[EAS] Why we need Text-To-Speech

[Sean Donelan]

On Mon, 19 Mar 2012, Dave Turnmire wrote:
> This HUMAN doesn't know how to pronounce the first or last example of
> above... and I'm afraid to try pronouncing the second...

Yep, that's why I said Richard was taking a better approach by not making
it about people.  Whether its a professional, trained emergency management 
official, an experienced state governor, professional broadcast announcer;
they could all sometimes make mistakes.  That's why there are so many 
blooper videos on Youtube.

If you listen to the original sound file Richard posted, the announcer
was fine.  It was a problem with the transmission/reception.

If you expect the emergency management official to adjust the text to 
work with TTS before sending the message so the TTS software does the 
right thing, you really need a way for the message originator to both 
proof-read and proof-listen to the message unless you don't mind the 
occasional weird result.  Training isn't a substitute for hearing the 
result.  That's just setting emergency management officials up for trouble 
when the TTS mangles a message that they could have fixed, if they had
been able to review it before sending.

I worked on DECtalk software (the same TTS previously used by weather 
radio) for libraries for the blind and physically handicapped throughout 
the 1990's. Although TTS systems have improved, customized speech 
dictionaries are not yet a replacement for proof reading and 
proof listening.