Hacker News Reader

Comment by TimWolla | Hacker News Reader

> Secondly, they didn't even capitalize Seafile's company name, but they did capitalize the competition's name (Dropbox).

That was the fault of Google Translate, see the original: http://www.golem.de/news/sperrung-paypal-entschuldigt-sich-b...

View on HN

Well, that's weird. Why would Google Translate make it a non-capital letter in the translation when it used a capital one in the original document? I don't think that should be happening in any scenario? Is this Google's "unsupervised" AI running amok?

> Well, that's weird. Why would Google Translate make it a non-capital letter in the translation when it used a capital one in the original document?

Because the German Language loves to put Capital Letters anywhere it feels like, which does not translate 1:1 to what would be appropriate in English Use.

Google Translate seems to have a poor heuristic to detect names as such.

>anywhere it feels like

It's not as arbitrary as you make it out to be: Names, nouns and the beginnings of sentences are capitalised.

German capitalizes a lot of words that aren't capitalized in English. English capitalizes "American" but in Spanish it's americano. So no, you can't always preserve that aspect of translated text.

"All nouns are capitalized in German". So it probably needs to do this somehow to get decent translations. German also throws together multiple words, right? So maybe it detected "seafile" as a compound noun of "sea file".

Compound words are common too in English. They're really just treated as new words, there isn't a real reason to syntactically split them. German does have some more comical examples though. My favorite was an archaic german book we had in highschool that for "typing" used the word, "schreibmachineschreiben" which is literally "write-machine-writing." It looks like the modern word is "Tippen."

All nouns in German are capitalized, not just proper nouns as in English.

For example:

"Das Mädchen trägt ein blau Hemd" becomes "The girl is wearing a blue shirt."

Mädchen -> girl, Hemd -> shirt both get downcased.

(Note: my translation might be a little off. I haven't used german in a good 15 years)

> (Note: my translation might be a little off. I haven't used german in a good 15 years)

Wow, very good! It was nearly perfect, it should have been "Das Mädchen trägt ein blaues Hemd".

Yeah. Modifying adjectives to match the case / gender always got me.

Reply on news.ycombinator.com