If I'm understanding you correctly, you could just collect things in a hash, keyed by the first column:

my %rows;
while (<DATA>) {
    my ($key, $c1, $c2, $c3) = split;
    push @{ $rows{$key} }, "${c1}_$c2,$c3";
}

for my $key (keys %rows) {
    print join("\t", $key, @{ $rows{$key} }), "\n";
}

__DATA__
frog-n    as    novelty-n    5.8504
frog-n    be    yellow-n    6.1961
frog-n    be-1    Asia-n    5.0937
frog-n    coord    zebra-n    5.9279
frog-n    coord-1    Canuck-n    6.3363
frog-n    nmod-1    mule-n    4.2881
amphibian-n    success-1    surprising-j    14.6340
amphibian-n    such_as    alligator-n    11.5265
amphibian-n    than    work-n    5.9948
amphibian-n    though    stalk-n    13.2228
[download]

Output:

frog-n  as_novelty-n,5.8504     be_yellow-n,6.1961      be-1_Asia-n,5.
+0937      coord_zebra-n,5.9279    coord-1_Canuck-n,6.3363 nmod-1_mule
+-n,4.2881
amphibian-n     success-1_surprising-j,14.6340  such_as_alligator-n,11
+.5265     than_work-n,5.9948      though_stalk-n,13.2228
[download]

(Note that the rows in the unsorted hash printout are only "by accident" in the same order the keys occurred the input data.  In case ordering matters, you'd need to take care of it separately (e.g. using Tie::IxHash).)