Updated for Keras 2.2.2 and Theano 1.0.2

This release has updated code to be compatible with new versions of Theano (1.0.2) and Keras (2.2.2).

Note that making these changes required us to remove the dropout layer from the model, which will lead to different results.

For the original code, use v1.0.

JAMIA Publication version

This is the code used to produce the results reported in:

Cocos, Fiks, and Masino, "Deep learning for pharmacovigilance: recurrent neural network architectures for labeling adverse drug reactions in Twitter posts." J Am Med Inform Assoc. 2017 Jul 1;24(4):813-821. doi: 10.1093/jamia/ocw180.