source: trunk/gsdl/perllib/Kea.pm@ 9674

package Kea;

# This function is called by BasPlug.pm when a flag in a collection
# configuration document specifies that keyphrase metadata must be gathered for
# that collection.
# It is passed as arguments, the documents text and possibly some options for
# how the keyphrase data is to be collected if the keyphrase option flag was
# set in the collection configuration file.  This module then writes the
# documents text to a file in a temporary directory because the stand-alone program Kea which will be
# called to do the actual extraction of the keyphrases expects a directory with one or more files as argument.
# Once Kea has been called upon, the file containing the keyphrase data
# gathered by Kea should be stored in gsdl/tmp and this file is read, the data
# we are interested in is extracted and passed back to BasPlug.pm in an
# appropriate format.

# returns a string containing comma-separated keyphrases
sub extract_KeyPhrases {

    # Parsing arguments of the function
    my $doc = shift(@_); # documents text  
    my $args = shift(@_); # any options
    my @optionlist = split(/\s+/, $args) if (defined($args)); #list of options

    # Specifying directory names
    my $gsdlhome = $ENV{'GSDLHOME'}; 
    my $keahome = "$gsdlhome/packages/kea/kea-3.0";
    my $defaultmodel = "$keahome/CSTR";
    
    # Initializing variables:
    my $command = ""; 
    my @keylist; 
    my @options = ();
    $modelspec = 0;

    # Settings for the java executable: 

    # CLASSPATH:
    $java_classpath = ".:$keahome";

    # See if java executable is on path
    my $java_exec="";
    if (system("which java >/dev/null 2>/dev/null")==0) {
    $java_exec=`which java`;
    chomp $java_exec;
    } else {
    $java_exec="$java_home/bin/java";
    }
    
    # The actual java command is based on these other variables:
    $java_command = "$java_exec -classpath \"$java_classpath\"";
    
    # end of java settings

    # Parsing options for keyphrase extraction:
    if (@optionlist) {
    foreach $element (@optionlist){ #for each option
        if (length($element) == 1) {
        push(@options, "-$element");
        } else {
        $option = substr($element, 0, 1);
        $value = substr($element,1);
        if (($option eq "m") && (-e "$keahome/$value")) {
            $modelspec = 1;
            push(@options, "-$option $keahome/$value");
        } elsif ($option eq "m") {          
            $modelspec = 1;
            print STDERR "Couldn't find model $value. Using the default model instead\n";
            push(@options, "-$option $defaultmodel");
        } else {
            push(@options, "-$option $value");
        }
        
        }
    }
    # if none of the option specifies the model, set the default one:
    if ($modelspec != 1) {
        push(@options, "-m $defaultmodel"); 
    }
    $options = join(" ",@options);
    # print STDERR "OPTIONS: $options\n";
    } else {
    # If no options were specified: Set default value for the model
    $options = "-m $defaultmodel";  
    }

    # Remove all HTML tags from the original text
    $doc =~ s/<P[^>]*>/\n/sgi;
    $doc =~ s/<H[^>]*>/\n/sgi;
    $doc =~ s/<[^>]*>//sgi;
    $doc =~ tr/\n/\n/s;

    # Write text to a temporary file doc.txt 
    open(OUT, ">$gsdlhome/tmp/doc.txt") or die "In Kea.pm doc.txt could not be created\n";  
    print OUT $doc;
    close(OUT);


    # EXECUTE KEA with specific options:
    $command = "$java_command KEAKeyphraseExtractor -l $gsdlhome/tmp $options";
    system ("$command");

    # Read the resulting doc.key, which contains keyphrases:

    open(IN, "<$gsdlhome/tmp/doc.key") or return "";
                                          #this means doc.key does not exist
                                          #either because an option was wrongly specified
                                      #or no keyphrases were found
    while(<IN>){
    chomp;
    push(@keylist,$_);
    }
    close(IN);

    $keylist = join(", ", @keylist);

    # Delete doc.key so that in future it will not be opened and read.
    # Otherwise KEA sees it as more keyphrases!

    unlink("$gsdlhome/tmp/doc.key");
    
    return $keylist;
}

1;