One thing I've never seen brought up in the Chinese Room argument is the later Wittgenstein's view of language where meaning is use. If that's the case, then a system that perfectly translates English to Chinese does not understand anything, because symbol translation is not fundamentally different form looking up words in a dictionary, and definitions are not meaning under language is use.

We would have to say that such a system, if it's possible, cannot understand language unless it can do all the other things with words that humans do. It must participate in the various languag games. So a robot that perfectly uses English or Chinese in all the situations humans use it would understand by definition.

Therefore, on the view that meaning is use, Searle was correct, at least the way he set up the thought experiment.