A friend of mine pointed out the Twitter account of Mirko Gosch, who is identified in his profile as "Translator, Level 2."
There have been all kinds of rumors about Twitter some day rolling out translation. When I heard these rumors in the past, I was skeptical, because I knew true, idiomatically accurate, automated translation was basically impossible.
Biz Stone concluded a post on the Twitter blog last month by writing, "We're very excited that more people will be able to use Twitter in their native language!" I found this confusing (not to nit pick) because being able to use a site in your own language requires localization, the translation of all of the site's menus, screens, prompts, etc. The ability to pass Japanese between two Japanese speaking users is a technical capability, but not a terribly difficult one (and one Twitter already has.) The ability of a French-speaking person and an English-speaking person to communicate with each other, without knowing each other's language, is a much more difficult one.
I don't completely understand Twitter's approach, but it appears as if it involves a combination of human translators approved by Twitter, and the capture of translations of commonly occurring phrases which are used to build an online dictionary.
According to its Twitter Translate page, the company is working on Italian, French, and German. The service is "currently in an invite-only mode while we improve our tools, but may allow any Twitterer to help translate at some time in the future. We've completed English and Japanese."
Here's how Twitter describes Twitter Translate functionality:
"Twitterers enabled with our translation tools will see a special set of options on their screen. By clicking the "Translate this page" option on any Twitter.com page, the Translate Box will appear on the screen, pointing to a certain word or phrase. You can type in your own translation or vote on another translation from within the box. The box then jumps to the next phrase until you've translated everything on the page. After that, you move to another page.
We have another view which allows you to translate everything on the site in a single list. In either mode, you can translate as much or as little as you'd like, and can stop when you'd like without losing your translations. Once we've collected a good number of translations for every phrase across the site, we will review them and pick the most accurate one. It's a fun and easy way for you to help make Twitter accessible to more people around the world."
If I understand Twitter's approach correctly, they are on the right track. I used to work in the Machine Translation (MT) industry, and accurate, fully computerized translation isn't difficult, it is, so far, impossible. This is because different languages conjugate verbs differently, handle plurals differently and place adjectives and punctuation in different places, among other things, so simply translating words doesn't work.
Today, at the low end of automated translation solutions, are free, online applications like freetranslation.com and Google's translation tools. These will supply what is called a "gist" translation, as in "you get the gist of it." I wouldn't recommend using either of these to translate instructions for a surgical procedure, though.
At the other end of the spectrum are highly sophisticated Machine Translation and Translation Memory products. Depending on implementation, both of these almost always involve native-speaking human inputting of words and phrases to produce acceptable translations.
Even if Twitter masters automated, on-the-fly translations of tweets, there will be significant IT challenges. Today Twitter experiences lag and "Fail Whales" publishing 140-character updates without translation. Providing smooth, quick translation to millions of users on top of Twitter's current workload will be a formidable task.
Tags: Twitter Translate, Machine Translation, Localization, Localisation
Link to original post