{"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/asynchronous-bidirectional-decoding-for","title":"Asynchronous Bidirectional Decoding for Neural Machine Translation","arxiv_id":"1801.05122","date":"2018-01-16","proceeding":null,"authors":["Xiangwen Zhang","Jinsong Su","Yue Qin","Yang Liu","Rongrong Ji","Hongji Wang"],"abstract":"The dominant neural machine translation (NMT) models apply unified\nattentional encoder-decoder neural networks for translation. Traditionally, the\nNMT decoders adopt recurrent neural networks (RNNs) to perform translation in a\nleft-toright manner, leaving the target-side contexts generated from right to\nleft unexploited during translation. In this paper, we equip the conventional\nattentional encoder-decoder NMT framework with a backward decoder, in order to\nexplore bidirectional decoding for NMT. Attending to the hidden state sequence\nproduced by the encoder, our backward decoder first learns to generate the\ntarget-side hidden state sequence from right to left. Then, the forward decoder\nperforms translation in the forward direction, while in each translation\nprediction timestep, it simultaneously applies two attention models to consider\nthe source-side and reverse target-side hidden states, respectively. With this\nnew architecture, our model is able to fully exploit source- and target-side\ncontexts to improve translation quality altogether. Experimental results on\nNIST Chinese-English and WMT English-German translation tasks demonstrate that\nour model achieves substantial improvements over the conventional NMT by 3.14\nand 1.38 BLEU points, respectively. The source code of this work can be\nobtained from https://github.com/DeepLearnXMU/ABDNMT.","url_abs":"http://arxiv.org/abs/1801.05122v2","url_pdf":"http://arxiv.org/pdf/1801.05122v2.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":"asynchronous-bidirectional-decoding-for","repo_url":"https://github.com/DeepLearnXMU/ABD-NMT","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null},{"paper_slug":"asynchronous-bidirectional-decoding-for","repo_url":"https://github.com/ManzoorElahi/organic-chemistry-reaction-prediction-using-NMT","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"decoder","task_name":"Decoder"},{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"nmt","task_name":"NMT"},{"task_slug":"translation","task_name":"Translation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1801.05122","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}