Modeling Ideological Salience and Framing in Polarized Online Groups with Graph Neural Networks and Structured Sparsity
Modeling Ideological Salience and Framing in Polarized Online Groups with Graph Neural Networks and Structured Sparsity
复制标题
利用图神经网络和结构化稀疏性对两极分化的在线群体中的意识形态显着性和框架进行建模
DOI:
10.18653/v1/2022.findings-naacl.41
复制
发表时间:
2021
期刊:
影响因子:
--
通讯作者:
Hinrich Schütze
中科院分区:
文献类型:
--
作者:
Valentin Hofmann;Xiaowen Dong;J. Pierrehumbert;Hinrich Schütze
The increasing polarization of online political discourse calls for computational tools that automatically detect and monitor ideological divides in social media. We introduce a minimally supervised method that leverages the network structure of online discussion forums, specifically Reddit, to detect polarized concepts. We model polarization along the dimensions of salience and framing, drawing upon insights from moral psychology. Our architecture combines graph neural networks with structured sparsity learning and results in representations for concepts and subreddits that capture temporal ideological dynamics such as right-wing and left-wing radicalization.
DOI:
10.1109/tnnls.2020.2978386
发表时间:
2021-01-01
影响因子:
10.4
作者:
Wu, Zonghan;Pan, Shirui;Yu, Philip S.
通讯作者:
Yu, Philip S.
DOI:
--
发表时间:
2018
期刊:
2018 Conference on Empirical Methods in Natural Language Processing (EMNLP
影响因子:
--
作者:
Field, Anjalie;Kliger, Doron;Wintner, Shuly;Pan, Jennifer;Jurafsky, Dan;Tsvetkov, Yulia
通讯作者:
Tsvetkov, Yulia