[ACL2017]Neural Word Segmentation with Rich Pretraining
来源:互联网 发布:mac 刻录机 编辑:程序博客网 时间:2024/05/20 18:15
中文分词的文章,非常有意思
之所以有意思是因为这篇文章将transition-based的方法引入到了neural network, 而且用来分词,当然transition-based用来做parsing的有很多文章,本文的关键问题是如何表示word, P, 和character. 并不是初始化而已,也是本文的创新点,就是pretraining word embedding, P embedding 和 character embedding.
预训练可以看如下图示:
我们以punc.为例,输入序列是5个窗口的字符,输出是判断中间字符之前是否出现标点符号,以此训练字符的embedding, 还有MLP之间的参数
另外需要给出解释的是:
silver data: 非人工标注的语料,可能是已有的工具标注的语料
gold data: 人工标注的语料
在paper的section 4的上面一段的解释:
4个预训练语料的比例是10/1/1/1, 不是句子的混合比例,是instances的混合比例。
punctuation corpus的例子是怎么取的呢?
标点符号之前和标点符号之后两个instance: 正好是正反两个instance
阅读全文
0 0
- [ACL2017]Neural Word Segmentation with Rich Pretraining
- [ACL2017]Adversarial Multi-Criteria Learning for Chinese Word Segmentation
- Vehicle Logo Recognition System Based on Convolutional Neural Networks With a Pretraining Strategy
- [ACL2017]Sequence-to-Dependency Neural Machine Translation
- Recurrent Neural Networks with Word Embeddings¶
- 翻译&学习|ACL2017|Modeling Situations in Neural Chat Bots
- Automating the Creation of Data-Rich Business Documents with Word 2007 and Visual Basic 2005
- Automating the Creation of Data-Rich Business Documents with Word 2007 and ASP.NET
- Perceptron for Word Segmentation
- Theano-Deep Learning Tutorials 笔记:Recurrent Neural Networks with Word Embeddings
- Associating Neural Word Embeddings With Deep Image Representations Using Fisher Vector源码解析
- Rich feature hierarchies for accurate object detection and semantic segmentation
- Rich feature hierachies for accurate object detection and semantic segmentation
- Rich feature hierarchies for accurate object detection and semantic segmentation
- Rich feature hierarchies for accurate object detection and semantic segmentation
- Rich feature hierarchies for accurate object detection and semantic segmentation
- Rich Internet Applications with Adobe Flex & Java
- Decoupled deep neural network for semi-supervised semantic segmentation 笔记
- spring,mybatis事务管理配置与@Transactional注解使用
- Thinkphp 3.2 显示html内容的时候 多了斜杠 导致样式和图片不能正常显示问题解决
- [BZOJ1011][HNOI2008]遥远的行星(近似)
- mysql优化
- Qt 学习之路 2(11):布局管理器
- [ACL2017]Neural Word Segmentation with Rich Pretraining
- try catch 用包装技术 抛出异常
- 一元三次方程的实数根
- 第三节-网页制作案例
- 萌新求助,搞了好久emmm……
- python里使用正则表达式搜索单词
- Remove Element
- JDBC之Statement,PreparedStatement,CallableStatement的区别
- 1. Struts2是什么?