diff --git a/newshomepages/analyze/drudge.py b/newshomepages/analyze/drudge.py
index 1abc0e41d86..0c3f232ee7c 100644
--- a/newshomepages/analyze/drudge.py
+++ b/newshomepages/analyze/drudge.py
@@ -57,8 +57,8 @@ def get_lemma(headline: str):
         # Read it into our NPL thing
         doc = nlp(headline)
 
-        # Parse out all the words
-        token_list = [token.strip() for token in doc if token.strip()]
+        # Parse out all the words, and remove empty ones
+        token_list = [token for token in doc if token.text.strip()]
 
         # Remove stop words
         token_list = [t for t in token_list if not t.is_stop]