按年被提到
20162016 · 4 次2026
大模型与语言模型›序列模型与注意力机制·arXiv:1609.08144 · cs.CL

Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation

谷歌神经机器翻译系统:弥合人类与机器翻译之间的差距
Yonghui Wu, Mike Schuster, Zhifeng Chen, Quoc V. Le, Mohammad Norouzi, Wolfgang Macherey, Maxim Krikun, Yuan Cao, Qin Gao, Klaus Macherey, Jeff Klingner, Apurva Shah 等 31 人 · 2016

8 层 LSTM 加注意力的谷歌神经机器翻译系统

开始阅读 ↗PDFalphaXiv 讨论
被推荐 · 9 人首次 2016-09-27 · 最近 2020-02-18
9不同的人
6活跃月份
9讨论串
5,315被引(OpenAlex)

论文介绍 GNMT:编码器和解码器各 8 层的深度 LSTM,带注意力和残差连接;注意力把解码器底层连到编码器顶层以提高并行度,推理时用低精度算术加速,并把词拆成 wordpiece 子词单元来处理罕见词,束搜索使用长度归一化和覆盖惩罚。在 WMT'14 英法和英德上取得有竞争力的结果,在简单句的人工并排评测中,比谷歌的短语式生产系统平均减少 60% 的翻译错误。

HN 为什么提它

评论者把它当作「谷歌用神经翻译取代统计翻译」的出处,并借它讨论 BLEU 评测、用强化学习直接优化 BLEU,以及开源复现与谷歌模型的差异。

方法难度 3 / 5约 50 分钟经得起时间
读之前最好知道
1LSTM 与序列到序列模型
2注意力机制
3BLEU 与机器翻译评测
4子词分词(wordpiece)

HN 书评

推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及
推荐ma2rten · 2017-01

“Google published a paper that explains exactly what they are doing and provides metrics as well [1,2].”

指出谷歌发表了论文,详细说明做法并给出评测指标
Why Deep Learning Cannot Be Applied to Natural Languages Easily原评论 ↗
批评anewhnaccount2 · 2018-03

“This result is on an independently set up evaluation framework, compared to Google's which used their own framework.”

Using AI to match human performance in translating news from Chinese to English原评论 ↗
引用aub3bhat · 2016-09

“Also there are metrics e.g. in the report released yesterday you can find BLEU scores on WMT 2014 tasks, and the clear conclusion is that Google is way ahead.”

引用报告里 WMT 2014 的 BLEU 分数,说明谷歌明显领先
Facebook, Amazon, Google, IBM and Microsoft Create Partnership on AI原评论 ↗
引用dennybritz · 2016-10

“A practical example would be Google's recently announced Machine Translation System. It uses RL to directly optimize BLEU scores on translated sentences.”

拿它当实例:用 RL 直接优化译文的 BLEU 分数
Learning Reinforcement Learning, with Code, Exercises, and Solutions原评论 ↗
引用srush · 2017-01

“The code implements a pretty close model to what Google Translate has published.”

说明这份代码实现的模型与谷歌翻译发表的论文相当接近
OpenNMT: Open-Source Neural Machine Translation原评论 ↗
引用circuithunter · 2017-02

“See the recent NY Times feature [1] and the arxiv paper [2].”

附上这篇论文,介绍谷歌神经机器翻译系统的实际应用
Oxford Deep NLP – An advanced course on natural language processing原评论 ↗
看全部 7 条 →
引用tommit · 2020-02

“starting with Google introducing Neural Machine Translation[1] to replace their statistical model”

把它当作谷歌引入神经机器翻译取代统计模型的起点
The new business of AI and how it’s different from traditional software原评论 ↗
被提交为 story · 2 次