GET
所属分类:图神经网络
开发工具:Python
文件大小:105683KB
下载次数:0
上传日期:2022-07-04 12:57:15
上 传 者:
sh-1993
说明: 【WWW 2022】“基于图神经网络的证据感知假新闻检测”源代码
([WWW 2022] The source code of "Evidence-aware Fake News Detection with Graph Neural Networks")
文件列表:
.DS_Store (6148, 2022-04-21)
Evaluation (0, 2022-04-21)
Evaluation\.DS_Store (6148, 2022-04-21)
Evaluation\mzEvaluator.py (1734, 2022-04-21)
Fitting (0, 2022-04-21)
Fitting\.DS_Store (6148, 2022-04-21)
Fitting\FittingFC (0, 2022-04-21)
Fitting\FittingFC\.DS_Store (6148, 2022-04-21)
Fitting\FittingFC\char_man_fitter_query_repr1.py (34859, 2022-04-21)
Fitting\FittingFC\declare_fitter.py (23182, 2022-04-21)
Fitting\FittingFC\multi_level_attention_composite_fitter.py (21917, 2022-04-21)
Fitting\densebaseline_fit.py (2832, 2022-04-21)
LICENSE (1067, 2022-04-21)
MasterFC (0, 2022-04-21)
MasterFC\master_get.py (13319, 2022-04-21)
Models (0, 2022-04-21)
Models\BiDAF (0, 2022-04-21)
Models\BiDAF\__pycache__ (0, 2022-04-21)
Models\BiDAF\__pycache__\wrapper.cpython-36.pyc (12129, 2022-04-21)
Models\BiDAF\bidaf_model.py (7676, 2022-04-21)
Models\BiDAF\wrapper.py (14970, 2022-04-21)
Models\FCWithEvidences (0, 2022-04-21)
Models\FCWithEvidences\DeClare (0, 2022-04-21)
Models\FCWithEvidences\DeClare\__pycache__ (0, 2022-04-21)
Models\FCWithEvidences\DeClare\__pycache__\pack.cpython-36.pyc (2802, 2022-04-21)
Models\FCWithEvidences\DeClare\pack.py (2799, 2022-04-21)
Models\FCWithEvidences\__pycache__ (0, 2022-04-21)
Models\FCWithEvidences\__pycache__\basic_fc_model.cpython-36.pyc (4072, 2022-04-21)
Models\FCWithEvidences\__pycache__\graph_based_semantic_structure.cpython-36.pyc (9772, 2022-04-21)
Models\FCWithEvidences\__pycache__\hierachical_multihead_attention.cpython-36.pyc (9857, 2022-04-21)
Models\FCWithEvidences\basic_fc_model.py (5801, 2022-04-21)
Models\FCWithEvidences\graph_based_semantic_structure.py (13853, 2022-04-21)
Models\__pycache__ (0, 2022-04-21)
Models\__pycache__\base_model.cpython-36.pyc (6160, 2022-04-21)
Models\base_model.py (7419, 2022-04-21)
__init__.py (23, 2022-04-21)
formatted_data (0, 2022-04-21)
formatted_data\.DS_Store (6148, 2022-04-21)
... ...
# GET
Source code and datasets for the paper "Evidence-aware Fake News Detection with Graph Neural Networks".
## Requirements
We use Pytorch 1.9.1 and python 3.6. Other requirements are in requirements.txt.
```
pip install -r requirements.txt
```
## Data
We utilize two widely used datasets.
* Snopes: http://resources.mpi-inf.mpg.de/impact/dl_cred_analysis/Snopes.zip
* PolitiFact: http://resources.mpi-inf.mpg.de/impact/dl_cred_analysis/PolitiFact.zip
## Usage
You can run the commands below to train and test our model on Snopes Dataset.
```
python MasterFC/master_get.py --dataset="Snopes" \
--cuda=1 \
--fixed_length_left=30 \
--fixed_length_right=100 \
--log="logs/get" \
--loss_type="cross_entropy" \
--batch_size=32 \
--num_folds=5 \
--use_claim_source=0 \
--use_article_source=1 \
--path="formatted_data/declare/" \
--hidden_size=300 \
--epochs=100 \
--num_att_heads_for_words=5 \
--num_att_heads_for_evds=2 \
--gnn_window_size=3 \
--lr=0.0001 \
--gnn_dropout=0.2 \
--seed=123756 \
--gsl_rate=0.6
```
You can also simply run the bash script.
```
sh run_snopes.sh
```
or
```
sh run_politifact.sh (on the PolitiFact dataset)
```
## Citation
Please cite our paper if you use the code:
```
@misc{xu2022evidenceaware,
title={Evidence-aware Fake News Detection with Graph Neural Networks},
author={Weizhi Xu and Junfei Wu and Qiang Liu and Shu Wu and Liang Wang},
year={2022},
eprint={2201.06885},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
```
## Acknowledgement
The general structure of our codes inherites from the open-source codes of [MAC](https://github.com/nguyenvo09/EACL2021), we thank them for their great contribution to the research community of fake news detection.
近期下载者:
相关文件:
收藏者: