Re: [tlug] Free program translates Euro languages to/from English

["Micheal E Cooper" <network-admin@example.com>]

> I'm still amazed at the ambiguity permitted in technical Japanese.  For

I second that. Considering that people's lives are on the line with
factory manuals, it amazes me that they can be so vague. When I asked
about it, the more seasoned, veteran engineers at the factory said that
they read over the manual at the beginning and then don't touch it again.
Basically, you read the manual as an intro to making your own set of
procedures, the content of which is based on the manual + your own
knowledge and reasoning. This is what I was told. When I asked why the
manuals are not clearer and more complete, they said that the user will
customize procedures and even the machine itself, so why bother?

> between plural and singular is a constant headache.  And as far as
> esthetics go, I don't think I'll ever understand how 300-character-long
> sentences containing 5 or 6 dependent clauses could be considered
> elegant.  And of  course the ambiquity problem is at least as bad with

Again true, but to be fair, my first bachelors was in English, and I can
tell you that literary theory, criticism, and scholarly work on the whole
is thick with needlessly long, complex sentences the purpose of which is
10% to convey meaning and 90% to raise the author in the intellectual
pecking order.

But about automated translation...

I really do doubt that we will be able to encode enough context and
knowledge to make machine translation viable for anything but the most
basic of communication. One of my last translation jobs for the prefecture
was the disclaimer to their automated translation page, which, though
better than most, was still pretty fortune-cookie funny. Even if the
designers got all the denotational meaning encoded (which is more or less
already done in digital dictionaries) and got all the context recognized
(which would be VERY hard), they would still have to contend with the
style and the connotations of the words being used (more or less
impossible, so they could leave it out) and they would have to keep
updating all of this very dense information to keep pace with the changes
in the language and target culture (maybe a wiki-like thing where users
correct the system and changes are fed back in?) or design the system to
read blogs and websites and then update itself (this is sci-fi territory).
Either way, I doubt it will happen in my lifetime.

My disclaimer: I am definitely a jack-of-all-trades, master of nothing,
and I am admittedly a chimney-sweep in the digital kingdom. I qualify all
of this with the recognition that I am not nearly as knowledgeable or as
l33t 5kild as the folks on this list.

That having been said, it seems to me that, rather than trying to make a
system that will translate for you (so that you can fire the translator)
or an online educ package that will teach for you (so that you can fire
some teachers), it seems more sensible to focus on creating tools that
will allow translators and teachers to do what they do better and more
efficiently.

And the earlier comments about translation packages choosing the most
common meaning, even within a subject or context setting, illustrates how
important this is. Instead of trying to replace the translator, the
program should be trying to assist the translator, to do the grudgework
for her.

In a last comment, a lot of MS adverts use the logic that, if you pay more
and use a Windows network, you won't have to pay for an expensive,
arrogant network administrator. You can do it yourself or have one of your
employees double as admin because Windows is so much easier to administer
than Unix. That is MS's TCO argument in a nutshell. In other words, pay us
more so that you can reduce staff and save that much money. The fully
automated translator market says the same to businesses: why pay a
translator per document when you can just buy software from us to do it
for a yearly flat fee?

The industrial revolution replaced craftsmen with human assembly lines,
and then automotive companies replaced the humans with machines. That has
worked for physical, repetitive tasks, but can AI replace human intellect
and judgement?