The idea is that you enter a German word and the site returns pairs of sentences containing that word in German and English.
I looked for a recent problem, Evidenzfall. It wasn't there (unsurprisingly - the site is new).
I then looked for Evidenz. IMO Evidenz is not always evidence, but can have to do with things being plain or manifest. I translated Evidenzfall as case of a plainly void administrative act, and I came to this conclusion mainly with the help of German online definitions and in particular a textbook on German administrative law.
The hits I get with Linguee almost exclusively have Evidenz, bold, in the German half and evidence, bold, in the English half. 'More results...' produced more examples, without bold. I didn't see any examples from administrative law.
One result was given three stars out of three (most had two):
Die Evidenz, dass auch Pflanzen Stammzellen haben müssen
The evidence that plants also have stem cells
Note, incidentally, that the translation omitted 'müssen' - probably OK here. But one obviously can't rely on translations being good - when one searches the Web, it is more in a spirit of hope.
Certainly some of the translations were not good. 'We will keep your application in evidence'? It rather looks to me as if the program is designed to look for 'evidence' when I enter 'Evidenz'.
I'm oversimplifying it, because this problem arises only when I enter Evidenz. If I enter evidence, I get a very useful page giving a variety of equivalents, from which I can choose. And it's not just English words this happens with - the same happens with Aufgabe. So perhaps my Evidenz is too obscure.
I do think, however, that the examples given have been collected and sorted in advance. This gives the service of LEO or dict.cc. It interposes a brain of some sort between me and the corpus, without the useful forum discussions found on other sites.
What impressions do other people have?
(Via Übersetzer-Logbuch - originally via Twitter)
Defined tags for this entry: internet, translation
I love Linguee at first sight. Many times I just want to come up with a good formulation for something - by entering a set of two or three words I can find it now. Of course, as you say, you often have to take a look at the sources.
I still hope that they will increase the size of their database in the future and the quality of their translations. Then it will be way better than leo or dict.cc.
Normally I search for English equivalents, and by casting an eye over the Ghits, I see which are going to be more reliable. Or I select site:uk, to exclude a lot of non-native English sites.
Here, I would have to click through to see where the sites are and what they are. First I am shown the translations, some of them bad, and later I get down to the basics. This is a bit too longwinded for me.
One possible drawback lies in its very design: it is a database of TRANSLATED texts, so there is always a risk of "translatorese". It does little to help find the specialist monolingual texts by native speaker subject experts which are often so important.
So I use it as an extra tool (and have even defined a short cut to call it up with IntelliWebSearch/IWS straight out of DejaVu or other programs), but like all such tools I take it with a pinch of salt.
I think it will be better into German, because the Web is full of bad English, and I get shown the bad English - together with the good - as the first step.
The problem is that the inputs clearly aren't filtered for quality - hardly surprising, given the sheer volume of text that has been aligned and stored in the system. Ultimately, this is the same problem that faces stats-based MT systems: no matter how good the algorithms may be (and some of them are now clearly very good indeed, such as Language Weaver and Google), the whole exercise comes crashing down because of the GIGO syndrome.
Somewhere I have a list of sample financial reporting terms I used to apply to new dictionaries. If I can find it, I'll run the terms through Linguee and see what the results are. However, based on a rough sample of the translation sources I've seen so far, I have the feeling that the people at Linguee should go back to basics and apply more rigorous quality evaluation criteria to the parallel corpora they use to build their database - just because a company has a household name doesn't mean it produces usable translations!
When I try to find terms in Google, I get a page that gives me the URLs, and so I can see quickly which to home in on. Or, as I said, I choose site:uk.
Admittedly, if I search for Evidenz or Evidenzfall, I'm being unfair, in that I would never use such a difficult term on a site like that. But it does show up the problems, as does your Aufwandsrückstellungen.
leo.org does not have any results for "Aufwandsrückstellungen".
dict.cc does not have any results for "Aufwandsrückstellungen".
In linguee I see some good information and at least get an idea of how other people translated it (of course, I do not know which translations are completely right).
I will definitely give it a try next time I have to translate something.
The term "Aufwandsrückstellung" is a standard German GAAP term that appears in most HGB financial statements (although it's about to be abolished by the BilMoG), which is why I would expect to be translated correctly at least once in a sample of five translations.