Share via


Exchange 2013 UM Features and Architecture

Voicemail Preview

- Due to UCMA 4.0 and the Microsoft Speech Platform there are many enhancements to Voicemail Preview

- Voicemail Preview feature uses automatic speech recognition to convert the spoken voicemail into text

- Voicemail Preview is updated in Exchange 2013 to be more accurate and provide better performance

- Transcription now happens to all messages

Thanks to UCMAP 4.0 and the Microsoft Speech Platform (ver 11.0), there are many enhancements to Voicemail Preview. The Voicemail Preview feature was introduced in Exchange 2010 and uses automatic speech recognition (ASR) to convert the spoken voicemail into text. Voicemail Preview is beaning updated in Exchange 2013 to be more accurate and provide better performance. Transcription now happens to all messages even if they are over 75 sec unlike previous versions. If a message is over 75 sec, it is transcribed, but the text for the message isn't included

Call Answering Rules

- Call Answering Rules were first introduced in Exchange 2010 , These rules can be setup based if you are on vacation or if a specific caller is trying to reach you, you can have UM forward the call to a different number as a few examples. Parts of this feature have changed with UM 2013

- Feature empowered UM subscribers to control their Unified Messaging mailbox with rules

- UM can forward a call to a different number as an examples

- For GA, we will support ‘personas’ – persona is abstract representation of the various contacts for one person. Most like the contacts linking

Enhanced Caller ID support

- Previously if you were not in the same dial plan and someone called and left a voicemail you would get "Voicemail from Anonymous Caller“

- Exchange 2010 sp1 made some improvements on the heuristics of matching name to number

- Exchange 2013 looks in the default personal contacts folder and also in user and system created personal contacts folders

- This is important since Exchange 2013 supports Unified Contact Store with Lync server

 

Enhancements to Speech Platform and Speech Recognition

- In Exchange Server 2013 the speech GAL grammar generation no longer happens on the server

- Generation happens periodically using the Mailbox Assistant, on the Mailbox server running the Microsoft Exchange Unified Messaging

- GAL speech grammar file is stored in the arbitration mailbox for an organization

- Grammar file downloaded to all Mailbox servers

- Mailbox Assistant runs every 24 hours

Exchange 2013 now uses the Microsoft Speech Platform version 11.0. This helps to provide a more accurate Voice Mail Preview experience for subscribers. It also helps with the global address list (GAL) grammar generation that has been changed with Exchange 2013. This GAL grammar is used to help determine spoken commands and names of contacts in the users mailbox and in the organization.

In Exchange Server 2013, to address ASR grammar generation scalability issues for UM, the speech GAL grammar generation no longer happens on the server with the Unified Messaging server role installed. Instead, it happens periodically using the Mailbox Assistant, on the Mailbox server running the Microsoft Exchange Unified Messaging service that hosts the organization’s arbitration mailbox. The GAL speech grammar file is stored in the arbitration mailbox for an organization and then later downloaded to all Mailbox servers in that Exchange organization. By default, the Mailbox
Assistant runs every 24 hours

Protect Voicemail using AD RMS

- Prevent forwarding of voicemail

- Integration with AD RMS and Exchange unified messaging

- Permissions designated by sender (by marking the message as private) or by administrative policy

Exchange 2013 UM Architecture 

  

If you notice here, the connection points from CAS to MBX is always going to be the same protocol that the client uses to connect to CAS.

CAS will proxy requests from clients on the appropriate protocol to the appropriate back-end MBX server.

For SMTP- once the CAS decides to accept the message, the SMTP session is then proxied.

One unique difference is Exchange UM. The nature of real time traffic in the RTP media stream does not lend itself to a proxy scenario. The signaling traffic (SIP) is used to redirect
the RTP stream to the back-end mailbox server

?All calls are front-ended by CAS servers ?Point SIP Trunks to CAS (or CAS Load Balanced FQDN), not Mailbox Servers ?Unified Messaging role is now a service on the Mailbox Server. 

?Unified Messaging Call Router Service running on CAS redirects the call to a specific Mailbox server, we put UM direct to Mailbox role is because the media stream should be fast to delivery

 

Basic Unanswered Call Flow Diversion to UM

Comments

  • Anonymous
    May 02, 2014
    Are you serious? Have you seen the output of voice transcription? It's terrible and almost never accurate.
  • Anonymous
    September 11, 2014
    "Transcription now happens to all messages even if they are over 75 sec unlike previous versions. If a message is over 75 sec, it is transcribed, but the text for the message isn't included"

    Then why waste the resources to transcribe it?
  • Anonymous
    September 29, 2014
    Does transcription happen for UM only or it can happen if voice message is submitted via SMTP/EWS ?
    Thanks.
  • Anonymous
    October 14, 2014
    Question, does this mean that the 2013 UM server can redirect to 2010/2007 UM servers if the users mailbox is hosted within 2010/2007 ?
  • Anonymous
    January 20, 2015
    こんばんは。Lync サポートの久保です。
    Lync と Exchange UM の連携、第二回目です。 第一回はこちら 。
    Lync Server 2013 と Exchange 2013
  • Anonymous
    March 17, 2015
    The comment has been removed