Machine Translation

Debbie Garside debbie at
Mon Sep 14 19:55:51 CEST 2009

Hi All

I am just trying to catch up on all this following a family bereavement...

It may be a few days before I have an opportunity to read through all the
traffic.  Rest assured, I will be feeding this back to colleagues.  However,
in the meantime, I am very interested in comments from individuals who have
knowledge of or build/supply machine translation applications.  Would the
use case defined be better served by some sort of partial generic MT
generated tagging e.g. by using the <Author/> tag for the MT (perhaps
<Author>MT_ApplicationName<Author/>) with the MT aspect being a generic part
of the tag coupled with such tags as <SourceLang/> etc. ? This would almost
certainly solve the problem of corpus building from web text.

With regard to a variation or extension... As Editor of ISO 639-6, I would
consider a tag for WelshWrittenLatnMT as I think it could be classed as a
Linguistic Register.  Although, I would generally associate register with
spoken discourse I think there could be sufficient ambiguity within the term
(purportedly invented in 1956) to allow for a written, non-human aspect.
Nevertheless, that doesn't mean to say that this would be the best way to
progress matters.

Try not to clobber me for my thoughts as they are a little awry at the

Best regards


> -----Original Message-----
> From: ietf-languages-bounces at
> [mailto:ietf-languages-bounces at] On Behalf Of Doug Ewell
> Sent: 12 September 2009 22:39
> To: ietf-languages at
> Subject: Re: Machine Translation
> CE Whitehead <cewcathar at hotmail dot com> wrote:
> > Personally, I'd like to be told when I retrieved a document
> whether it
> > was a human translation or machine translation or original, and if
> > either of the former, where the original was lodged--if it
> is lodged
> > online, and particularly so before I put the document through an
> > online translator.  (For example, if I retrieved a document
> translated
> > from French to English with a machine, and I were a native
> speaker of
> > French, and I put it back through a machine to get a French
> version,
> > that would be ----- dumb, I guess.)
> >
> > The w3c's policy for translations--requiring at the top or
> bottom of a
> > document (in a header or footer that 'wraps' the document)
> a statement
> > indicating that a document is a translation, the the
> original document
> > is the normative version of the document, and with the URL of the
> > original document--does result in this information's appearing in
> > search results (this is helpful!).  I've not been able to
> locate much
> > about translation policy otherwise online but note that joomla
> > translations does have a translation policy:
> All the complexity described here does tend to imply that BCP
> 47 language tags might not be an appropriate place for
> declaring "translated mechanically" after all.
> I hope some of the feedback on this list is being relayed
> back to Debbie's colleagues who requested this in the first
> place.  Maybe they would be satisfied with a higher-level
> protocol such as this.
> --
> Doug Ewell  |  Thornton, Colorado, USA  |
> RFC 5645, 4645, UTN #14  |
> ietf-languages @
> _______________________________________________
> Ietf-languages mailing list
> Ietf-languages at

More information about the Ietf-languages mailing list