According to the Wikipedia article, the L in $LR(k)$ means "left-to-right scan", and the "R" means "rightmost derivation." However, in Knuth's original paper on $LR(k)$ grammars, he defines $LR(k)$ (on page 610) as a language that is "translatable from left to right with bound $k$."
I am guessing that this new terminology was chosen to complement $LL(k)$ parsing's "left-to-right scan, leftmost derivation." That said, I don't know when the terminology changed meaning.
Does anyone know where the newer acronym for $LR(k)$ comes from?