How Accurate is Google Translate?

December 10, 2008 by
Filed under: Translate, Translator 

Many users ask this question after using Google Translate for the first time. They want to make sure it’s a reliable translation tool before they can brag that they can speak in another language (well, they can until someone proves otherwise).

Since only native speakers and professional translators can tell, their question most of the time is left unanswered.

Actually, there are instances that Google’s language translator can provide a translation as if a human translates it. There are also instances that it produces poor translation. Because of this translation flaw, users should only use this tool then to get an idea of what a foreign text could possibly mean in 41 languages (well, 40 minus the language to be translated).

How Is Google Translate Different From Other Online Translation Tools?

Google Translate, in comparison to other language translation tool, is using a “statistical translation system for the language pairs” instead of the rule-based approach that “requires a lot of work to define grammar and vocabularies.”

What the company actually means with their technology is that they “feed the computer billions of words of text, both monolingual text in the target language, and aligned text consisting of examples of human translations between the languages. We then apply statistical learning techniques to build a translation model.” Google first got their linguistic data from United Nations’ documents that are normally available in six languages, and acquire more from other resources.

(Their explanation still seemed vague to me, but using my imagination seemed to help me understand what Google really means. It’s like feeding their computers with text from a Bible that was written in English on the first column and in a different language on the right column.)

Can it Be Improved?

Despite the flaw of Google translator, the translations can be improved. The company is constantly working on its perfection, and even users can contribute if they see that a word or a text has been poorly translated by the computer. This can be done by native speakers who would like to test the accuracy of the tool, or by even those intermediate speaker. Both of them can easily identify a poorly translated part of the text since they have a good background of grammar rules, and idiomatic expressions.

So to contribute, all they have to do is expand the Suggest a Better Translation link. It will display a box where they can type their translation. After typing, they click the contribute button.

It may take time though for Google to update their database since they might still review what have been submitted, but at least everyone can help on improving the quality of thier language translator tool by suggesting a better translation.

Comments

5 Comments on How Accurate is Google Translate?

  1. CJ on Fri, 12th Dec 2008 8:26 am
  2. I am a native French-English speaker and did machine translation for my thesis.

    I tested it:

    http://scienceforseo.blogspot.com/2008/11/google-translate-tried-and-tested.html

    You write a nice and complete summary – nice :)

  3. acky on Fri, 21st Aug 2009 5:43 pm
  4. Sa pamamagitan ng aking course at sa mga activities na aming ginagawa gaya ng mga router malaki ang maitutulong nito para sa aking kinabukasan at para makamit ang aking pinapangarap na makapagtapos ng pag aaral.

  5. levent on Sun, 9th May 2010 6:50 am
  6. english turkish translation is really really bad. a bunch of meaningless words.
    it would be much better if google didnt support turkish.

  7. Janet on Mon, 20th Dec 2010 6:18 am
  8. We offer professional language translation service and found Google less reliable, for the reasons mentioned by Reah above.

    Though it is a great tool to understand the subject of a discussion and actually take help of it in participating in discussions.

    I have done this with our clients, and proves helpful…

  9. Anonymous on Mon, 3rd Jan 2011 3:52 pm
  10. seems pretty good for me, in frrench english at least.