A user perspective to quantify controversy on Twitter using Graph Neural Networks

Anonymous

TL;DR: Theorical approach to quantify controversy based on the probability of user to participate to a controversial topic.

Abstract:

This paper investigates the quantification of controversy in online discussions, focusing on
social media platforms, notably Twitter.
Emphasizing the prevalence of echo chambers, where users are exposed to opinions aligned with their own, we propose a novel approach leveraging Large Language Models (LLM) and Graph Neural Networks (GNN).
Our methodology integrates both structural and textual information in social networks to provide a nuanced understanding of controversy. Contributions include a theoretical model for quantifying controversy based on the expected probability of user participation in controversial topics.
We introduce an empirical estimation method using a GNN-based model. Unlike existing approaches focused on structural polarity, our model captures the rich textual content.
Empirical evaluations on Twitter topics demonstrate the effectiveness of our methodology, outperforming variant methods using only textual or structural information, as well as state-of-the-arts methods.
In conclusion, we introduce an innovative approach to controversy quantification, emphasizing user participation within social networks.

Paper Type: long

Research Area: NLP Applications

Contribution Types: NLP engineering experiment, Data analysis, Theory

Languages Studied: english