0

I have used google translator. And it has horrible grammatical problems. I am looking for some library, or something that is more versatile then that. Does anyone know anything about it?

I once heard that pear has a translator package. Does anyone know how efficient it is?

JasonMArcher
  • 14,195
  • 22
  • 56
  • 52
Krishna Kant Sharma
  • 1,567
  • 1
  • 12
  • 18
  • What languages do you want to translate? – Bkkbrad Aug 21 '09 at 12:31
  • 5
    Any translation tool **will have** horrible grammatical problems. Specially if the two translating languages are extremely different (i.e.: English -> Japanese). But most of the time, it's acurate enough to understand the underlying message. – Esteban Küber Aug 21 '09 at 12:33
  • Yes voyager, you are right. I am just asking that what is the best method two translate text from one language to another. (Not manually :) – Krishna Kant Sharma Aug 21 '09 at 12:42
  • Google Translate is among the best automatic translation systems existing today. If it's not good enough for you I don't believe you'll find a system which is. Machine translation has been one of the toughest problems tackled with digital computers since they were invented. – hippietrail Aug 04 '11 at 07:42
  • I'd stick with Google definitely. Though it is not always as good still it produces the **best** results. And Google is working on the semantics recognization to serve even better results each day... – shadyyx Mar 07 '13 at 11:32
  • While looking into Machine Translation for a project of mine, I came across Moses. I have never used it, but it sounds pretty cool, and is open source. feed it a library of texts translated manually between the two target languages, and it constructs a translation database on its own. It supports strict word-for-word translation and intelligent grammatical tree-based translation too. It's a C++ application, so if you want to use it on the web, you may need to write a wrapper for it in PHP or Node or your preferred server-side language with commandline capabilities. http://www.statmt.org/moses/ – Adrian May 08 '13 at 16:24
  • Yandex's translation service is an alternative for translating websites. After a moderate search, I didn't find very many services that do it, anymore. If you're not looking to translate websites specifically, there are a lot more alternatives. – Brōtsyorfuzthrāx Oct 20 '22 at 22:02

3 Answers3

3

It's unlikely you'll find a much better solution. There are two primary methods to machine translation - rule based and statistical. Rule based translation has the problem that a great deal of text involves idiomatic expressions, grammar, or vocabulary that are outside the set of rules. Statistical methods can sometimes provide better solutions if there is a large dataset (and large data sets are Google's business), but also has trouble with idiomatic expressions and differences in word order between languages. For more information, check the Wikipedia Article on Machine Translation, and more specifically the articles on Inter-Lingual and Statistical Machine Translation methods.

T.R.
  • 7,442
  • 6
  • 30
  • 34
1

Try babelfish: http://babelfish.yahoo.com/

1

You will not find any package that can automatically translate a text gramatically correct. Computers simply can't understand text well enough to translate it.

The main problem with automatic translation is understanding the text that is translated, the actual translation part is trivial in comparison and could be done by a computer.

Take for example a simple sentence like "When the going gets tough, the tough gets going.". It has ambigous words, implicit references and cultural undertones, which is very hard to write an algorithm for deciphering.

I once read about a project of developing an independent, unambigous intermediate translation language, which sounded promising. You could translate your text (partly manually) to this intermediate language, which then could be automatically translated into any language. I haven't heard anything more about it, though...

Guffa
  • 687,336
  • 108
  • 737
  • 1,005
  • 1
    This seems intuitive but isn't actually true. The best automatic translators today like Google Translate don't even try to understand the text at all, they use statistics instead. In fact Fred Jelinek (in)famously said "Every time I fire a linguist my system improves" about the machine translation he was working on. – hippietrail Aug 04 '11 at 08:09
  • @hippietrail: That is only true to a certain level, i.e. statistics can only get translations from *horrible* to *bad*, to get a good translation you need to understand the text that you translate. – Guffa Aug 04 '11 at 08:09
  • 1
    I tend to agree but the consensus seems not to agree \-: Though everybody agrees computers will never be able to translate as well as people. "Context" would be the bigger obstacle though. Lack of general knowledge used to be cited as well but I don't see this so much now that large knowledge bases exist and machine translation still sucks. – hippietrail Aug 04 '11 at 08:14