[WWW'23] KHAN: Knowledge-Aware Hierarchical Attention Networks for Accurate Political Stance Prediction

This repository provides an implementation of KHAN as described in the paper: KHAN: Knowledge-Aware Hierarchical Attention Networks for Accurate Political Stance Prediction by Yunyong Ko, Seongeun Ryu, Soeun Han, Youngseung Jeon, Jaehoon Kim, Sohyun Park, Kyungsik Han, Hanghang Tong, and Sang-Wook Kim, In Proceedings of the ACM Web Conference (WWW) 2023.

The overview of KHAN

• Datasets
  • To reflect the different political knowledge of each entity, we build two political knowledge graphs, KG-lib and KG-con. Also, for extensive evaluation, we construct a large-scale political news datatset, AllSides-L, much larger (48X) than the existing largest political news article dataset.
• Algorithm
  • We propose a novel approach to accurate political stance prediction (KHAN), employing (1) hierarchical attention networks (HAN) and (2) knowledge encoding (KE) to effectively capture both explicit and implicit factors of a news article.
• Evaluation
  • Via extensive experiments, we demonstrate that (1) (accuracy) KHAN consistently achieves higher accuracies than all competing methods (up to 5.92% higher than the state-of-the-art method), (2) (efficiency) KHAN converges within comparable training time/epochs, and (3) (effectiveness) each of the main components of KHAN is effective in political stance prediction.

Datasets

1. News articles datasets (SemEval, AllSides-S, AllSides-L)

Dataset	# of articles	Class distribution
SemEval	645	407 / 238
AllSides-S	14.7k	6.6k / 4.6k / 3.5k
AllSides-L	719.2k	112.4k / 202.9k / 99.6k / 62.6k / 241.5k

2. Knowledge Graphs (YAGO, KG-conservative, KG-liberal)

KG dataset	# of source poses	# of entities	# of raltions
YAGO	-	123,182	1,179,040
KG-lib	219,915	5,581	29,967
KG-con	276,156	6,316	33,207

3. Pre-trained KG embeddings (common, conservative, liberal)

Dependencies

Our code runs on the Intel i7-9700k CPU with 64GB memory and NVIDIA RTX 2080 Ti GPU with 12GB, with the following packages installed:

python 3.8.10
torch 1.11.0
torchtext 0.12.0
pandas
numpy
argparse
sklearn

How to run

python3 main.py \
  --gpu_index=0 \
  --batch_size=16 \
  --num_epochs=50 \
  --learning_rate=0.001 \
  --max_sentence=20 \
  --embed_size=256 \
  --dropout=0.3 \
  --num_layer=1 \
  --num_head=4 \
  --d_hid=128 \
  --dataset=SEMEVAL \
  --alpha=0.6 \
  --beta=0.2

Citation

Please cite our paper if you have used the code in your work. You can use the following BibTex citation:

@inproceedings{ko2023khan,
  title={KHAN: Knowledge-Aware Hierarchical Attention Networks for Accurate Political Stance Prediction},
  author={Ko, Yunyong and Ryu, Seongeun and Han, Soeun and Jeon,Youngseung and Kim, Jaehoon and Park, Sohyun and Han, Kyungsik Tong, Hanghang and Kim., Sang-Wook},
  booktitle={Proceedings of the ACM Web Conference (WWW) 2023},
  pages={1572--1583},
  year={2023},
  isbn = {9781450394161},
  publisher = {Association for Computing Machinery (ACM)},
  doi = {10.1145/3543507.3583300},
  location = {Austin, TX, USA},
}