A while ago with the teacher's class, learned some language and translation and other algorithms, the correct match rate is relatively low.

That's not even asking for a reduction in spatial performance.

In layman's terms, let's say so-and-so school has a banner that says "Welcome to the new teacher's life".

Reading it carefully, it's a misnomer: does it say "Welcome new teachers" or "Welcome new teachers"?

But it doesn't affect our human understanding, because under normal circumstances we humans would understand that it's not possible to say that they should be allowed to dine "before they are born".

It's a joke, but in machine translation, it could be translated as "let them come and eat while they are alive".

The corpus is perfect, the algorithm is not, and the whole thing is still not.

There is a relationship with the hardware, but the relationship can be said to be big, can also be said to be not big, because the hardware development is quite fast, the constraints are not particularly big, even if your algorithms are particularly complex, the corpus is particularly large, the language is high performance, we still have a trick: networking, ah, the need to deal with the data to the server a throw.

As for, the answer upstairs is yes, it's a matter of time, but it may be a long time, a long time, a long time.

Why? No one can say when it comes to a critical point, an invincible machine translation algorithm is suddenly born, like deep learning algorithms directly promote the rapid development of AI.