[2010.05997] Look It Up: Bilingual and Monolingual Dictionaries Improve Neural Machine Translationopen searchopen navigation menucontact arXivsubscribe to arXiv mailings

Despite advances in neural machine translation (NMT) quality, rare words continue to be problematic. For humans, the solution to the rare-word problem has long been dictionaries, but dictionaries cannot be straightforwardly incorporated into NMT. In this paper, we describe a new method for "attaching" dictionary definitions to rare words so that the network can learn the best way to use them. We demonstrate improvements of up to 3.1 BLEU using bilingual dictionaries and up to 0.7 BLEU using monolingual source-language dictionaries.

1 mentions: @davidweichiang
Date: 2020/10/14 02:22

Referring Tweets

@davidweichiang My student Xing Jie Zhong's paper, appearing at WMT, on using dictionaries in NMT: t.co/r45BFdBYY7

Related Entries

Read more [2001.07715] TEASER: Fast and Certifiable Point Cloud Registrationcontact arXivarXiv Twitter
0 users, 1 mentions 2020/01/23 09:51
Read more [2001.00610] Representing Unordered Data Using Multiset Automata and Complex Numbersopen searchopen ...
0 users, 1 mentions 2020/06/01 15:51
Read more GitHub - IndicoDataSolutions/Enso: Enso: An Open Source Library for Benchmarking Embeddings + Transf...
0 users, 0 mentions 2020/09/12 21:52
Read more CMU Multilingual NLP (7): Machine Translation/Sequence-to-sequence Models - YouTube
0 users, 1 mentions 2020/09/18 09:52
Read more [2010.04674] Learning Context-Free Languages with Nondeterministic Stack RNNsopen searchopen navigat...
0 users, 1 mentions 2020/10/12 02:21