inference and make answer Memory Network Attentive reader (2016 Stanford > 2015 Deepmind)Bidirectional LSTM