[EAS] Test-To-Speech must be allowed - NOW!

[Sean Donelan]

On Fri, 2 Mar 2012, Marlin Jackson wrote:
> If this list server will allow there are 2 attachments. Both are our 
>RMT's for February from Washington State EOC. One is from our Sage and 
>the other from a Warning Systems Inc (WSI) converter connected to a TFT.

Thanks for the samples.  The reason why I asked for a long, complicated 
example instead of of a simple "this is a test" was the same reason NOAA's
evaluation used long, complex samples.  People can understand/tolerate 
most short synthesized voices, such as voice prompts.  But listening and
understanding a 90-120 second message, even read by a human, is more
challenging.

There isn't one "best" answer.  There are many different ways to measure
text to speech systems, and different techniques work in different 
environments.  For example, the TTS software selected by NOAA a few years
ago scored relatively low in recent rankings by an industry trade 
magazine.
http://www.asrnews.com/TTS_acc_website.pdf

Words per minute.  Human announcers vary between about 120 WPM 
(Walter Cronkite) and about 200 WPM (sports announcers), and use a variety
of techniques to help a listener understand a message. CAP devices seem to 
use around 180 WPM to 200+ WPM for text to speech, although there isn't
any set maximum WPM.  For example, some EAS participants have set the 
visual text crawl to be as fast as possible.  Does there need to be any 
rules or guidance for EAS participants?  Some EAS participants try to 
make it as pleasent and useful experience for their listeners and viewers, 
other EAS participants seem to just want to get it over even if the 
message is difficult to understand.