{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/adding-interpretable-attention-to-neural","title":"Adding Interpretable Attention to Neural Translation Models Improves Word Alignment","arxiv_id":"1901.11359","date":"2019-01-31","proceeding":null,"authors":["Thomas Zenkel","Joern Wuebker","John DeNero"],"abstract":"Multi-layer models with multiple attention heads per layer provide superior\ntranslation quality compared to simpler and shallower models, but determining\nwhat source context is most relevant to each target word is more challenging as\na result. Therefore, deriving high-accuracy word alignments from the\nactivations of a state-of-the-art neural machine translation model is an open\nchallenge. We propose a simple model extension to the Transformer architecture\nthat makes use of its hidden representations and is restricted to attend solely\non encoder information to predict the next word. It can be trained on bilingual\ndata without word-alignment information. We further introduce a novel alignment\ninference procedure which applies stochastic gradient descent to directly\noptimize the attention activations towards a given target word. The resulting\nalignments dramatically outperform the naive approach to interpreting\nTransformer attention activations, and are comparable to Giza++ on two publicly\navailable data sets.","url_abs":"http://arxiv.org/abs/1901.11359v1","url_pdf":"http://arxiv.org/pdf/1901.11359v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"adding-interpretable-attention-to-neural","repo_url":"https://github.com/shuoyangd/meerkat","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"translation","task_name":"Translation"},{"task_slug":"word-alignment","task_name":"Word Alignment"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1901.11359","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}