Facebook vs. Google; Translation Methods
A recent news article detailed Facebook and Googles “competing” translation methods. Facebook uses a crowd-sourcing/wiki methodology, which allows armies of rabid fans around the world to translate the essential parts of the Facebook interface into the languages of the world. Google, on the other hand, trusts its core ability to program a reliable, automated software interface to handle all translations (see my earlier post here).
These aren’t competing systems so much as completely different environments. Facebook has the problem of enormously expanding popularity. They had to quickly translate the interface widely in order to forestall any international competitors. And, with Facebook’s popularity, it could quickly crowd-sourcing using English fluent individuals in each of the languages. Facebook is now in 65 languages, with 30 new ones coming online soon. Wow!
This is relatively similar to the system used by Second Life to handle their translations. Faced with mounting costs to get their text translated well and into non-English languages, Second Life abandoned professional translation in order to embrace crowd-sourcing for most of it’s translations. They’ve been quite successful with this. Eve, a massively multiplayer videogame based in Iceland, has also embraced fans to get localizations completed.
Google has another path, and another goal. Google, for all of it’s popularity as an interface, does not generally have rabid fans ready to turn every link into 100+ languages for it’s own websites (they did use crowd-sourcing for the Language Tools interface, however). And since the basis of Google’s search business is the linkage between OTHER peoples’ sites, language becomes a significant barrier. Google is most powerful when the maximum number of people are speaking clearly to one another (network effects). It’s only natural that Google try to use it’s core competency (cloud programming) to try to maximize its impact.
I recently used Google’s Language Tools for the first time in several months. They’ve added a number of languages, and the translations are definitely clearer than they once were. Currently, they claim over 2500 language pairs between 51 languages (though I imagine most of these pairs are functionally Language One -> English -> Language Two).
Personally, I feel that the crowd-sourcing is a marvelous expedient to those platforms popular enough to support the solution and willing to publish first in English (or another language) and then have the others languages lag by weeks or months. Wikis have proven themselves to be excellent at self-correcting errors and abuses, so long as a minimum of volunteers participate.
The future is clearly with Google and automated translation. The only question is “when”? And I’ve been asking that question since I first met the CEO of Dragon Systems, the first developer of linguistic software, in San Francisco in 1991. We’re closer now than we were then … but still far away. It may be a case of moving goalposts, rather like the definition of Artificial Intelligence (see discussion with Charles Stross and Paul Krugman here).
[...] Facebook vs. Google; Translation Methods | Apogee Communications Blog I recently used Google’s Language Tools for the first time in several months. They’ve added a number of languages, and the translations are definitely clearer than they once were. Currently, they claim over 2500 language pairs between … [...]
Even though they’ve added more languages, they are still translation tools, not human translators. Everyone already knows the capability and limitations of these tools right, and how they provide translations. Now its really your call which of the both methods do you prefer when translating few words to large documents.