Sampling Informative Training Data for RNN Language Models

Code for training RNN language models used in ACL-SRW 2018 paper. N-gram language models were trained using SRILM.

(Code refactor & cleanup in progress.)