Fetching the paper…

Adding Interpretable Attention to Neural Translation Models Improves Word Alignment · Around