Home
Softono

Ebook Lsi Cnn

Open source Jupyter Notebook
11
Stars
2
Forks
0
Issues
2
Watchers
7 years
Last Commit

 About Ebook Lsi Cnn

基于潜在语义索引(LSI)和卷积神经网络(CNN)的深度智能阅读模型

Platforms

Web Self-hosted

Languages

Jupyter Notebook

Links

Need Help Installing Ebook Lsi Cnn?

We provide expert installation service for this software. Our team will install, configure, and secure Ebook Lsi Cnn on your server. plans start at just $30.

Ebook Lsi Cnn

View on GitHub

智能阅读模型

概述

总体分为两部分:第一部分关键词匹配,第二部分精准匹配,最后选择置信度最高的正确回答。

第一部分可使用 TF-IDF、LSI 等传统方法。

第二部分可使用 基于深度学习的问答系统。

本文主要构建第二部分,第一部分可参考:https://github.com/littleredhat1997/doc-similarity

下载数据

  1. http://www.tipdm.org/jingsa/1253.jhtml
    train_data_complete.json、test_data_sample.json、submit_sample.txt => main/data 文件夹

  2. https://spaces.ac.cn/archives/4338
    me_train.json => generalization/data 文件夹

运行项目

$ ......
1. word2vec/step.ipynb -> 
  word2vec/word2vec.ipynb
2. main/data/data.ipynb -> 
  main/data/1xxxxxx.ipynb 
  main/data/2xxxxxx.ipynb 
  main/data/3xxxxxx.ipynb 
  main/data/4xxxxxx.ipynb 
  main/data/5xxxxxx.ipynb 
  -> main/evaluate.ipynb
3. test/data/newdata.ipynb -> 
  test/data/data.ipynb -> 
  test/predict.ipynb -> 
  test/evaluate.ipynb

$ tree
.
├── word2vec
│   ├── step.ipynb
│   └── word2vec.ipynb
├── main
│   ├── data
│   │   └── data.ipynb
│   ├── 1_FastText.ipynb
│   ├── 2_CNN1.ipynb
│   ├── 3_CNN2.ipynb
│   ├── 4_BiLSTM.ipynb
│   ├── 5_Attention.ipynb
│   └── evaluate.ipynb
└── test
    ├── data
    │   ├── data.ipynb
    │   └── newdata.ipynb
    ├── evaluate.ipynb
    └─── predict.ipynb

模型设计

  1. FastText

alt text

  1. CNN

alt text

  1. Bi-LSTM

alt text

  1. Attention

alt text