Saturday, November 22, 2008

Bringing The Global Web To India

A few weeks ago I blogged about search suggestions, a feature that made it easy to enter queries in Indian language scripts. Once you are able to do that and start searching, you'll quickly discover another barrier that non-English speaking users face in India - there just isn't much good content available online. No matter how good the search engine, it can't return relevant information if it doesn't exist! This situation is beginning to change in a few areas -news and entertainment for example - where lots of Indian language websites are being created. However we still have some way to go before the Indic web has enough high-quality content to satisfy all the information needs of users.

What can we do in the meantime? Well, Google has an interesting approach to this issue - automatic translation. If you do a search in Hindi and scroll down to the last search result on the first page, you'll see a link to a result that's been translated from English. For example, try querying for सरकारी नौकरी and scroll down to the bottom of the results page. You'll see a link to a translated query result. Clicking on the link takes you to a translated query results page. Here's how this works:

1) We will take your Hindi query - "सरकारी नौकरी" - and translate it into English - "government job".
2) We'll then run the English query and get back English results.
3) We'll translate those results back into Hindi for you.

All these translations are done automatically, using a machine translation engine developed at Google. This technology allows you to translate any text or web page instantly. Here's the Times of India homepage automatically translated into Hindi. Of course, because these are machine-generated translations they will never be as good as human translations (and they can even be quite funny) but the quality should be good enough for you to understand the sense of what you're looking for.

A neat and unique way of using technology to help bring information to users, even when it doesn't exist in their language.