diff --git a/dan/datasets/extract/arkindex.py b/dan/datasets/extract/arkindex.py
index 93742351d4a73bd13f5299d1cc1bd5268472ae1e..12dff600a54c7b8768c973b8d759bae515a36911 100644
--- a/dan/datasets/extract/arkindex.py
+++ b/dan/datasets/extract/arkindex.py
@@ -106,9 +106,10 @@ class ArkindexExtractor:
             raise NoTranscriptionError(element.id)
 
         transcription = random.choice(transcriptions)
+        stripped_text = transcription.text.strip()
 
         if not self.tokens:
-            return transcription.text.strip()
+            return stripped_text
 
         entities = get_transcription_entities(
             transcription.id,
@@ -117,7 +118,7 @@ class ArkindexExtractor:
         )
 
         if not entities.count():
-            return transcription.text.strip()
+            return stripped_text
 
         return self.translate(
             entities_to_xml(