Feature request to help find junk translations

I've come across a lot of translations in the various languages I've worked on, where the translator, either because they are learning, testing, or getting it plain wrong, has just submitted a "translation" of the English phrase where they haven't actually translated anything, i.e. it is still in English.

A helpful feature, especially for a manager, would be to be able to search for any "translations" that exactly match their original phrase. I know there are instances where this will be correct, i.e. if the phrase is very short and the two languages share the same word, e.g. "Blog" or "Wiki" in the footer, etc.

But there are still some phrases where people have submitted "{actors} replied to the project discussion {topic}" as the German "translation". These need to be weeded out, and at the moment, I only come across them by accident.

Basically, my request is to filter a list of translations that exactly match their original phrase, so that these can be reviewed and deleted/ voted down (or kept if correct).

Examples?

In Norwegian we decide to NOT translate the translation tools phrases, - i.e - we just use the English text as is.

I'm all for seeing corrected translations where "junk translations" (I've probably posted quite a few) have been placed. I usually try to keep the original text with the translation for just that purpose.

Bjørn, I don't have any examples, because I've fixed the ones I've found and the ones I can't find, I need this requested feature for. What I mean is that someone has clicked Translate on the phrase, for example, "{actor} replied to the discussion" and has submitted the translation with the exact text "{actor} replied to the discussion", despite the fact that the language is German.

I'd argue that your decision (not that it's any of my business) in Norwegian not to translate the Tr8n Tools interface is not a good idea. It is more helpful to Norwegian translators to use a Norwegian interface, in my opinion.

Ben, I don't know what you mean by "I usually try to keep the original text with the translation for just that purpose."

It's difficult to weed out what is a "junk" translation by an automatic search (how can the computer know whether "kjsdvkjwoev kwjvnv sldv9eufw dkflj" is more/ less correct than "{actor} hat auf die Diskussion geantwortet"?)

Something that could easily be automated is my original point: finding submitted phrases which have not been translated (i.e. submitted without changing the text).

Another point would be a filter showing submitted translations where the tokens used are either wrongly written (e.g. "{actor} commented" = "{atcor} hat kommentiert") or left out entirely (e.g. "{actor} commented" = "Aktor hat kommentiert"). There should be a filter showing that the number and type of tokens in the translation matches the number and type of tokens in the original phrase.

Private User & Bjørn P. Brox
If this filter/tool/feature will be available - it will be nice.
Even there will be "noise" like Bjørn's non-translated Tr8n pages phrases or for some cases sentences where the token is missing after translation (in Estonian we are not usually using for one item count/num and list rules translation tokens which will give 1 page, 1 profile, 1 person but word One page, Only one profile etc instead). Anyhow the results list will be much shorter (tan all 9000+ Geni phrases) and it will be then easy to manage.

I think Lauri Kreen, Ͼ is exactly right. There may be some (many?) translations which are "noise" i.e. they are picked up by such a filter but not necessarily wrong. But the time saved in finding those translations which *are* wrong will be very useful.

As an illustration of my second point about the wrong number/ type of tokens in the translation, here's an example from Chinese (Simplified):
http://www.geni.com/tr8n/phrases/view?translation_key_id=192722&amp... Aside from the fact that the translation is actually not correct (i.e. it does not convey the correct meaning), the token used is visibly wrong.

A filter which highlighted translations which had tokens that didn't match its source phrase would be useful for finding such errors as these. Again, the computer won't be able to tell that "回复" doesn't mean "Your reply to" (it's just "Reply"), but it will be able to highlight that "{target}" is not the same as "{topic}".

Kit, - i f users are not able to understand the English menus in the translation tools they should stay away from translation job ;-)

Bjørn, you do have a point! :)

Bjørn P. Brox Good point, but then your language translation will be never 100% complete ;-)
Of course you can translate these phrases last...

@Lauri. That' is correct but our thinking has been that it is better to use our limited resources on translating the part of the UI the average end user faces than striving for 100%.
BTW; I have sent a feature request to Geni on doing the statistics based on translated+locked ("locked" means that the phrase has been handled (a conscious decision to lock it has been made by someone; hence the need for the comment to document the who/why part of the equation).

I have made a document keeping track of this where I do my own statistics (as my feature request has been given the standard Geni response (=ignored)).

To break down the remaining work I have categorized it as :
A)- Translated
B)- Non-translated minus "locked as non-translated" (=handled)
C)- Phrases where we wait for a more intelligent handling from Geni (e.g. gifts)

The stuff to work on = All phrases minus A minus B minus C

https://spreadsheets.google.com/ccc?key=0All5GJ8d-2LLdDBiQVVKaXdyaV...

(Note : Norwegian language content)

I sometimes switch language to see how others have solved some specific phrases using rules, and in such case is it nice to have a common interface.

Language selection phrases should not be translate either if the original phrase is not English. If you for example accidentally change your language to Russian the user will be stuck if someone have translated the menu item to switch back to his own language to Russian. Keeping the original text, for example the Hebrew link at the bottom of the page is important.

When it comes to 100% translation you simply solve it by adding the translations for these phrases as untranslated text, .- i.e. just copy the original text and store it as the translated text.