diff --git a/test/datasets/test_iwslt2016.py b/test/datasets/test_iwslt2016.py
index 03b681f6f1..c5afd37664 100644
--- a/test/datasets/test_iwslt2016.py
+++ b/test/datasets/test_iwslt2016.py
@@ -45,7 +45,6 @@ def _generate_uncleaned_train():
         "<title",
         "<speaker",
         "<doc",
-        "</doc",
     ]
     for i in range(100):
         rand_string = " ".join(random.choice(string.ascii_letters) for i in range(10))
@@ -53,7 +52,8 @@ def _generate_uncleaned_train():
         # to ensure cleaning happens appropriately
         if random.random() < 0.1:
             open_tag = random.choice(xml_tags) + ">"
-            close_tag = "</" + open_tag[1:] + ">"
+            # Open tag already contains the closing >
+            close_tag = "</" + open_tag[1:]
             file_contents.append(open_tag + rand_string + close_tag)
         else:
             examples.append(rand_string + "\n")
diff --git a/test/datasets/test_iwslt2017.py b/test/datasets/test_iwslt2017.py
index 375ca4525d..6f462e8599 100644
--- a/test/datasets/test_iwslt2017.py
+++ b/test/datasets/test_iwslt2017.py
@@ -38,7 +38,6 @@ def _generate_uncleaned_train():
         "<title",
         "<speaker",
         "<doc",
-        "</doc",
     ]
     for i in range(100):
         rand_string = " ".join(random.choice(string.ascii_letters) for i in range(10))
@@ -46,7 +45,8 @@ def _generate_uncleaned_train():
         # to ensure cleaning happens appropriately
         if random.random() < 0.1:
             open_tag = random.choice(xml_tags) + ">"
-            close_tag = "</" + open_tag[1:] + ">"
+            # Open tag already contains the closing >
+            close_tag = "</" + open_tag[1:]
             file_contents.append(open_tag + rand_string + close_tag)
         else:
             examples.append(rand_string + "\n")