tensor2tensor

Don't use this. (Lua)Torch is obsolete. Instead, if you want OpenNMT specifically for some reason then you should use either https://github.com/OpenNMT/OpenNMT-tf (TensorFlow) or https://github.com/OpenNMT/OpenNMT-py (PyTorch). I'm not sure what the best open source neural translation model is, but Google's Transformer models are very strong (and can be trained on TPUs, which is cheap(er) and fast): https://github.com/tensorflow/tensor2tensor/#translation

This is the kind of problem for which LSTM RNNs -- and more recently, fully-attention-based deep neural nets -- produce state-of-the-art results. I wonder if the author ever tried using, say, an AWD LSTM RNN[a] or a Transformer-like model[b] for this task. Using an RNN or an attention model for this would eliminate the need for feature engineering such as: feature_1 = 1 if x_t is capitalized and y_t equals "NAME"; 0 otherwise. This is one of seven carefully engineered feature functions listed in the article, and the author states that the seven are only a partial list. Moreover, using a modern RNN or attention model likely would produce better predictions, with much better generalization. [a] https://arxiv.org/abs/1708.02182 / https://github.com/salesforce/awd-lstm-lm [b] https://arxiv.org/abs/1706.03762 / https://github.