Firstly; and I could be wrong, but it's my expeirence that; the serches here ignore the second word. Well actually a search is performed for each word and the result lists are merged.

A regexp search while probakly not too hard since this site uses MySQL, which has RLIKE, could be mighty dangerous. Search is already kind of heavy on resources, and I could foresee an improperly constructed regexp running out of control.

--
perl -pew "s/\b;([mnst])/'$1/g"