diff --git a/doc/TRAINING.rst b/doc/TRAINING.rst
index 68007457..0463ba26 100644
--- a/doc/TRAINING.rst
+++ b/doc/TRAINING.rst
@@ -287,6 +287,8 @@ UTF-8 mode
 DeepSpeech includes a UTF-8 operating mode which can be useful to model languages with very large alphabets, such as Chinese Mandarin. For details on how it works and how to use it, see :ref:`decoder-docs`.
 
 
+.. _training-data-augmentation:
+
 Augmentation
 ^^^^^^^^^^^^