Article Reranking by Memory-enhanced Key Sentence Matching for Detecting Previously Fact-checked Claims.

Overview

MTM

This is the official repository of the paper:

Article Reranking by Memory-enhanced Key Sentence Matching for Detecting Previously Fact-checked Claims.

Qiang Sheng, Juan Cao, Xueyao Zhang, Xirong Li, and Lei Zhong.

Proceedings of the Joint Conference of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (ACL-IJCNLP 2021)

PDF / Poster / Code / Chinese Dataset / Chinese Blog 1 / Chinese Blog 2

Datasets

There are two experimental datasets, including the Twitter Dataset, and the firstly proposed Weibo Dataset. Note that you can download the Weibo Dataset only after an "Application to Use the Chinese Dataset for Detecting Previously Fact-Checked Claim" has been submitted.

Code

Key Requirements

python==3.6.10
torch==1.6.0
torchvision==0.7.0
transformers==3.2.0

Usage for Weibo Dataset

After you download the dataset (the way to access is described here), move the FN_11934_filtered.json and DN_27505_filtered.json into the path MTM/dataset/Weibo/raw:

mkdir MTM/dataset/Weibo/raw
mv FN_11934_filtered.json MTM/dataset/Weibo/raw
mv DN_27505_filtered.json MTM/dataset/Weibo/raw

Preparation

Tokenize

cd MTM/preprocess/tokenize
sh run_weibo.sh

ROT

cd MTM/preprocess/ROT

You can refer to the run_weibo.sh, which includes three steps:

  1. Prepare RougeBert's Training data:

    python prepare_for_rouge.py --dataset Weibo --pretrained_model bert-base-chinese
    
  2. Training:

    CUDA_VISIBLE_DEVICES=0 python main.py --debug False \
    --dataset Weibo --pretrained_model bert-base-chinese --save './ckpts/Weibo' \
    --rouge_bert_encoder_layers 1 --rouge_bert_regularize 0.01 \
    --fp16 True
    

    then you can get ckpts/Weibo/[EPOCH].pt.

  3. Vectorize the claims and articles (get embeddings):

    CUDA_VISIBLE_DEVICES=0 python get_embeddings.py \
    --dataset Weibo --pretrained_model bert-base-chinese \
    --rouge_bert_model_file './ckpts/Weibo/[EPOCH].pt' \
    --batch_size 1024 --embeddings_type static
    

PMB

cd MTM/preprocess/PMB
  1. Prepare the clustering data:

    mkdir data
    mkdir data/Weibo
    

    and you can get data/Weibo/clustering_training_data_[TS_SMALL] <[TS_LARGE].pkl after running calculate_init_thresholds.ipynb.

  2. Kmeans clustering. You can refer to the run_weibo.sh:

    python kmeans_clustering.py --dataset Weibo --pretrained_model bert-base-chinese --clustering_data_file 'data/Weibo/clustering_training_data_[TS_SMALL]
         
          <[TS_LARGE].pkl'
    
         

    then you can get data/Weibo/kmeans_cluster_centers.npy.

Besides, it is available to see some cases of key sentences selection in key_sentences_selection_cases_Weibo.ipynb.

Training and Inferring

cd MTM/model
mkdir data
mkdir data/Weibo

You can refer to the run_weibo.sh:

CUDA_VISIBLE_DEVICES=0 python main.py --debug False --save 'ckpts/Weibo' \
--dataset 'Weibo' --pretrained_model 'bert-base-chinese' \
--rouge_bert_model_file '../preprocess/ROT/ckpts/Weibo/[EPOCH].pt' \
--memory_init_file '../preprocess/PMB/data/Weibo/kmeans_cluster_centers.npy' \
--claim_sentence_distance_file './data/Weibo/claim_sentence_distance.pkl' \
--pattern_sentence_distance_init_file './data/Weibo/pattern_sentence_distance_init.pkl' \
--memory_updated_step 0.3 --lambdaQ 0.6 --lambdaP 0.4 \
--selected_sentences 3 \
--lr 5e-6 --epochs 10 --batch_size 32 \

then the results and ranking reports will be saved in ckpts/Weibo.

Usage for Twitter Dataset

The description of the dataset can be seen at here.

Preparation

Tokenize

cd MTM/preprocess/tokenize
sh run_twitter.sh

ROT

cd MTM/preprocess/ROT

You can refer to the run_twitter.sh, which includes three steps:

  1. Prepare RougeBert's Training data:

    python prepare_for_rouge.py --dataset Twitter --pretrained_model bert-base-uncased
    
  2. Training:

    CUDA_VISIBLE_DEVICES=0 python main.py --debug False \
    --dataset Twitter --pretrained_model bert-base-uncased --save './ckpts/Twitter' \
    --rouge_bert_encoder_layers 1 --rouge_bert_regularize 0.05 \
    --fp16 True
    

    then you can get ckpts/Twitter/[EPOCH].pt.

  3. Vectorize the claims and articles (get embeddings):

    CUDA_VISIBLE_DEVICES=0 python get_embeddings.py \
    --dataset Twitter --pretrained_model bert-base-uncased \
    --rouge_bert_model_file './ckpts/Twitter/[EPOCH].pt' \
    --batch_size 1024 --embeddings_type static
    

PMB

cd MTM/preprocess/PMB
  1. Prepare the clustering data:

    mkdir data
    mkdir data/Twitter
    

    and you can get data/Twitter/clustering_training_data_[TS_SMALL] <[TS_LARGE].pkl after running calculate_init_thresholds.ipynb.

  2. Kmeans clustering. You can refer to the run_twitter.sh:

    python kmeans_clustering.py --dataset Twitter --pretrained_model bert-base-uncased --clustering_data_file 'data/Twitter/clustering_training_data_[TS_SMALL]
         
          <[TS_LARGE].pkl'
    
         

    then you can get data/Twitter/kmeans_cluster_centers.npy.

Besides, it is available to see some cases of key sentences selection in key_sentences_selection_cases_Twitter.ipynb.

Training and Inferring

cd MTM/model
mkdir data
mkdir data/Twitter

You can refer to the run_twitter.sh:

CUDA_VISIBLE_DEVICES=0 python main.py --debug False --save 'ckpts/Twitter' \
--dataset 'Twitter' --pretrained_model 'bert-base-uncased' \
--rouge_bert_model_file '../preprocess/ROT/ckpts/Twitter/[EPOCH].pt' \
--memory_init_file '../preprocess/PMB/data/Twitter/kmeans_cluster_centers.npy' \
--claim_sentence_distance_file './data/Twitter/claim_sentence_distance.pkl' \
--pattern_sentence_distance_init_file './data/Twitter/pattern_sentence_distance_init.pkl' \
--memory_updated_step 0.3 --lambdaQ 0.6 --lambdaP 0.4 \
--selected_sentences 5 \
--lr 1e-4 --epochs 10 --batch_size 16 \

then the results and ranking reports will be saved in ckpts/Twitter.

Citation

@inproceedings{MTM,
  author    = {Qiang Sheng and
               Juan Cao and
               Xueyao Zhang and
               Xirong Li and
               Lei Zhong},
  title     = {Article Reranking by Memory-Enhanced Key Sentence Matching for Detecting
               Previously Fact-Checked Claims},
  booktitle = {Proceedings of the 59th Annual Meeting of the Association for Computational
               Linguistics and the 11th International Joint Conference on Natural
               Language Processing, {ACL/IJCNLP} 2021},
  pages     = {5468--5481},
  publisher = {Association for Computational Linguistics},
  year      = {2021},
  url       = {https://doi.org/10.18653/v1/2021.acl-long.425},
  doi       = {10.18653/v1/2021.acl-long.425},
}
Owner
ICTMCG
Multimedia Computing Group, Institute of Computing Technology, Chinese Academy of Sciences. Our official account on WeChat: ICTMCG.
ICTMCG
Notepy is a full-featured Notepad Python app

Notepy A full featured python text-editor Notable features Autocompletion for parenthesis and quote Auto identation Syntax highlighting Compile and ru

Mirko Rovere 11 Sep 28, 2022
RL agent to play μRTS with Stable-Baselines3

Gym-μRTS with Stable-Baselines3/PyTorch This repo contains an attempt to reproduce Gridnet PPO with invalid action masking algorithm to play μRTS usin

Oleksii Kachaiev 24 Nov 11, 2022
Block-wisely Supervised Neural Architecture Search with Knowledge Distillation (CVPR 2020)

DNA This repository provides the code of our paper: Blockwisely Supervised Neural Architecture Search with Knowledge Distillation. Illustration of DNA

Changlin Li 215 Dec 19, 2022
P-Tuning v2: Prompt Tuning Can Be Comparable to Finetuning Universally Across Scales and Tasks

P-tuning v2 P-Tuning v2: Prompt Tuning Can Be Comparable to Finetuning Universally Across Scales and Tasks An optimized prompt tuning strategy achievi

THUDM 540 Dec 30, 2022
[NIPS 2021] UOTA: Improving Self-supervised Learning with Automated Unsupervised Outlier Arbitration.

UOTA: Improving Self-supervised Learning with Automated Unsupervised Outlier Arbitration This repository is the official PyTorch implementation of UOT

6 Jun 29, 2022
Official implementation of "Open-set Label Noise Can Improve Robustness Against Inherent Label Noise" (NeurIPS 2021)

Open-set Label Noise Can Improve Robustness Against Inherent Label Noise NeurIPS 2021: This repository is the official implementation of ODNL. Require

Hongxin Wei 12 Dec 07, 2022
Official code for: A Probabilistic Hard Attention Model For Sequentially Observed Scenes

"A Probabilistic Hard Attention Model For Sequentially Observed Scenes" Authors: Samrudhdhi Rangrej, James Clark Accepted to: BMVC'21 A recurrent atte

5 Nov 19, 2022
🐸STT integration examples

🐸 STT 0.9.x Examples These are various examples on how to use or integrate 🐸 STT using our packages. It is a good way to just try out 🐸 STT before

coqui 92 Dec 19, 2022
利用Tensorflow实现基于CNN的中文短文本分类

Text Classification with CNN 使用卷积神经网络进行中文文本分类 CNN做句子分类的论文可以参看: Convolutional Neural Networks for Sentence Classification 还可以去读dennybritz大牛的博客:Implemen

Jeremiah 4 Nov 08, 2022
An official reimplementation of the method described in the INTERSPEECH 2021 paper - Speech Resynthesis from Discrete Disentangled Self-Supervised Representations.

Speech Resynthesis from Discrete Disentangled Self-Supervised Representations Implementation of the method described in the Speech Resynthesis from Di

Facebook Research 253 Jan 06, 2023
GANSketchingJittor - Implementation of Sketch Your Own GAN in Jittor

GANSketching in Jittor Implementation of (Sketch Your Own GAN) in Jittor(计图). Or

Bernard Tan 10 Jul 02, 2022
MAUS: A Dataset for Mental Workload Assessment Using Wearable Sensor - Baseline system

MAUS: A Dataset for Mental Workload Assessment Using Wearable Sensor - Baseline system Getting started To start working on this assignment, you should

2 Aug 06, 2022
Fuse radar and camera for detection

SAF-FCOS: Spatial Attention Fusion for Obstacle Detection using MmWave Radar and Vision Sensor This project hosts the code for implementing the SAF-FC

ChangShuo 18 Jan 01, 2023
Educational 2D SLAM implementation based on ICP and Pose Graph

slam-playground Educational 2D SLAM implementation based on ICP and Pose Graph How to use: Use keyboard arrow keys to navigate robot. Press 'r' to vie

Kirill 19 Dec 17, 2022
MQBench Quantization Aware Training with PyTorch

MQBench Quantization Aware Training with PyTorch I am using MQBench(Model Quantization Benchmark)(http://mqbench.tech/) to quantize the model for depl

Ling Zhang 29 Nov 18, 2022
Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering

Graph ConvNets in PyTorch October 15, 2017 Xavier Bresson http://www.ntu.edu.sg/home/xbresson https://github.com/xbresson https://twitter.com/xbresson

Xavier Bresson 287 Jan 04, 2023
A quick recipe to learn all about Transformers

Transformers have accelerated the development of new techniques and models for natural language processing (NLP) tasks.

DAIR.AI 772 Dec 31, 2022
Pytorch implementation of our method for high-resolution (e.g. 2048x1024) photorealistic video-to-video translation.

vid2vid Project | YouTube(short) | YouTube(full) | arXiv | Paper(full) Pytorch implementation for high-resolution (e.g., 2048x1024) photorealistic vid

NVIDIA Corporation 8.1k Jan 01, 2023
Storchastic is a PyTorch library for stochastic gradient estimation in Deep Learning

Storchastic is a PyTorch library for stochastic gradient estimation in Deep Learning

Emile van Krieken 140 Dec 30, 2022