HN Explorer

Google's wrong translation

1 points by cvrajeesh ↗ HN
When translated "India is my country" from English to Urdu http://translate.google.com/#es/ur/India%20is%20my%20country

and reversed the same http://translate.google.com/#ur/en/%D9%BE%D8%A7%DA%A9%D8%B3%D8%AA%D8%A7%D9%86%20%D9%85%DB%8C%D8%B1%D8%A7%20%D9%85%D9%84%DA%A9%20%DB%81%DB%92

resulted in "Pakistan in my country".

Can some one explain why is that?

2 comments

[ 2.8 ms ] story [ 18.9 ms ] thread
balnaphone
You don't need the reverse translation, you can see it's Pakistan (پاکستان says "pakistan") directly in the Urdu. If you want to see how to translate India into Urdu correctly, choose Arabic and take the first word (الهند says "al-hind" or "the india"). The core google translation matrix is derived from the six official languages of the UN (English, French, Spanish, Mandarin, Arabic, Russian), since around 20 million parallel translations are available to the public for these six.

For languages like Urdu, they try to derive translations from other sources, but key among them is the corrections that users of the translation service make themselves. So, if say a student translates an English language article into Urdu, but chooses arbitrarily to edit the translation to replace India with Pakistan, then the engine will eventually adopt this translation, if it's seen enough times.

Note also that the Urdu translation says "ALPHA" in red letters above it? That's because they know it doesn't work, there likely isn't enough sample data to make a strong translation matrix.

Lastly, you might want to replace your huge URLs in your post with tiny ones (e.g. using tinyurl.com) so the page formatting isn't screwed up.