predict case patterns of verbs using not Goi-taikei but a probabilist…

…ic model.
mamorlis · Mar 29, 2007 · 75ee2da · 75ee2da
1 parent 9b9b995
commit 75ee2da
Show file tree

Hide file tree

Showing 3 changed files with 36 additions and 31 deletions.
diff --git a/resolveZero/script/NCVTool.pm b/resolveZero/script/NCVTool.pm
@@ -32,11 +32,16 @@ sub new {
         type  => 'newswire',
         prob  => 'MI',
         model => 'n1000',
+        pat   => 'ncv',
         @_,
     );
     my $self  = {};
     bless $self, $class;
 
+    opendir 'DIR', $model_dir{$scorer_params{type}};
+    my @file = grep /$scorer_params{model}/, readdir DIR;
+    closedir DIR;
+
     my $model_path = $model_dir{$scorer_params{type}}.'/'.$scorer_params{model};
     my $model_prob = $scorer_params{prob};
     if ($model_prob eq 'chi2') {
@@ -45,6 +50,8 @@ sub new {
     }
     die "Cannot find models in $model_path" if ! -e $model_path.'.pz.da';
     my ($pid, $out, $in);
+    $scorer = 'scorer -u 2 -m Pos -p ' if ($scorer_params{pat} eq 'cv');
+
     $pid = open2 $out, $in, "$scorer $model_prob -d $model_path";
 
     $self->{pid} = $pid;

diff --git a/resolveZero/script/predictCasePattern.pl b/resolveZero/script/predictCasePattern.pl
@@ -0,0 +1,27 @@
+#!/usr/bin/env perl
+
+use strict;
+use warnings;
+
+use FindBin qw($Bin);
+unshift @INC, $Bin;
+use NCVTool;
+my $ncvtool = new NCVTool(model=>'n50', pat=>'cv');
+
+sub predicat_case_pattern {
+    my $pred = shift; # Bunsetsu
+    my @type = ('GA');
+    if ($pred->PRED) {
+	my %case = ( '¤ò'=>'WO', '¤Ë'=>'NI' );
+	for my $case (keys %case) {
+ 	    my $q = $case.':'.$pred->PRED;
+	    my $score = $ncvtool->get_score($q);
+	    push @type, $case{$case} if ($score and $score > 0) ;
+	}
+    } else {
+	@type = ('GA', 'WO', 'NI') if ($pred->HEAD_POS =~ /^Æ°»ì/);
+    }
+    return @type;
+}
+
+1;
diff --git a/resolveZero/script/resolveZero.pl b/resolveZero/script/resolveZero.pl
@@ -31,23 +31,7 @@ BEGIN
 require 'centerList.pl';
 require 'extractFeatures.pl';
 require 'cab.pl';
-
-use FindBin qw($Bin);
-my $v2type = $Bin.'/../../dict/db/v2type.db';
-my %db;
-{
-    no strict "subs";
-    if (eval "require BerkeleyDB; 1") {
-        tie %db, 'BerkeleyDB::Hash',
-            -Filename => $v2type,
-            -Flags    => DB_RDONLY,
-            -Mode     => 0444
-            or die "Cannot open $v2type:$!";
-    } elsif (eval "require DB_File; 1") {
-        tie %db, 'DB_File', $v2type, O_RDONLY, 0644
-            or die "Cannot open $v2type:$!";
-    }
-}
+require 'predictCasePattern.pl';
 
 my $rootPath = $scriptPath; $rootPath =~ s|[^/]+/$||;
 
@@ -120,30 +104,17 @@ sub resolve_zero {
             }
 	    if ($b->PRED) {
 		my $pred = $b;
-		my @type = ();
-		if ($db{$b->PRED}) {
-		    @type = split ' ', $db{$b->PRED};
-		} else {
-		    if ($b->HEAD_POS =~ /^動詞/) {
-			@type = ('GA', 'WO', 'NI');
-		    } else {
-			@type = ('GA');
-		    }
-		}
+ 		my @type = &predicat_case_pattern($pred);
 		for my $type (@type) {
-# 		    print STDERR 'intra ant', "\t", $type, "\n";
 		    my $intra_ant = &identify_antecedent_intra_bact($pred, $s, $type, $cl, $mRef->{intra}->{ant}->{$type});
-# 		    print STDERR 'intra ana', "\n";
 		    my $intra_score = 0;
 		    if ($intra_ant) {
 			$intra_score = &detemine_anaphoricity_intra_bact($s, $pred, $intra_ant, $type, $cl, $mRef->{intra}->{ana}->{$type});
 		    }
 		    if ($intra_score > $intraParam{$type}) {
 			$pred->{$type} = $intra_ant;
 		    } else {
-# 			print STDERR 'inter ant', "\n";
 			my $inter_ant = &identify_antecedent_inter_bact($pred, $s, $t, $type, $cl, $mRef->{inter}->{ant}->{$type});
-# 			print STDERR 'inter ana', "\n";
 			my $inter_score = 0;
 			if ($inter_ant) {
 			    $inter_score = &detemine_anaphoricity_inter_bact($s, $pred, $inter_ant, $type, $cl, $mRef->{inter}->{ana}->{$type});