TSN4PI: Unified AI Framework Tracks Evolving Political Ideologies on Social Media
A new framework named TSN4PI, detailed in an arXiv preprint (2608.17987), seeks to monitor the evolution of political ideologies on social media platforms. It consists of two components: PIDN, which utilizes large language models for style transfer and unsupervised domain adaptation to eliminate non-political material and identify ideologies within noisy datasets, and PIPN, which employs temporal graph neural networks to predict future ideological changes. This combination facilitates the tracking of ideology's presence, intensity, and development. The initiative addresses issues such as limited data, high annotation costs, bias, and excessive irrelevant information. Additionally, two extensive datasets are being made available for noncommercial research to aid in the study of political polarization.
Key facts
- TSN4PI is a unified framework for tracing political ideology evolution on social media.
- The framework was announced on arXiv as 2608.17987v1.
- PIDN uses large language models with style transfer and unsupervised domain adaptation.
- PIDN filters irrelevant content from noisy cross-domain data.
- PIPN uses temporal graph neural networks to predict future ideological shifts.
- The framework enables analysis of ideology presence, intensity, and evolution.
- The study addresses data scarcity, non-political content, costly and bias-prone manual annotation, and difficulty in modeling future ideological inclinations.
- Two large-scale datasets are released for noncommercial research use.
Entities
—