#!/usr/local/bin/perl -w
#-----------------------------------------------
#
#
#  三階のマルコフ連鎖による学習とテキスト生成
#
#
#-----------------------------------------------
#
#　使用法
# $ perl l1.pl < text


use strict;
use Text::ChaSen;

my %Textfiber=();
my $Maxhop=10;

&initialize();

# ----------------------
#        M A I N
# ----------------------

srand;
my $input;
my @tokens;

#一文ずつ学習
while(<>)
{
	$input=$_;
	@tokens = &analyze($input);
	&learn3(@tokens);
}

#記憶のダンプ
#&dump3;

#ランダム文字列生成
for(0..50)
{
	print &generate3("_HEAD")."\n";
}

# ------------------------------------------------------------------------
#
#                             L E A R N I N G
#
#                         順方向で三階のMarkov連鎖
#
# ------------------------------------------------------------------------

sub learn3(@)
{
	my @input = @_;
	my $prefix1 = "_HEAD";
	my $prefix2 = "_HEAD2";
	my $surfix;
	
	push (@input,"_TAIL");
	for $surfix (@input)
	{
		push(@{$Textfiber{$prefix1}},"$prefix2/$surfix");
		$prefix1=$prefix2;
		$prefix2=$surfix;
	}
	
	push(@{$Textfiber{$prefix1}},"_TAIL");
}

sub dump3()
{
	my ($key,$k);
	print "dumping %Textfiber\n";
	for $key (keys %Textfiber)
	{
		print "{".$key."}-> ".join(" , ",@{$Textfiber{$key}})."\n";
	}
}

sub generate3($)
{
	my $prefix=shift;
	my (@output,@cand,$surfix,$surfix2);

	#headはpushしない
	
	LOOP:
	for(1..$Maxhop)
	{
		@cand=@{$Textfiber{$prefix}};
		last if($#cand == -1);
		$surfix=$cand[rand($#cand+1)];
		
		@cand=split (/\//,$surfix);

		for $surfix2 (@cand)
		{
#			print "[$surfix2]";
			last LOOP if $surfix2 eq "_TAIL";
			push (@output,$surfix2) unless ($surfix2 eq "_HEAD2");
			$prefix=$surfix2;
		}
	}
	return join("/",@output);
}

# ------------------------------------------------------------------------
#
#
#                              A N A L Y S I S
#
#
# ------------------------------------------------------------------------

sub analyze($)
{
	my $input = shift;
	my @tokens;
	
	# 特定文字のエスケープ
	
	for($input)
	{
		s/([%,)(\/])/'%'.unpack("H2",$1)/eg;
	}

	# 分かち書き
	$input=Text::ChaSen::sparse_tostr($input);
	chomp $input;
	@tokens=split /\//,$input;
	
	return @tokens;
}



sub initialize()
{
	Text::ChaSen::getopt_argv('chasen','-F','%m/');
}