[ACL2017]Sequence-to-Dependency Neural Machine Translation
来源:互联网 发布:北京软件开发外包 编辑:程序博客网 时间:2024/05/29 18:10
conventional NMT的翻译过程:
(1)an encoder reads in source sentences of various lengths
(1)encoder 读取变长的原句子
(2) 将其转换为中间隐变量表示:
(3)计算a
目前提出的NMT方法中大都致力于解决的问题包括
(1)out-of-vocabulary words
(2)designing attention mechanism
(3)to more efficient parameter
(4)using source-side syntactic trees for better encoding and so on
不同于以上的关注点:本文的主要想法是使用目标语言的句法知识以提高翻译的质量
在NMT framework 中利用target syntactic structures存在以下难点:
(1)如何用RNN生成句法树
(2)如何同时生成目标词序列和构建目标语的句法结构
(3)如何有效利用句法结构提高翻译质量
本文给出的解决方案是:decoder时给出两个RNN:
A: word-RNN 用于生成目标语
B: Action-RNN用于构建句法树(这个就是生成依存树的过程,采用的arc-standard algorithm)
此时又出现了一个严重问题:词序列和action序列长度不一致
解决方案:当Action-RNN预测的标签为SH(shift)时,word-RNN才生成目标词
为了使得生成的词序列与action序列保持一致,这里引入了虚拟的词序列标签
意思是说当action-RNN生成的标签是SH时,此时word-RNN生成目标词y
为了有效利用action-RNN的结果,本文将公式(5)修改成了了公式(10)
公式(10)s
- [ACL2017]Sequence-to-Dependency Neural Machine Translation
- Neural Machine Translation and Sequence-to-sequence Models: A Tutorial
- Towards String-to-Tree Neural Machine Translation
- A novel approach to neural machine translation
- Improving Neural Machine Translation with Conditional Sequence Generative Adversarial Nets
- Adversarial Neural Machine Translation
- NEURAL MACHINE TRANSLATION BY JOINTLY LEARNING TO ALIGN AND TRANSLATE
- Effective Approaches to Attention-based Neural Machine Translation
- [EMNLP2015]Effective Approaches to Attention-based Neural Machine Translation
- Google's Neural Machine Translation System
- Neural Machine translation中的Attention机制
- Neural Machine Translation(NMT)技术概述
- Modeling Coverage for Neural Machine Translation
- Massive Exploration of Neural Machine Translation Architectures
- 【nlp论文阅读】Adversal Neural Machine Translation
- 神经网络机器翻译Neural Machine Translation: Attention Mechanism
- Neural Machine Translation论文阅读笔记
- 【论文阅读】Neural Machine Translation By Jointly Learning To Align and Translate
- James Whittaker系列——10分钟测试计划
- awk BEGIN END操作
- BZOJ 1787 Meet 紧急集合 (LCA)
- Windows 驱动机制-定时、DPC及延时
- Python 对象
- [ACL2017]Sequence-to-Dependency Neural Machine Translation
- 在有序的数组中获取目标值的地址——二分查询算法(java实现)
- 百度定位问题
- ApiDemos学习知识点之media-MediaPlayer_Audio&MediaPlayer_Video(11)
- dcloud 本地存储
- linux系统多线程编程
- DPC(延迟过程调用)的技术细节
- Android数据加密之Base64编码算法
- 深入浅出JMS(二)--ActiveMQ简单介绍以及安装