[EAS] Multilingual support for CAP messages

[Ed Czarnecki]

>> But... what of the situation that some of you have described where you
have dozens... or more... languages in your community?  
A:  There are very few CAP originator products that currently enable
authoring of a multilingual CAP message.  The DASEOC is one that does (as
deployed in Minnesota and other jurisdictions, these is a useful case
studies).  

>> A downstream station can't "choose" a "primary" language... if it isn't
in the CAP message to begin with.
A:  Actually, we've demonstrated that downstream stations can, in fact,
choose languages that are not in the CAP message ... or the broadcast EAS
message, if that is what is received.  

TPT in Minnesota has effectively demonstrated transmission of alert messages
in multiple languages, when the source alert was English-only CAP, or
conventional English-only broadcast EAS.  In these cases, we are using rich
CAP source information in multiple languages, TTS in multiple languages, and
standard EAS sentences where the rich CAP text is not available.  The
Minnesota EOC (HSEM) has resources at its disposal for human translations.
However, there can be significant lag in getting these messages translated,
particularly at odd hours.  Which is one of the reasons we developed the
approach we did, that combines automated capabilities at both the originator
and the EAS participant.  *** On a voluntary basis.  By willing
broadcasters.  Who know they are addressing the language populations in
their audience. ***

Regarding automated translation of "free form" CAP text (the info in the
<instruction> and <description> parameters), I would argue (strongly) that
the state of the industry is nowhere near reliable enough to support
automated free form translation of critical life-safety information.  If
you've ever used Google translate, you can see the translation abberrations
even in some simple statements - never mind trying to translate abstract
terms like "shelter in place".  I would not recommend relying on automated
translation for either alert originators, or EAS participants.  Too much
room for error.

This is what the FCC calls "machine translation" in their recent NPRM.  We
do not support the idea of fully automated translation yet, due to the still
developing state of the technology.  We do, however, propose an alternate
approach that has already been deployed in Minnesota.

So, as for what can be realistically done today with existing technology,
I'd again cite the Minnesota example, which we briefed in some detail at
last years NAB conference.

Ed

-----Original Message-----
From: eas-bounces at radiolists.net [mailto:eas-bounces at radiolists.net] On
Behalf Of Dave Turnmire

Many thanks to Harold and Ed for their providing input regarding their
equipment and other insight.  And to the others who have commented.  One
never knows whether a question here will elicit a yawn or a lively response!

Unless I have missed something (quite possible!), this discussion has