Old News travel Slow

[_rck_]

Sometimes it takes a while for information to get to me, but looks like Google cleaned up the Machine Translation Competition of NIST this year; they won best in every category.

BlondKiwi always predicted that this was going to happen, because they have enough documents and enough machines to do the learning of parallel corpus translations that

Comments

[anonymous]

If I understand the scores correctly, it looks like even google has a way to go yet, especially wrt Chinese. And it looks like some of the other teams, like ISI and UMD are still competitive.

Given Google's performance in the Large Data Track (which I believe all teams shared) it would seem their algorithms are better independently of their larger training corpus. Now I really want to grok the details and differences of each team's approach. Good thing I'm taking an NLP course in the fall term!

if you find out ...

[_rck_]

... please post here; I am not sure that I have enough of the background needed to study what they are doing or working on