Introduction to Statistical Machine Translation

&#62; The biggest single advance seems to have been a movement away from words as the unit of language, and towards phrase-based models, which give greatly improved performance.<p>This really struck me, as someone who both teaches and studies language. If it works better for computers, I wonder how much better it would work for people. If there's anything I've noticed in my time in Japan, it's that the Japanese approach seems to be to nail down every single word with a single Japanese meaning and stick to that meaning all the time, which leads to a lot of very Japanese-sounding English.

&#62; Note: English is my only language, which makes it hard for me to construct translation examples!<p>Ouch.<p>Building MT systems without knowing foreign languages is a bit like deaf people building a speech recognizer. No offence meant to anyone, but it works much better when you know what you're doing.