[EAS] EAS Digest, Vol 15, Issue 16

[Harold Price]

Tom Taggart wrote:
>So the FCC's rationale for disapproving text-to-speech based on lack of
>consistency in this software seems misplaced.

Only the presence/absence of TTS was checked, not the quality.  The 
test report for each device also notes issues with the tests, for 
example, one vendor's test resulted in audio that exceeded the two 
minute limit for the TTS test.  Because TTS was only an "observe" 
criteria for the test, it was not within the scope of accreditation.

Tuning of the TTS is required, and it isn't easy.  Speed, pauses 
between words/sentences, expansion of abbreviations, local place name 
pronunciations, all lead to differences from one vendor to 
another.  Even simple things - in NOAA text, MPH is used.  The ENDEC 
says "miles per hour".  Perfect Paul says "miles an hour".  Just a 
small example, but little things add up.

Out of the box, the various TTS implementation are 
inconsistent.  Local areas will need to train each device type to say 
their place names.  Trust me, this is pretty tedious to do even once.

I think the FCC got it right, TTS is problematic in some cases, and 
parts of the listener community will be more sensitive that others to 
machine generated voice.  They want a "good" experience for 
listeners, and mandated against TTS.  The unexpected side effect will 
be, perhaps, far less acceptance by the origination community, and 
the CAP system will be used less often than was hoped.  As has been 
said, maybe the FCC (and listeners) can accept "good enough".

The FCC is currently hearing additional input on the matter.

Harold