diff --git a/speech/speech_recognition/conformer/pytorch/README.md b/speech/speech_recognition/conformer/pytorch/README.md
index 0dd3659b7cd30336ef47aeb99f3f009231b03635..35cc3b57d63451d566bef9d356f896a310ad471e 100755
--- a/speech/speech_recognition/conformer/pytorch/README.md
+++ b/speech/speech_recognition/conformer/pytorch/README.md
@@ -12,8 +12,8 @@ pip3 install -r requirements.txt
 
 ## Step 2: Training
 
-dataset is data_aishell.tgz and resource_aishell.tgz from wenet.  
-You could just run the whole script
+Dataset is data_aishell.tgz and resource_aishell.tgz from wenet.
+You could just run the whole script, which will download the dataset automatically.
 ```
 bash run.sh --stage -1 --stop-stage 6
 ```
@@ -50,4 +50,4 @@ bash run.sh --stage 6 --stop-stage 6
 
 
 ## Reference
-https://github.com/wenet-e2e/wenet
\ No newline at end of file
+https://github.com/wenet-e2e/wenet