NOTE - A reference implementation of the monotonic attention mechanism is now built into TensorFlow in the tf.contrib.seq2seq module. See here for documentation.
This repository contains an example implementation of the monotonic alignment decoder proposed in "Online and Linear-Time Attention by Enforcing Monotonic Alignments", by Colin Raffel, Minh-Thang Luong, Peter J. Liu, Ron J. Weiss, and Douglas Eck (arXiv:1704.00784).
It also contains (in the benchmark folder) the synthetic benchmark used to compare the speed of the monotonic attention mechanism and standard softmax-based attention.