Genesys Cloud - Main

 View Only

Sign Up

  • 1.  Copilot Transcription Accuracy – Should Trunk Language Match Dialect?

    Posted 03-19-2026 05:03
    Edited by Phaneendra Avatapalli 03-19-2026 05:11

    Hi everyone,

    We’re currently reviewing transcription accuracy for branded terms (e.g., “Okta” being transcribed as “Octa”). We’re seeing this behaviour both in the interaction transcript and in Copilot live transcription (which I understand use the same transcription layer).

    We’ve identified that our trunk language is configured as en-US, while Dictionary Management is configured as en-AU. We’re assessing whether dialect alignment may be contributing to these transcription inconsistencies.

    Before making any trunk-level changes, we’d appreciate insights from other organisations:

    • Are you typically keeping the trunk language aligned with your primary spoken dialect?
    • Or are you running a different trunk dialect and tuning recognition via Dictionary Management and boost values?
    • Have you seen measurable differences in transcription accuracy when trunk language matches (or does not match) your spoken dialect?

    Interested to understand best practices when using Copilot together with Speech & Text Analytics.

    Appreciate any experiences or recommendations.



  • 2.  RE: Copilot Transcription Accuracy – Should Trunk Language Match Dialect?
    Best Answer

    Posted 03-19-2026 06:08

    Hi Phaneendra, 

    You're onto something with that dialect mismatch. You should be looking to have your trunk match your actual spoken dialect. Having en-US on the trunk but en-AU speakers is likely causing those "Okta" > "Octa" issues.

    What some customers do is align everything to the dialect people actually speak (en-AU for you) and then use Dictionary Management with values to handle tricky branded terms.

    Quick wins before touching the trunk:

    • Override at the queue level with en-AU (safer to test - doesn't impact everything at once and takes priority over trunk settings)
      • https://help.genesys.cloud/articles/set-the-media-language-on-voice-calls/
    • Update "Okta" and other problem words into Dictionary Management
    • Make sure Dictionary Management is set to en-AU



      ------------------------------
      Josh Coyle
      Senior Professional Services Consultant
      ------------------------------



    1. 3.  RE: Copilot Transcription Accuracy – Should Trunk Language Match Dialect?

      Posted 03-19-2026 06:40

      Hi Josh,

      Thanks for the suggestion regarding the queue-level override.

      We've already set the language to en-AU within our In-Queue flow (using the Set Language action), and our Dictionary Management is configured as en-AU.

      I also have "Okta" configured in Dictionary Management, including phonetic variations such as "Octa" and other similar sound-like entries.

      However, we don't currently see a separate Media Language or transcription language override option at the queue level.

      Could you please clarify exactly where this queue-level override can be configured? Is this within the Queue configuration itself, or under Telephony/Trunks or Media Profiles which I don't have access to at the moment.



      ------------------------------
      Phaneendra
      Technical Solutions Consultant
      ------------------------------

      Original Message


    2. 4.  RE: Copilot Transcription Accuracy – Should Trunk Language Match Dialect?

      Posted 03-19-2026 06:52
      Edited by Josh Coyle 03-19-2026 06:52

      Hi Phaneendra, 

      The language can be set at queue level itself within the General tab of the Queue settings, image below:



      ------------------------------
      Josh Coyle
      Senior Professional Services Consultant
      ------------------------------

      Original Message


    3. 5.  RE: Copilot Transcription Accuracy – Should Trunk Language Match Dialect?

      Posted 03-19-2026 07:15

      Hi Josh,

      Ah, wow I completely overlooked that setting under the Queue General tab 😄

      Thank you for pointing that out and for sharing the screenshot much appreciated!

      We'll update a few selected queues to en-AU first and monitor whether this improves transcription accuracy for "Okta" and similar terms before making any broader changes.

      Thanks again for the guidance.



      ------------------------------
      Phaneendra
      Technical Solutions Consultant
      ------------------------------

      Original Message


    Related Content