Putting Sign Language AI into Users Hands(deepmind.google) |
Putting Sign Language AI into Users Hands(deepmind.google) |
The videos show it translating ASL as it's naturally used; without making the users adapt their signing to English grammar. That's a big leap forward.
Because iconicity (signs that look like what they mean - the sign language equivalent to onomatopoeia in spoken languages) is so common in sign languages, it's much easier for users of different languages to communicate. Something like 30% to 50% of vocabulary in sign languages is iconic, so it's much easier to guess at the meanings of words from someone else's language.
It's like if the delegates showed up to the first UN meeting and tried to communicate like this: https://old.reddit.com/r/TikTokCringe/comments/1vmqc2i/peppa...
Speakers are prioritizing words and grammar they think will be understood by their conversation partners. Then, over time, people started to standardize, and in some cases write down, which words were best used in that context.