diff --git a/Language-Scaling/Mandarin/asr-finetune-conformer-ctc-nemo-zh.ipynb b/Language-Scaling/Mandarin/asr-finetune-conformer-ctc-nemo-zh.ipynb
index b45972c..b45303d 100644
--- a/Language-Scaling/Mandarin/asr-finetune-conformer-ctc-nemo-zh.ipynb
+++ b/Language-Scaling/Mandarin/asr-finetune-conformer-ctc-nemo-zh.ipynb
@@ -319,7 +319,7 @@
     "id": "fjHrIfDj_mMD"
    },
    "source": [
-    "#### 2.4.1 Change vocabulary"
+    "#### 2.4.1 Change vocabulary (optional)"
    ]
   },
   {
@@ -328,7 +328,9 @@
     "id": "4zBKAO36_mME"
    },
    "source": [
-    "Character based models don't need the tokenizer creation as only single characters are regarded as elements in the vocabulary. "
+    "Character based models don't need the tokenizer creation as only single characters are regarded as elements in the vocabulary.\n",
+    "\n",
+    "***HINTS*** if you don't want to change the vocabulary, just directly skip this section!"
    ]
   },
   {
@@ -646,7 +648,9 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "Let us freeze the encoder for easier initial convergence and faster training. On a smaller dataset when retraining the decoder, this is often a good idea."
+    "Let us freeze the encoder for easier initial convergence and faster training. On a smaller dataset when retraining the decoder, this is often a good idea.\n",
+    "\n",
+    "***HINTS*** If you want to finetune all the parameters of the model, you can skip this section!"
    ]
   },
   {