Re: Translations

> From: koen@win.tue.nl (Koen Holtman)
> Subject: Re: Translations
> 
> M.T. Carrasco Benitez:
> >
> >Should translations request be included in the same model as linguistic 
> >versions.  Translations request (to humans or machine) could be 
> >considered as "virtual linguistics versions"; i.e., the do not exist at 
> >this moment, but they can be available.
> 
> Well, in the models used by HTTP and transparent content negotiation, the
> server would list every linguistic version it currently has available,
> whether virtual or non-virtual.  HTTP is not concerned with how the server
> produces its responses internally, so there is no distinction between a
> version that is stored in a file and a version that is produced on request
> by a translation engine.  Both produce the same kind of bits on the wire.
> 
> I don't know if you would want to take the HTTP model though; it depends on
> what you want to use the model for.  The possibility to make a request to a
> human translator does not really fit into the HTTP model.
> 
> >Tomas
> 
> Koen.
> 

One of the recent URI discussions dealt with the URL addressing an object or
a service or a transaction of some kind. This was significant to some people
who thought that GET on a mailto URL seemed inappropriate.

HTTP might be an insufficient model for real commercial translation services.
An interesting aspect about mail is that it might be targeted at an individual
or broadcast to a number of users. In the commerce arena the broadcast nature 
of bid/asking prices makes a lot more interesting interactions possible.
e.g. Suppose the nature of the original request was I have $1000 to spend
on translating a ten page document within the next two weeks. The responses
to this negotiated transaction could involve alternate costs, schedules, 
unspecified results (e.g. I'll give you Japanese and French in the same time
period for $1001 or I can give you 12 languages in draft machine translation
in 1hr for $100 and let you pick the final translation languages for $20
a page).

This is way off the original note of how to label existing content, but the 
static/dynamic nature of content on the web begs the question of whether or
not future content possibly generated over a longer lasting transaction
is not part of the overall model.

Received on Tuesday, 14 January 1997 12:12:36 UTC