voice-push.com

Pushing VoiceXML to the masses!



Home

Blog

VoiceXML
VoiceXML 2.0 & 2.1
VoiceXML 3.0
State Control XML
ASR & TTS
VoiceXML Applications

Video & VoiceXML
Video Apps

VUI vs. GUI
Client vs. Server Apps

Voice User Interfaces
DTMF vs. ASR
Target Audience
Dialog States
Global Commands
Zeroing Out
Personality
NLU vs. Directed Dialogs
Prompts - Wording
Prompts - Snippets
Prompts - Randomising
Prompts - Recording
Grammar Design
Waiting
Error Handling

Project Phases
User Requirements
Technical Spex
VUI Specifications
Development
Going Live!

Links

Contact

Snippets

Try and think in terms of reusable prompts or snippets. Don’t record prompts like this:

Welcome to our news desk! For sports, press 1. For weather, press 2. For traffic, press 3. And for help, press 0.

This should be broken up into seperate prompts:

Prompt Name Text
welcome.wav Welcome to our news desk!
sports.wav For sports,...
weather.wav For weather,...
traffic.wav For traffic,...
help.wav And for help,...
press1.wav ..., press 1.
press2.wav ..., press 2.
press3.wav ..., press 3.

Now you can be more flexible. If the customer decides that they want to rearrange the order in which the prompts are played then you can easily put them together in a new way - without having to re-record everything!

To really be on the safe side, it’s advisable to record each item in two positions in the list - as a normal item and as the last item:

Prompt Name Text
sports.wav And for sports,...
weather.wav And for weather,...
traffic.wav And for traffic,...
help.wav For help,...

Now it may be decided to drop the help prompt altogther and to put traffic first, followed by weather, followed by sports. No problem:

Welcome to our news desk! For traffic, press 1. For weather, press 2. And for sports, press 3.

Obviously this idea of snippets is particularly useful when it comes to recording dates, times and quantities. However, if you really want it to sound good, then you have to remember one more thing: pitch! What’s that I hear you ask - well it’s whether the voice is rising or falling and it’s covered in the next section.

Open & closed snippets

How a word is stressed depends on where it comes in a sentence. If a word is at the end of a sentence then it is normally closed - i.e. the pitch or tone of the speaker’s voice drops at the end of the word to indicate that the sentence is finished. Unless of course it’s a question, in which case the speaker’s tone actually rises. If a word is in the middle of a sentence then the pitch depends on what the following word is. Words at the start of a sentence also have a different intonation.

So when you’re recording your snippets, you have to ask yourself where they come in a sentence. In the tables above you will notice that many prompts have ‘...’ before or after them. This is to indicate that the sentence isn’t over and that the intonation should reflect that accordingly.

One way of assuring this, is to get the voice actor to say the whole sentence and have the sound engineer cut in into snippets. However, if you’re voice actor is good enough, there shouldn’t be any need to do this, as they should know how to make the snippet sound open or closed.



If you have any comments, ideas, issues, etc. about this topic why not try the voice-push forums






© voice-push.com