Fix references

tensorflow/contrib/keras/python/keras/layers/recurrent.py

@@ -985,7 +985,7 @@ class LSTM(Recurrent):
 
   References:
       - [Long short-term
-        memory](http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf)
+        memory](http://www.bioinf.jku.at/publications/older/2604.pdf)
         (original 1997 paper)
       - [Supervised sequence labeling with recurrent neural
         networks](http://www.cs.toronto.edu/~graves/preprint.pdf)

tensorflow/contrib/rnn/python/ops/rnn_cell.py

@@ -79,7 +79,7 @@ class CoupledInputForgetGateLSTMCell(rnn_cell_impl.RNNCell):
 
   The default non-peephole implementation is based on:
 
-    http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf
+    http://www.bioinf.jku.at/publications/older/2604.pdf
 
   S. Hochreiter and J. Schmidhuber.
   "Long Short-Term Memory". Neural Computation, 9(8):1735-1780, 1997.

tensorflow/docs_src/tutorials/word2vec.md

@@ -351,7 +351,7 @@ to evaluate embeddings is to directly use them to predict syntactic and semantic
 relationships like `king is to queen as father is to ?`. This is called
 *analogical reasoning* and the task was introduced by
 [Mikolov and colleagues
-](http://msr-waypoint.com/en-us/um/people/gzweig/Pubs/NAACL2013Regularities.pdf).
+](http://www.anthology.aclweb.org/N/N13/N13-1090.pdf).
 Download the dataset for this task from
 [download.tensorflow.org](http://download.tensorflow.org/data/questions-words.txt).
 

tensorflow/python/ops/rnn_cell_impl.py

@@ -401,7 +401,7 @@ class LSTMCell(RNNCell):
 
   The default non-peephole implementation is based on:
 
-    http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf
+    http://www.bioinf.jku.at/publications/older/2604.pdf
 
   S. Hochreiter and J. Schmidhuber.
   "Long Short-Term Memory". Neural Computation, 9(8):1735-1780, 1997.