changed tokenization

ref_seg_ger.py

@@ -57,7 +57,7 @@ _URLS = {
 
 _LABELS = [
     'publisher', 'source', 'url', 'other', 'author', 'editor', 'lpage',
-    'volume', 'year', 'issue', 'title', 'fpage', 'identifier', 'edition'
+    'volume', 'year', 'issue', 'title', 'fpage', 'edition'
 ]
 
 _FEATURES = datasets.Features(
@@ -229,6 +229,8 @@ class RefSeg(datasets.GeneratorBasedBuilder):
 
                 # tokenized_input = row['token'].split(' ')
                 tkn = self.TOKENIZER.pre_tokenize_str(row['token'])
+                if row['label'] == 'identfier':
+                    row['label'] = 'other'
                 if not tkn:
                     continue
                 tokenized_input, offsets = zip(*tkn)