diff --git a/examples/csmsc/voc1/README.md b/examples/csmsc/voc1/README.md
index 252c2b920..b9ea2e906 100644
--- a/examples/csmsc/voc1/README.md
+++ b/examples/csmsc/voc1/README.md
@@ -3,6 +3,12 @@ This example contains code used to train a [parallel wavegan](http://arxiv.org/a
 ## Dataset
 ### Download and Extract
 Download CSMSC from it's [official website](https://test.data-baker.com/data/index/TNtts/) and extract it to `~/datasets`. Then the dataset is in the directory `~/datasets/BZNSYP`.
+The structure of the folder is listed below.
+```text
+datasets/BZNSYP
+└── Wave
+    └──XXXX.wav files
+```
 
 ### Get MFA Result and Extract
 We use [MFA](https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner) results to cut silence at the edge of audio.