sort {length $columnA <=> length $columnB} 

##</code><code>##

sort {length $a <=> length $b} 

##</code><code>##

Global symbol "$columnA" requires explicit package name (did you forget to declare "my $columnA"?) 

##</code><code>##

print "matches first" if "a (word) in context " =~ m/\b\(word\)\b/; 
print "matches second" if "a (word) in context " =~ m/\(word\)/; 

##</code><code>##

print "matches third" if "a (word) in context " =~ m/(?<=\s)\(word\)(?=\s)/; 

##</code><code>##

print "matches fourth" if "a a in context (word) " =~ m/(?<!\w)\(word\)(?=\s|$)/; 


##</code><code>##

#!/usr/bin/env perl
use strict;
use warnings;
use Data::Dumper;

my %dic;

while (<DATA>) {
    chomp;
    my($columnA, $columnB) = split("\t", $_);
    $dic{$columnA} = $columnB;
}
print Dumper \%dic;

my $regex = join ( "|", map {quotemeta} sort { length $a <=> length $b } keys %dic ); 
   $regex = qr/(?<!\w)($regex)(?=\s|$)/;
print $regex;

my $sentence = "this word word (word) is a word\n";
$sentence =~ s/$regex/$dic{$1}/g; 
print $sentence;

__DATA__
word word	parola parola
(word)	(parola)