[Foundation-l] Push translation

[Andreas Kolbe]

--- On Sat, 31/7/10, Nikola Smolenski <smolensk at eunet.rs> wrote:
> Interestingly, I have had a completely opposite
> experiences. When reading a 
> Google translation, it is easy for me to decipher what does
> it mean even if 
> it is not gramatically correct. When translating, I often
> hang on deciding 
> what sentence structure to use, or on remembering how a
> specific words 
> translates. GTT solves both problems. My estimate is that I
> retain half and 
> rewrite half of every sentence it produces.


I'm afraid if that is how you proceed, you are already up the creek without a paddle. You say, "When reading a Google translation, it is easy for me to decipher what does it mean even if it is not gramatically correct." 

If you are translating, you should not be able to decipher what the Google output means, you should be able to decipher what the *original* says, *from looking at the original*. 

Because the Google Translator Toolkit, at times, translates "there is one such system" as "there is no such system", or it translates "A is governed by B" as "A governs B". Don't ask me why, it does. Even in mainstream language pairs like English and German. I shudder to think what it does in Hindi and Tamil. 

So when you are working on a text about maths, or physics, that is supposed to go into an encyclopedia, deducing the meaning of the original from the Google translation is really quite fatal.

And to someone who is fully fluent in the source language, and wants to compose a text in the target language, Google Translator Toolkit is, at present, worthless. Word-processing the Google output to arrive at a readable, written text creates more work than it saves. 

Remember, we are talking translation here: that means composing a well-written, correctly formatted text for others to read. We are not talking about "figuring out what it probably means."

If Google want to build up their translation memory, I suggest they pay publishers for permission to analyse existing, published translations, and read those into their memory. This will give them a database of translations that the market judged good enough to publish, written by people who (presumably) understood the subject matter they were working in. 

This seems a much better idea than to pay for and collect memories from haphazard Wikipedia translations done by amateurs which, judging by the feedback from the relevant Wikipedia communities, are garbage. Why feed that garbage into the system? 

There should be alarm bells ringing at Google here. 

A. 



> Andreas Kolbe написа:
> > Having tried it tonight, I don't find the Google
> translator toolkit all
> > that useful, at least not at this present level of
> development. To sum up:
> >
> > First you read their translation.
> >
> > Then you scratch your head: What the deuce is that
> supposed to mean ...?
> >
> > Then you check the original language version.
> >
> > Then you compare the two.
> >
> > Then you start wondering: How did *this* turn into
> *that*?
> >
> > Then you shake your head.
> >
> > (Note: everything up to this point is unproductive
> time.)
> >
> > Then you look at the original again and try to
> translate it.
> >
> > As you do, you invariably end up leaving the Google
> shite where it is and
> > writing your own text.
> >
> > In the end, you delete the Google shite, and then, as
> you do so, you kick
> > yourself because there were two words in there that
> you needn't have typed
> > yourself.
>