#!/usr/local/bin/perl

################################################################
#  Simple IR system
#  
#  Usage:
#    simpleir.perl keywords...
#
#  Description:
#    Read the source code.
#    It assume IREX SGML data for the target.
#  
#  Author:
#    Satoshi Sekine (New York University)
#
#  You can copy, modify, redistribute the code, as long as
#  this header is kept. The source code is provided "AS IS".
#  The author has no responsibility for any damage or anything
#  caused by the code. No copyright is claimed.
################################################################

$SYSTEM_ID = 1199;
$TOPIC_ID = 1003;

@FILES=( "$ENV{IREX_ROOT}/MAINICHI/mai94.sgml",
         "$ENV{IREX_ROOT}/MAINICHI/mai95.sgml" );

for $f ( @FILES ){
  open( FILE,"< $f" )|| die "Can't open file : $f";
  while( <FILE> ){
    if( /<DOCNO>(\d+)<\/DOCNO>/ ){
	$docno = $1;
    }
    $len += length( $_ );
    for $argv ( @ARGV ){
      $match += grep( /$argv/,$_ );
    }
    next unless $_ eq "</DOC>\n";

    if( $match > 0 ){
        $sim{$docno} = $match / log( $len );  
        # You can make this more sophisticated.
    }

    $len=0;
    $match = 0;
  }
  close FILE;
}

# Print IREX format output
#
print "<SYSTEM-ID>$SYSTEM_ID</SYSTEM-ID>\n";
print "<IR-RESULT>\n";
print "<TOPIC-ID>$TOPIC_ID</TOPIC-ID>\n";
$i=1;

foreach $docno ( sort {$sim{$b} <=> $sim{$a}} keys %sim ) {

       print "<A-ID RANK=$i SIM=$sim{$docno}>$docno</A-ID>\n";
       if($i++>=300){ last; }
}

print "</IR-RESULT>"