Comment by liampulles 2 days ago
If you take the embedding for king, subtract the embedding for male, add the embedding for female, and lookup the closest embedding you get queen.
The fact that dot product addition can encode the concept of royalty and gender (among all other sorts) is kind of magic to me.
I have seen this particular work example to work. You don't get the exact match but the closest one is indeed Queen.
Yes but it doesn't generalize very well. Even on simple features like gender. If you go look at embeddings you'll find that man and woman are neighbors, just as king and queen are[0]. This is a better explanation for the result as you're just taking very small steps in the latent space.
Here, play around[1]
mother - parent + man = woman
father - parent + woman = man
father - parent + man = woman
mother - parent + woman = man
woman - human + man = girl
woman - man + man = girl
man - man + man = woman
woman - woman + woman = man
Thank you for the comment!
This led me to do a bit more research, and I see indeed the queen result is in itself infact "cheating" a bit: https://blog.esciencecenter.nl/king-man-woman-king-9a7fd2935...
#TheMoreYouKnow
Shouldn't this itself be a part of training?
Having set of "king - male + female = queen" like relations, including more complex phrases to align embeddings.
It seems like terse, lightweight, information dense way to address essence of knowldge.
This was actually shown to not really work in practice.