An attentional neural machine translation model is provided for the task of machine translation that, according to some embodiments, leverages the hierarchical structure of language to perform natural language processing without a priori annotation. Other features are also provided.