Speech to Text is an optional, licensed feature of NuPoint UM that converts voice mail messages to text, allowing users to discreetly access voice messages in a text format. This feature uses a third-party transcription service that requires Internet connectivity. Your DNS configuration must allow for transcription requests being sent to the Speech to Text provider, Nuance Communications.
Speech to Text service allows you to:
Discretely
check voice messages
When in meetings or in social settings, it is simply awkward to listen
to voice messages. This situation is especially frustrating if an
important message is expected. The STT feature allows you to "read"
new voice messages virtually anywhere.
Quickly
find important voice messages
In most cases, you can scan a transcribed message faster than you can
listen to the original voice message. After performing this quick
review, you can listen to the voice message to catch nuances of expressions
and confirm key details.
Easily
respond to voice message with an email
After reading a transcribed message, you can “reply” to it by forwarding
it to email. This eliminates the need to "re-cap" the voice
message by typing it into the email message.
Notes:
Administrators and users must agree to the terms of the transcription service provider's licensing agreement. Part of this agreement specifies that the transcription process may involve human beings listening to your recorded audio messages.
The Speech to Text feature must be configured using the Web Console. (Exception: The Text Console can be used to configure STT for UM Standard.)
Although the transcription process is fully automated, voice messages are kept for 24 hours during which random spot checks are performed to assess quality. All data (voice message and transcriptions) are then electronically "shredded" and retained in a library for research purposes. This is analogous to a physical paper shredding process—audio and transcriptions from all sources are "cut up" and mixed up together so that a message cannot be pieced together again.
If an account balance lapses, transcription requests are denied and alarms are raised. NuPoint UM disables the Speech to Text feature until the administrator corrects the account balance and activates the feature again.
There may be limitations on the type and/or length of messages being transcribed. If transcription cannot be performed, an alternative email is sent to notify the user.
Nuance does not provide language detection and/or translation services. For this release, messages received in languages other than North American English will not be transcribed.
The maximum message length is 60 seconds. For longer messages, users can dial in and listen to the entire original voice mail. (If necessary, your Mitel reseller can contact the transcription service provider to extend the maximum message length for your account.)
The Speech to Text service uses a complex computer model to transcribe spoken words into text. While transcription accuracy can be very good, there may be instances when the transcription does not accurately represent the spoken message. If in doubt, please listen to the original voice message.
If the system cannot understand the message, it will not transcribe it. Instead, the user will receive an email notification. Typically, about 15% of messages are deemed "untranscribable" due to non-English accents or language, jargon, background noise or other issues.
When Speech to Text is enabled, users receive a text transcription email for each of their voice mail messages. You can add a copy of the audio file by enabling "Include audio attachments with email messages" in the system-wide settings.
For Advanced UM, only email messages containing audio attachments are synchronized with the user's voice mailbox. Emails containing text transcriptions are not synchronized and do not affect the status of the original voice mail or MWI.
A message number (ID) is contained within the body of email messages. When two emails are generated for a single voice mail, they will have the same message number.
Speech to Text is supported in International English only, and in North America only.
Confidential, Record-a-Call, and Fax messages are not transcribed.
Text transcription emails can be handled like other emails (mark as read, forward, reply, etc.) but operations performed on them are not synchronized with the user's NuPoint voice mailbox.
Users must accept the Speech to Text license agreement in Web View before the feature can be used. Administrators cannot accept the license on the user's behalf.
The transcription service provider is Nuance Communications, Incorporated.
You must contact your Authorized Mitel Reseller in order to set up an account and purchase conversion credits from Nuance. When the credits run low or are depleted, NuPoint UM will begin generating alarm messages. You must then purchase more conversion credits from your reseller.
Because it takes time for the system to transcribe voice messages to text, email with text transcriptions will arrive after emails with audio attachments. Delays of 10 minutes or more can be encountered depending on the length of the voice messages and the number of transcription requests the system is currently processing. Internal corporate mail routing delays may also affect delivery time. The transcription process rarely takes less than two minutes.
Standard Messages
A typical transcribed message looks like this:
"Hi there.
I was just returning your call about dinner at the weekend. Not very sure.
I'll let you know by tomorrow what it’s looking like. Bye."
Spoken through Nuance.
The quotation marks delineate the contents of the original voice mail message.
Messages with Special Characters
Special characters contained in the body of the message indicate the following:
Question Mark: Indicates that the service cannot understand the complete word. It therefore spells the word phonetically. For example, "Hi Ashby(?), are you still working? Bye."
Underscore: Indicates that the service cannot understand any part of the word. It therefore replaces the word with an underscore. for example, "Hi _ . My number is 777777777. I repeat 777777777. Thank you."
Unconverted Messages
The transcription service will not convert messages for the following reasons:
Too much background noise
Speaker's accent not understood
Message was recorded in a language other than English
Message spoken too fast
Message contains too many special characters indicating that text has been replaced or spelled phonetically
Message is a fax
Message is marked "confidential"
If any of these conditions occur, NuPoint UM users will receive an email instructing them to call the voicemail system to check the original message.