Alignment-Enhanced Transformer for Constraining NMT with Pre-Specified Translations
Abstract
We investigate the task of constraining NMT with pre-specified translations, which has practical significance for a number of research and industrial applications. Existing works impose pre-specified translations as lexical constraints during decoding, which are based on word alignments derived from target-to-source attention weights. However, multiple recent studies have found that word alignment derived from generic attention heads in the Transformer is unreliable. We address this problem by introducing a dedicated head in the multi-head Transformer architecture to capture external supervision signals. Results on five language pairs show that our method is highly effective in constraining NMT with pre-specified translations, consistently outperforming previous methods in translation quality.
Cite
Text
Song et al. "Alignment-Enhanced Transformer for Constraining NMT with Pre-Specified Translations." AAAI Conference on Artificial Intelligence, 2020. doi:10.1609/AAAI.V34I05.6418Markdown
[Song et al. "Alignment-Enhanced Transformer for Constraining NMT with Pre-Specified Translations." AAAI Conference on Artificial Intelligence, 2020.](https://mlanthology.org/aaai/2020/song2020aaai-alignment/) doi:10.1609/AAAI.V34I05.6418BibTeX
@inproceedings{song2020aaai-alignment,
title = {{Alignment-Enhanced Transformer for Constraining NMT with Pre-Specified Translations}},
author = {Song, Kai and Wang, Kun and Yu, Heng and Zhang, Yue and Huang, Zhongqiang and Luo, Weihua and Duan, Xiangyu and Zhang, Min},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2020},
pages = {8886-8893},
doi = {10.1609/AAAI.V34I05.6418},
url = {https://mlanthology.org/aaai/2020/song2020aaai-alignment/}
}