Well, if you can assume that the data will always follow the pattern you presented, then you can split on a space:

my @words=(
  "12 chem1",
  "1/2 chem2",
  "(N+1) chem3",
  "2M chem4",
  "N chem5"
);

# given an array of words, returns an 
# array of corresponding prefixes

sub get_prefixes {
  map { (split)[0] } @_;
}

print join(",",get_prefixes(@words));
[download]

-Vlad