Changeset 2064

Timestamp:

2001-02-24T14:03:24+13:00 (23 years ago)

Author:

paynter

Message:

Sort thesaurus phrases by frequency then type.

File:

• trunk/gsdl/perllib/classify/phind.pm (modified) (8 diffs)

trunk/gsdl/perllib/classify/phind.pm

@@ -1138 +1138 @@
     # 2. 
     # Read phrases file to find thesaurus entries that already
-    # have a phindid.  Store their phind-ids for later translation.
+    # have a phindid.  Store their phind-ids for later translation,
+    # and store their frequency for later sorting.
     my %thesaurustophindid;
-    my ($phindid);
+    my %phindidtofrequency;
+    my ($phindid, $freq);
 
     open(IN, "<$infile");
@@ -1152 +1154 @@
     $phindid = shift @fields;
     $symbols = shift @fields;
+    $freq = shift @fields;
 
     # do we have a thesaurus id corresponding to this phrase?
@@ -1157 +1160 @@
         $thesid = $symbolstothesid{$symbols};
         $thesaurustophindid{$thesid} = $phindid;
+        $phindidtofrequency{$phindid} = $freq;
     }
     }
@@ -1164 +1168 @@
 
     # 3.
-    # Create phind-id numbers for remaining thesaurus entries
+    # Create phind-id numbers for remaining thesaurus entries,
+    # and note that their frequency is 0 for later sorting.
     my $nextphindid = $phindid + 1;
 
@@ -1177 +1182 @@
     if (!defined($thesaurustophindid{$thesid})) {
         $thesaurustophindid{$thesid} = $nextphindid;
+        $phindidtofrequency{$nextphindid} = 0;
         $nextphindid++;
     }
@@ -1186 +1192 @@
     # phind-id numbers.
     my $newthesaurusfile = &util::filename_cat($phinddir, "$thesaurus.phindid");
-    my ($relations, $linkcounter, $linktext, $linktype, @linkdata, $link);
+    my ($relations, $linkcounter, $linktext, $linktype, @linkdata);
+    my (@links, $linkid, %linkidtotype, $newrelation);
 
     open(TH, "<$thesaurusfile");
@@ -1203 +1210 @@
 
     # convert each part of the relation string to use phind-id numbers
-    $newrelation = "";
-    $linkcounter = 0;
+    # at the same time, we want to sort the list by frequency.
+    undef %linkidtotype;
+    
     foreach $linktext (split(/;/, $relations)) {
         @linkdata = split(/,/, $linktext);
-
+        
         # remember the linktype (e.g. BT, NT)
         $linktype = shift @linkdata;
-        $newrelation .= "$linktype,";
-
-        # convert the link target identfiers
-        foreach $link (@linkdata) {
-        die unless (defined($thesaurustophindid{$link}));
-        $newrelation .= "$thesaurustophindid{$link},";
-        $linkcounter++;
+        
+        # store the type of each link
+        foreach $thesid (@linkdata) {
+        die unless (defined($thesaurustophindid{$thesid}));
+        $linkidtotype{$thesaurustophindid{$thesid}} = $linktype;
         }
-        $newrelation =~ s/\,$//;
-        $newrelation .= ";";
-    }
-    $newrelation .= ":";
-
-    print TO "$phindid:$symbols:$linkcounter:$newrelation\n";
+    }
+
+    # sort the list of links, first by frequency, then by type.
+    @links = sort { ($phindidtofrequency{$b} <=> $phindidtofrequency{$a}) 
+                        or ($linkidtotype{$a} cmp $linkidtotype{$b}) } (keys %linkidtotype);
+    $linkcounter = (scalar @links);
+
+    # create a string describing the link information
+    $linktype = $linkidtotype{$links[0]};
+    $newrelation = $linktype;
+    foreach $linkid (@links) {
+        if ($linkidtotype{$linkid} ne $linktype) {
+        $linktype = $linkidtotype{$linkid};
+        $newrelation .= ";" . $linktype;
+        }
+        $newrelation .= "," . $linkid;
+    }
+    $newrelation .= ";";
+    
+
+    # output the new line
+    print TO "$phindid:$symbols:$linkcounter:$newrelation:\n";
     }
     close TH;
@@ -1229 +1251 @@
 
     undef %thesaurustophindid;
+    undef %linkidtotype;
+    undef %phindidtofrequency;
 
     # 5.