Update README.md

Changed the snippet to load the correct data split.

README.md

@@ -777,10 +777,11 @@ Each instance has scalar attributes (canvas) and sequence attributes (elements).
 To get a label for categorical values, use the `int2str` method:
 
 ```python
+data = dataset['train'] # obtain the train set
 key = "font"
-example = dataset[0]
+example = data[0] # obtain first sample in train set
 
-dataset.features[key].int2str(example[key])
+data.features[key].feature.int2str(example[key]) # obtain the text equivalent of the encoded values
 ```
 
 ### Data Fields