https://www.jianshu.com/p/dfdfeaa7d01f
1 HMM模型
![](http://upload-images.jianshu.io/upload_images/2025501-803c6a2569a31cf3.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/799/format/webp)
image.png
马尔科夫过程:
![](http://upload-images.jianshu.io/upload_images/2025501-d86809c85e49b475.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/850/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-4276aeba657585e3.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/537/format/webp)
image.png
-
以天气判断为例:引出隐马尔科夫模型
image.png
image.png
以天气判断为例:由海藻信息推测天气
![](http://upload-images.jianshu.io/upload_images/2025501-b214943553046d80.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/425/format/webp)
image.png
于是我们可以将这种类型的过程建模为有一个隐藏的马尔科夫过程和一个与这个隐藏马尔科夫过程概率相关的并且可以观察到的状态集合。这就是本文重点介绍的隐马尔可夫模型。
隐马尔可夫模型(Hidden Markov Model) 是一种统计模型,用来描述一个含有隐含未知参数的马尔可夫过程。
-
马尔科夫假设
image.png
image.png
image.png
image.png
2 三个问题
![](http://upload-images.jianshu.io/upload_images/2025501-6697fff2e5b15c0c.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/858/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-4cd15d015ea9dbe3.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/945/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-20cf45167dd57592.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/882/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-256a146abdf0377b.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/888/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-d20c71807c29754d.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/905/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-ee3173c5237bad95.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/880/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-62f62aea630de2ad.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/867/format/webp)
image.png
- 源码架构
├── jieba
│ ├── analyse
│ │ ├── analyzer.py
│ │ ├── idf.txt
│ │ ├── init.py
│ │ ├── textrank.py
│ │ └── tfidf.py
│ ├── _compat.py
│ ├── dict.txt
│ ├── finalseg
│ │ ├── init.py
│ │ ├── prob_emit.p
│ │ ├── prob_emit.py
│ │ ├── prob_start.p
│ │ ├── prob_start.py
│ │ ├── prob_trans.p
│ │ └── prob_trans.py
│ ├── init.py
│ ├── main.py
│ └── posseg
│ ├── char_state_tab.p
│ ├── char_state_tab.py
│ ├── init.py
│ ├── prob_emit.p
│ ├── prob_emit.py
│ ├── prob_start.p
│ ├── prob_start.py
│ ├── prob_trans.p
│ ├── prob_trans.py
│ └── viterbi.py
├── LICENSE
├── MANIFEST.in
├── README.md
├── setup.py
└── test
![](http://upload-images.jianshu.io/upload_images/2025501-f82eaf46c348e768.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/707/format/webp)
image.png
3 Jieba源码分块解析
![](http://upload-images.jianshu.io/upload_images/2025501-aa6e2501c8939b09.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/904/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-a76962f2944eba42.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/858/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-287b286254d117e2.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/881/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-6e2f47cc24af67ba.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/952/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-9d64ad93b84a2040.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/935/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-8c4874413f8f4a61.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/902/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-a111f71ca290aa1a.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/924/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-fb5fc61c6d78296c.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/723/format/webp)
image.png
Jieba应用实践
![](http://upload-images.jianshu.io/upload_images/2025501-5edd22656f5b2c0d.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/799/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-c9f3051c63f9882f.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/402/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-45ebe2a0742e8917.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/398/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-96ac231be5887e09.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/461/format/webp)
image.png
![](http://upload-images.jianshu.io/upload_images/2025501-8cbf7c39502b8f63.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/324/format/webp)
image.png
小白学习 无关利益
感谢:
隐马尔可夫模型(HMM)攻略
HMM的(五个基本要素,三个假设,三个解决的问题)
HMM学习最佳范例七:前向-后向算法3
中文分词技术(中文分词原理)
鬼吹灯文本挖掘
https://blog.csdn.net/zhuzuwei/article/details/80775078
作者:林桉
链接:https://www.jianshu.com/p/dfdfeaa7d01f
来源:简书
简书著作权归作者所有,任何形式的转载都请联系作者获得授权并注明出处。