agent to complete work requirements after finishing an interaction. for voice virtual agents A software application that handles customer interactions in place of a live human agent. that use the Voicebot Gateway.
Supported Third-Party Transcription Services
In addition to Transcription, CXone Mpower supports the following third-party providers of Continuous Stream Transcription:
- Google Transcription: Available in all languages that Google Transcription Service
supports. - Microsoft Transcription: To use this option, you must have an Azure subscription with speech service enabled. You must also have access to the Azure portal. Microsoft Transcription is available in all languages that Azure speech service supports .
Check with your transcription service provider to find out what regions the service supports. For some transcription providers, you can configure the service to customize the transcription results.
Setup for Continuous Stream Transcription
To use Continuous Stream Transcription, you must create a transcription profile in Transcription Hub. The profile specifies which transcription service to use and the language to transcribe. Your Studio script must include at least one Continuous Transcription action. You must configure this action with the appropriate transcription profile. This connects the transcription profile to your script.
You can turn transcription on and off as your script runs so you only transcribe audio when you need it. To do this, you must have two Continuous Transcription actions in your script.
To transcribe multiple languages in the same script, you must create a transcription profile for each language. Each profile needs to have its own instance of Continuous Transcription. Only one Continuous Transcription action can run at a time in a script. This means only one language can be transcribed at a time.
If you're using Continuous Stream Transcription with a virtual agent, you must also set up the virtual agent to use the Voicebot Gateway.
Load or Change Behaviors for Continuous Stream Transcription
CXone Mpower can change how it handles Continuous Stream Transcription during an interaction. This is typically done when Continuous Stream Transcription is used with virtual agents. You can set behaviors to handle silence, interrupting, timeouts, and so on. Many of these behaviors are designed for use with virtual agents A software application that handles customer interactions in place of a live human agent..
Behaviors that affect Continuous Stream Transcription can only be set before transcription starts. This means that these behaviors must be configured before the CONTINUOUS TRANSCRIPTION action runs. You can use a Script Parameters Studio code snippet to pass data to CXone Mpower to configure transcription behaviors. If you're working with a script for a virtual agent that uses Voicebot Gateway, use the Default Next Prompt Behaviors snippet instead.
If you need to change a transcription-related behavior on a turn-by-turn basis, your script must be set up to start and stop transcription for each turn.
Google Transcription Class Token Support
You can use class tokens with Google Transcription service. Class tokens allow you to identify words or phrases as belonging to a particular class. This can improve the accuracy of the resulting transcription. For example, you can use a class token to inform the transcription engine that the string of numbers in the phrase "my address is 123 Oak Street" is an address number. You can also add a boost value to increase the chances of Google Transcription using the specified class token. Boosting does not guarantee that the service applies a class token accurately.
To use class tokens, you need to add them to the transcript profile in Transcription Hub. When CXone Mpower sends audio to Google Transcription, it includes the class tokens and the transcription service applies them. You can add class tokens to an existing profile or when you create a new profile.
Class tokens change the behavior of the transcription service. Just like other transcription behaviors, class tokens must be configured in the script before the CONTINUOUS TRANSCRIPTION action runs. Class tokens can be applied or changed after transcription has started, but you must set up the script to handle it.
You don't need to configure anything in your Google Transcription management console for this option to work. Refer to the Google documentation for more information about this feature and the supported tokens .
Google does not return an error if the token name was entered into the Transcription Hub profile incorrectly. The only indication of a potential problem is that the transcription results aren't affected by the token as expected.