Update training data link to WMT16

doc/Quickstart.md

@@ -92,6 +92,6 @@ Now you have a model which you can use to predict on new data. We do this by run
 
 This will output predictions into `pred.txt`. The predictions are going to be quite terrible,
 as the demo dataset is small. Try running on some larger datasets! For example you can download
-millions of parallel sentences for [translation](http://www.statmt.org/wmt15/translation-task.html)
+millions of parallel sentences for [translation](http://www.statmt.org/wmt16/translation-task.html)
 or [summarization](https://github.com/harvardnlp/sent-summary).