It's actually quite weird because Google only seems to need
about four Swedish comments to figure out the language but with
Finnish, it needs a lot more... This one works as it has
59 comments.
@Nikke: I think you're
right, when Google does figure out that it's Finnish the
translation is leaving a lot of words untranslated. I assume that
means that they're basing their language detection scheme not on
grammar or patterns but on the actual words it has in their
dictionaries.
Still trying to find threads written in other languages than
English. So far I've found Finnish, Swedish, Chinese, and Danish
that Google was able to translate.
Compare the Danish post with the long one in Finnsish and look
att the nouns. I'm quite certain that we are on to something here
about Google's language capabilities. They do use stemming as well,
but don't seem to be able to strip more than a few chars from the
end or beginning of the word.
@Nikke: Seems like there's
more information in today's DN
(in Swedish). It says it is non-rule/grammar based and depends on a
large dataset of words and their respective translation.
21 comments so far
great news... but aint it a bit heavier?
1 week ago by BUGabundo.
Chinese it can figure out right away but unfortunately it's not the same with Finnish.
1 week ago by lemonad.
@BUGabundo: I meant that I've just added a link to translate jaiku pages :)
1 week ago by lemonad.
i know
1 week ago by BUGabundo.
Yay for our zuper-zecreto FI language! ;)
1 week ago by jkniiv.
@jkniiv: You just wait, Google has got @jyri now ;)
1 week ago by lemonad.
It's actually quite weird because Google only seems to need about four Swedish comments to figure out the language but with Finnish, it needs a lot more... This one works as it has 59 comments.
1 week ago by lemonad.
google needs more info to make sure it isn't japanese?
1 week ago by cybette.
I think it's because of all those damn Finnish cases to the noun. It must be really difficult to find enough unstemmed words in just a few comments.
1 week ago by Nikke.
@cybette: Ah, so Google and I are not so different after all, we both understand Finnish very poorly :)
1 week ago by lemonad.
@Nikke: I think you're right, when Google does figure out that it's Finnish the translation is leaving a lot of words untranslated. I assume that means that they're basing their language detection scheme not on grammar or patterns but on the actual words it has in their dictionaries.
1 week ago by lemonad.
Still trying to find threads written in other languages than English. So far I've found Finnish, Swedish, Chinese, and Danish that Google was able to translate.
1 week ago by lemonad.
Hot damn, Google only needed one comment to figure out that this was Danish.
1 week ago by lemonad.
Russian (?) was way easy.
1 week ago by lemonad.
@lemonad: "We are not yet able to translate from Malay into English".
1 week ago by lemonad.
Compare the Danish post with the long one in Finnsish and look att the nouns. I'm quite certain that we are on to something here about Google's language capabilities. They do use stemming as well, but don't seem to be able to strip more than a few chars from the end or beginning of the word.
1 week ago by Nikke.
@lemonad hello tell me is linux OS good? and which version?
1 week ago by Aaqil.
@Nikke: Seems like there's more information in today's DN (in Swedish). It says it is non-rule/grammar based and depends on a large dataset of words and their respective translation.
6 days, 20 hours ago by lemonad.
Google's translation of the article
6 days, 20 hours ago by lemonad.
The article was really from Ny Teknik.
6 days, 20 hours ago by moonhouse.
@moonhouse: Ah, of course. They've initiated a collaboration.
6 days, 20 hours ago by lemonad.