It’s not really, ASL is it’s own distinct language from English, like most sign languages. ASL is most closely related to French Sign Language, which still isn’t anything related to French and has its own linguistic features and independent development. Because of this translating ASL to English (or any other language) is no easier than translating to or from any other spoken language.

Sure, but machines are actually pretty good at spoken and written translation these days, so I would expect that once the image recognition is solved they could handle ASL readily as well.

Oh absolutely, I agree. I just wanted to make certain it was clear that it isn’t as simple as classifying images as words.

Since ASL doesn't have a massive corpus of scrabeable training data (at least I would think). It's probably on the harder side when it comes to creating machine translators.

There must be a huge corpus of TV broadcasts, filmed live events etc. where a person speaking is being translated into sign language in real time.

Those are usually subtitled.

in theory with a mastery of speech recognization one would assume you could then apply that to any diction that has an accompanying ASL translator to derive ASL elements from the contexts given by the spoken word.

in practice that's all difficult, if at all possible -- but just sayin; we have decently good speech interpretation at this point, perhaps we aren't far from self-training ASL against something similar.

Usually, but I'm referring to the broadcasts that have a live sign language interpreter on-screen.

If Apple is filming its ASL employees, it’s about to get a lot more

I suspect that buoys and placement make signed languages harder for machine translation than spoken ones.

And to add on a little, ASL is like any other language in that it has regional dialects. It's why many people introduce themselves with their name and where they learned to sign.

One of the more common examples from teachers is the sign for SLOW. On the west coast, we sign it as the flat palm being pulled slowly back along the extended non-dominant arm. But in the south and parts of the east, it is signed as the Y sign, thumb against the cheek, being tilted forward.

The way you fix this casually is to ask the other person to finger spell or mime what they meant. I can't see AI being able to do that well for a while but maybe..?

Never, the whole point is to employ ASL fluent people to keep the culture and language alive.

> ASL seems like an image recognition task

Not trying to be trite, but I imagine you don't know ASL. It is an extremely loose and context dependent language really. I am sure it can be done eventually, but I still think it's very far off in terms of recognition. Would love to find out more about the state of the art for ASL or BSL recognition though.

You mean a video recognition task.