In a July 6, 2022 Facebook post, Zuckerberg explained why Meta AI’s recent No Language Left Behind (NLLB) project merits attention. Specifically, Meta AI tweeted, the company built an AI model capable of translating between 200 languages — for a total of 40,000 different translation directions. The authors reported that their model achieved a 44% improvement in BLEU, thus “laying important groundwork towards realizing a universal translation system.” But, as to be expected, improvement was not uniform across language pairs, with little to no improvement for pairs such as Armenian into English or French into Wolof.
Comments about this article
Cameroon
Local time: 03:12
French to English
+ ...
A 200- languages directory , it is what this tool looks like .It seems attractive yet I prefer to wait and see ...
Portugal
Local time: 03:12
Member (2007)
English to Portuguese
+ ...
What Meta claims is... a meta-claim!
France
Local time: 04:12
Member (2003)
English to Dutch
+ ...
Let's wait and see, but I'm not surprised.
Cheers,
Gerard
Netherlands
Member (2006)
German to Dutch
It's nice that they made their source code available to the open domain, but for me the most useful aspect of MT is that when tons of technical documents are harvested and indexed, chances are good that I find the correct translation of a rare and very specific technical term.
A small corpus won't be of any use for me in my professional life.
Local time: 06:12
English to Arabic
+ ...
Could somebody perhaps explain what Mark meant by "50 billion parameters" & "25 billion translations", because he didn't in the article? Or, perhaps he explained it somewhere else?
Also, are the "across our apps" paid or free? He didn't explain that either!
To report site rules violations or get help, contact a site moderator:
You can also contact site staff by submitting a support request »
This discussion can also be accessed via the ProZ.com forum pages.