# SummaRuNNer

**Repository Path**: cfjzy/SummaRuNNer

## Basic Information

- **Project Name**: SummaRuNNer
- **Description**: The PyTorch Implementation of SummaRuNNer
- **Primary Language**: Unknown
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 2
- **Created**: 2020-11-28
- **Last Updated**: 2020-12-19

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

## The PyTorch Implementation Of SummaRuNNer

[![License](https://img.shields.io/badge/license-MIT-000000.svg)](https://opensource.org/licenses/MIT)

### Statement

+ **Not the official implementation! Just For Learning and communication!**

### Models

1. RNN_RNN
<div  align="center">
<img src="images/RNN_RNN.jpg" width = "350" height = "350" align=center />
</div>

2. CNN_RNN
<div  align="center">
<img src="images/CNN_RNN.png" width = "350" height = "260" align=center />
</div>

3. Hierarchical Attention Networks
<div  align="center">
<img src="images/Hiarchical_Attn.png" width = "350" height = "350" align=center />
</div>

### Setup

Requires [pipenv](https://docs.pipenv.org/). Use `pip install pipenv` if not installed.

```
pipenv install
pipenv shell
```

### Usage  

```shell
# train
python main.py -device 0 -batch_size 32 -model RNN_RNN -seed 1 -save_dir checkpoints/XXX.pt
# test
python main.py -device 0 -batch_size 1 -test -load_dir checkpoints/XXX.pt
# predict
python main.py -batch_size 1 -predict -filename x.txt -load_dir checkpoints/RNN_RNN_seed_1.pt
```
## pretrained models

1. RNN_RNN(`checkpoints/RNN_RNN_seed_1.pt`)
2. CNN_RNN(`checkpoints/CNN_RNN_seed_1.pt`)
2. AttnRNN(`checkpoints/AttnRNN_seed_1.pt`)

## Result

#### DailyMail(75 bytes)  

| model  | ROUGE-1   | ROUGE-2 | ROUGE-L |
| ------ | :-----:   | :----:  | :----:  |
|SummaRNNer(Nallapati)|26.2|10.8|14.4|
|RNN-RNN|26.0|11.5|13.8|
|CNN-RNN|25.8|11.3|13.8|
|Hierarchical Attn Net|26.0|11.4|13.8|

### Blog

+ [用PyTorch搭建抽取式摘要系统](http://mp.weixin.qq.com/s/9X77MPcQOQPwZaOVIVfo9Q)

### Download Data:  

+ 百度云:[https://pan.baidu.com/s/1LV3iuuH1NjxuAJd0iz14lA](https://pan.baidu.com/s/1LV3iuuH1NjxuAJd0iz14lA) 密码:`ivzl`

+ Google Driver:[data.tar.gz](https://drive.google.com/file/d/1JgsboIAs__r6XfCbkDWgmberXJw8FBWE/view?usp=sharing)

+ Source Data:[Neural Summarization by Extracting Sentences and Words](https://docs.google.com/uc?id=0B0Obe9L1qtsnSXZEd0JCenIyejg&export=download)

### Evaluation

+ [Tools](https://github.com/hpzhao/nlp-metrics)

### Acknowledge

+ Thanks for @[AlJohri](https://github.com/AlJohri)'s and @[500swapnil](https://github.com/500swapnil)'s contribution