https://github.com/NonvolatileMemory/MXNET-SMN

Sequential Matching Network implemented by MXNET
https://github.com/NonvolatileMemory/MXNET-SMN

chatbot deep-learning mxnet

Last synced: 2 months ago
JSON representation

Sequential Matching Network implemented by MXNET

• Host: GitHub
• URL: https://github.com/NonvolatileMemory/MXNET-SMN
• Owner: NonvolatileMemory
• Created: 2017-12-07T12:15:37.000Z (about 7 years ago)
• Default Branch: master
• Last Pushed: 2019-03-11T07:09:29.000Z (almost 6 years ago)
• Last Synced: 2024-08-01T22:41:46.498Z (5 months ago)
• Topics: chatbot, deep-learning, mxnet
• Language: Python
• Size: 22.5 KB
• Stars: 18
• Watchers: 4
• Forks: 5
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• Awesome-MXNet - Sequential Matching Network(Gluon:star:)

README

        
# MXNET-SMN

Sequential Matching Network(ACL 2017) reimplemented by MXNET1.0-GPU//but you can always use cpu

用mxnet写的SMN，这篇文章发表在ACL 2017，使用MXNET1.0实现




# What's SMN

**SMN是啥？**

SoTA model of multi-turn retrieval-based conversation systems.

You can see the paper[SMN](http://www.aclweb.org/anthology/P/P17/P17-1046.pdf)

Don't forget reference it if you use.

Wu, Yu, et al. "Sequential Matching Network: A New Archtechture for Multi-turn Response Selection in Retrieval-based Chatbots." ACL. 2017.

SMN是目前最屌的多轮检索对话模型.

如果你是个大佬，请去读一下这篇论文[SMN](http://www.aclweb.org/anthology/P/P17/P17-1046.pdf)，这是我这一年中读过的为数不多的好文章.

它坦诚了SMN在某些方面的失败，对于baseline也作了完整归类，总结，所以ACL的质量还是非常高的。




# How to use?

**你能不能告诉我怎么用？**

1.get data from msra[douban corpus](https://1drv.ms/u/s!AtcxwlQuQjw1jF0bjeaKHEUNwitA)

2.get pre-trained word2vec.model use bash *python3.5 gen_w2v.py train.txt train_vec.model train_vec.vec*

3.get processed data use Process.py(also only support py3)

4.run the model

1.下载数据 [douban corpus](https://1drv.ms/u/s!AtcxwlQuQjw1jF0bjeaKHEUNwitA)

2.预训练Embedding矩阵*python3.5 gen_w2v.py train.txt train_vec.model train_vec.vec*

3.预处理数据使用Process.py

4.跑模型

额外赠礼：每个py文件的作用

|代码                              |作用                                |

|:------------------------------------:|------------------------------------|

|gen_w2v.py|用来生成预训练的词向量|

|Process.py|用来打包数据|

|model.py|用来训练模型|




# Params of Model

**模型的参数**

batch_size = 1000(with 1 titan xp)

embedding_size = 200

gru_layer = 1

max_turn = 10

max_len = 50

lr = 0.001

**警告：如果尝试修改参数，将会是一件非常痛苦的事情，因为我的代码高耦合**




# Why you use MXNET

**为什么是MXNET？**

**Fast!**

**快**! 不仅是开发快，运行快，训练也快。

> *当你使用MXNET的时候你会有一种闪电侠在中城奔跑的错觉。*

>                                           *-杜存宵*




# Other versions?

**我只会用该死的TensorFlow,怎么办？**

[There is Theano version.](https://github.com/MarkWuNLP/MultiTurnResponseSelection)

你可以选择学习theano或者mxnet，也可以自己实现一个。

*或者自杀*。




# Your code is not pythonic!

**你的代码就像shit一样**

Sorry to hear that.

你可以贡献你pythonic的代码，但是不好意思，我只用了三天的边角零头来开发这个模型。




# Can I use your code to do a chatbot

**我能用这个东西做个聊天机器人吗**

If you can use lucene then you can.

需要为每个query使用lucene检索出一个候选列表，之后进行排序




# How to tipping

**你真帅，我要为你生猴子**

You can sent your money to this [website](https://love.alipay.com/donate/index.htm)

你可以在支付宝E公益进行打赏。




## 你的屁话真多

不好意思，不好意思。

## TODO:

1.save model

2.add dropout

3.add evaluations

4.add inference