I saw a presentation where someone took thousands of English words, placed them in a high-dimensional space based on, I think, what other words they co-occurred with, ran PCA on this space, and analyzed the top 4 resulting dimensions. The top dimension was “good/bad”.
This sounds like semantic differential research. The standard finding is three dimensions: good-bad (evaluation), strong-weak (potency), and active-passive (activity).
I saw a presentation where someone took thousands of English words, placed them in a high-dimensional space based on, I think, what other words they co-occurred with, ran PCA on this space, and analyzed the top 4 resulting dimensions. The top dimension was “good/bad”.
You already said so in this very thread :)
This sounds like semantic differential research. The standard finding is three dimensions: good-bad (evaluation), strong-weak (potency), and active-passive (activity).
Do you remember the other three?
See here