Find empty targets: was Client not happy I used Google Translate
- View SourcePut this in the SQL box and find all segments with an empty target.
ID IN (SELECT ID FROM Sentences GROUP BY ID HAVING Count(ID) = 1)
Then you can select them all and remove them at one go.
On 28/04/2012 6:44 AM, gennady.lapardin wrote:
> Hi Herbert,
> I'm new here, and my starting post few days ago was if somebody was able to suggest how to clean TM from empty (not translated) segments/entries/lines, call them at your descretion. Got no reply at all.
> Back to this topic, this is how I work from the early '90s, when I used (and still using) Promt. First, I run the text through MT, align, delete empty (not translated) entries, get raw (I call them shadow) TM. Then, go ahead as usual. In the latest DVX2 I can use benefits of using 1. shadow TM (from GT), 2. mandatory TM, if any, 3. AA. That is why I'll never give up my DVX2.
> Could somebody share how to clean TM from not translated entries, please?
> Re: Client not happy I used Google Translate
> Posted by: "Herbert Eppel" herb@... hetranslations
> Fri Apr 27, 2012 6:21 am (PDT)
> It would be nice to see the GT and AA results side by side so
> that one can select the best bits from both.
> I seem to remember that various workarounds were discussed in the
> Can someone remind me what the preferred procedure is for
> achieving this?
> Using: DVX2 Build 533 WG, working in the grid
> Office 2010 (NL)
> Windows 7 Pro 64 bit (ENG)
> J> John Lejderman
- View SourceAt 20:11 30/04/2012, Michelle wrote:
>Are you trying to say that it is better not to trust GT? :-D
>Those who use GT should remember also that when GT sends a
>"translation" it asks whether the person receiving it wants to
>"improve" it. I suspect many do. Although not all so-called
>"improvements" really are that.
GT's very untrustworthyness is what makes it so useful for
professional translators. In the hands of anyone without an excellent
knowledge of their subject and good language skills, GT can be a very
dangerous tool, whereas we are in the position to reliably take
advantage of its power to speed things up and help with research.
And while some of GTs unreliability (to make an understatement) might
be due to the "improvements" you mention, most of it is due to the
fact that while the technology is amazing...it's still just not that
good. It's still no replacement for an human being with expertise,
despite its advantage in speed and memory.
Here is an example of my latest GT result:
"Another aspect that emphasizes the separation of the lateral bodies
than the central band is the bricks with which it is treated the top
of the first, while the hoof, which gets up to two meters from ground
level, runs smoothly throughout the building."
Sometimes I get near perfect results (kind of scary), and sometimes I
get a gem in amongst the gibberish (happily added to my
TDBs--confirming a term based on subject knowledge is generally much
faster than finding it yourself). And sometimes, as in the example
above, there are no gems, most of the terms are wrong, it makes
little sense, and yet GT still speeds things up by giving me a
structure to start dictating from, based on my understanding of the
original sentence and on the terms I know or have stored in my databases.
Your milage may vary. In any case, in case it is not clear to
everyone on this list, GT only asks people for "improvements" when
providing translations from their website. DVX2 DOES NOT sends our
translations to Google.
Happy May Day,