4 ms·

Statistical translation that Google uses works quite well to translate between most Indo-European languages. However, it has problems in other languages like Ch

Statistical translation that Google uses works quite well to translate between most Indo-European languages. However, it has problems in other languages like China and Finnish, but they are not necessarily insurmountable if you combain statistical approach with some domain logic and a lot of material. Developers of Google Translate described to me that Finnish is causing problems because the excessive inflection [1] in Finnish language needs a couple orders of magnitude more translated material to make statistical approach to work. At least that time it was not easy to obtain. Interestingly enough, the official EU documents and meeting translations are one of the best sources of 1-to-1 translations as they are translated to all languages in EU. [1] http://en.wikipedia.org/wiki/Inflection#Uralic_languages_.28agglutinative.29 http://en.wikipedia.org/wiki/Inflection#Uralic_languages_.28...

Mvandenbergh 15y ago
One of the reasons that the EU documents work so well is that they're painstakingly written simultaneously in all the official languages. There is not a master copy from which other translations are made. This also means that if, say, the German working group writes a clause that is only approximately translateable to French, both groups will have to re-write their versions so that there is as close to possible as an exact correspondence between them. Other possible corpuses are mostly literary, which are of course subject to significant rewriting for stylistic reasons.
xarien 15y ago
I agree to a point. For one, I would not group Finnish and Chinese together. Now, I can't speak much for Finnish, but I am positive a statistical approach to Chinese provides very poor results. However, with that said, I do agree that domain logic and a large relational data set is a good approach. Essentially relationships between words and symbols need to be established with false positive and false negative associations broken down and re-established in near real time to come close to a perfect solution. Basically, the problem is too expensive to solve.

Mvandenbergh 15y ago

One of the reasons that the EU documents work so well is that they're painstakingly written simultaneously in all the official languages. There is not a master copy from which other translations are made. This also means that if, say, the German working group writes a clause that is only approximately translateable to French, both groups will have to re-write their versions so that there is as close to possible as an exact correspondence between them. Other possible corpuses are mostly literary, which are of course subject to significant rewriting for stylistic reasons.

xarien 15y ago

I agree to a point. For one, I would not group Finnish and Chinese together. Now, I can't speak much for Finnish, but I am positive a statistical approach to Chinese provides very poor results. However, with that said, I do agree that domain logic and a large relational data set is a good approach. Essentially relationships between words and symbols need to be established with false positive and false negative associations broken down and re-established in near real time to come close to a perfect solution. Basically, the problem is too expensive to solve.