Skip-gram window (size 2)
"The cat sat on the mat"
center = cat
context = {The, sat, on}
network learns: dot(v_cat, u_sat) high, dot(v_cat, u_random) low
Embedding space (2D projection of learned vectors)
cat
dog
puppy
kitten
king
man
queen
woman
king - man ≈ queen - woman
co-occurring words cluster. analogy vectors are directions.