#!/usr/local/bin/perl

################################################################################################
# Name:     Bridget Thomson-McInnes
# Program:  Word Sentence Generator
################################################################################################
# Program Description:
#   This program reads in a training corpus - using it to create an n-gram
#   model.  The program then uses this n-gram model to generate sentences.
#   The sentences are generated by calculating the probability of a word 
#   the n-1 words before it; normalizing the probabilities of the possible
#   words. The program picks a random word by producing a random number
#   between 0 and 1 and then picking the appropriate word.
#
# Usage: ./generator.pl <n-gram> <number of sentences> <training corpus>
################################################################################################

#  Get the command line variable n and m
#  m-grams n-number of sentences
$m = shift @ARGV;
$n = shift @ARGV;

#  complain if less than 2 arguments
if (!defined $n || !defined $m) {
    print STDERR "Not enough arguments to calculate. Retry. \n";
    print STDERR "USAGE: generator.pl <n-gram> <no. of sentences> <corpus>\n";
    exit();
}

#  initialize variables
my %m_gram  = (); my %m_1gram  = (); my %hash  = (); 
my $m_total = 0;  my $m_1total = 0;  my @queue = (); 

#  initialize the input line
$line = "";  for(1..$m-1) { $line .= "<start> "; }

#  Load word sequences into a hashtable
while (<>) { 
    chomp; if(! (m/^\s*$/) ) { s/[,;:.!?]/ $& /g; s/[("'"-)]//g; $line .= $_;}
    
    #  store the n-gram word sequence in a hash table
    foreach ( (split/\s+/, $line) ) {               
	#  get the m-1 gram, push a word on the queue
	if($#queue == $m-2) { $m_1gram{ (join " ", @queue) }++; $m_1total++; } push @queue, $_;

	#  get the mgram and the hash
        if ($#queue == $m-1) {
	    my @array = @queue; $temp = pop @array; push @{$hash{ (join " ", @array) }}, $temp;
	    $m_gram{ (join " ", @queue) }++; shift @queue; $m_total++;
	    if($queue[$#queue]=~m/[.!?]/) { @queue = (); for(1..$m-1) { push @queue, "<start>"; } }
	} 
    } $line = "";
}


#  print out n sentences
for $j(1..$n) {
    print "Sentence $j: "; 
    
    #  initialize the start of the sentence
    my @sequence = (); for (1..$m-1) { push @sequence, "<start>"; } 
    
    do {  
	#  get the next word possiblities and store them in the choices hash;
        my @random = (); my $count = 0; my $seq = join " ", @sequence; my %seen = ();  
	my $freq = $m_1gram{$seq}/$m_1total; my @unique = grep { ! $seen{$_} ++ } @{$hash{$seq}};

	for (0..$#unique) {
            $mgram = $seq . " " . $unique[$_]; $mgram=~s/\s+$//g; $mgram=~s/^\s+//g;
	    if($m > 1) { $random[$_] = $count + ( ($m_gram{$mgram}/$m_total) / $freq ); }
	    else { $random[$_] = $count + ( $m_gram{$mgram}/$m_total ); } $count = $random[$_];
	}

	$chosenWord = ""; my $number = rand 1;      
	for (0..$#random) { if($number < $random[$_]) { $chosenWord = $unique[$_]; last; } }
	
        push @sequence, $chosenWord; shift @sequence; print "$chosenWord "; 
	
    } until($chosenWord=~/[.!?]/); print "\n\n";
    
}