commit 7e4cd6c854ac74ffb885907f172243c5ef8982a1
author Marc Kupietz <kupietz@ids-mannheim.de> Thu Dec 15 18:34:37 2022 +0100
committer Marc Kupietz <kupietz@ids-mannheim.de> Fri Dec 16 16:10:26 2022 +0100
tree 37080409a565e605259f8ef6608438727d624885
parent 12f64e4f3925ef0d0910633d1abe816ce4b324bf

Handle lonely docid attributes in NKJP KorAP-XML

Change-Id: Id439a98ac46e2bbd33e65b1719cca39cd11b6377

script/korapxml2conllu

diff --git a/script/korapxml2conllu b/script/korapxml2conllu
index 5e91c06..9cb25f8 100755
--- a/script/korapxml2conllu
+++ b/script/korapxml2conllu
@@ -159,7 +159,7 @@
         }
         $filename=$1 if(!eof(MORPHO_OR_TOKENPIPE) && /\s+inflating:\s+(.*)/);
       }
-    } elsif(m@^\s*<layer\s+.*docid="([^"]+)"@) {
+    } elsif(m@(?:^|\s)docid="([^"]+)"@) {
       last if($test && $text_no++ > 3);
       if(!$first) {
         closeDoc(0);
@@ -340,8 +340,11 @@
       $docid=$1;
       $log->debug("Getting plain text for $docid");
       $text_started=0;
-    } elsif(/<layer[^>]+docid="([^"]*)/) {
+    } elsif(/<raw_text\b/) {
+      $text_started=0;
+    } elsif(/(?:^|\s)docid="([^"]*)/) {
       $docid=$1;
+      $log->debug("Getting plain text for $docid");
     } elsif(m@<span @) {
         ($current_id) = /id="[^0-9]*([^\"]*)"/;
         ($current_from) = /from="([^\"]*)"/;