Universal Fake News Collection System using Debunking Tweets

TL;DR

This paper introduces a multilingual, rule-based system for collecting and identifying fake news by analyzing user debunking tweets, aiming to reduce manual effort and extend coverage beyond existing fact-checking sites.

Contribution

The proposed system is an unsupervised, rule-based framework that efficiently gathers potential fake news and clusters events, functioning in multiple languages with promising accuracy.

Findings

65% of identified event clusters are fake

System operates in English and Japanese

Potential for extension to other languages

Abstract

Large numbers of people use Social Networking Services (SNS) for easy access to various news, but they have more opportunities to obtain and share ``fake news'' carrying false information. Partially to combat fake news, several fact-checking sites such as Snopes and PolitiFact have been founded. Nevertheless, these sites rely on time-consuming and labor-intensive tasks. Moreover, their available languages are not extensive. To address these difficulties, we propose a new fake news collection system based on rule-based (unsupervised) frameworks that can be extended easily for various languages. The system collects news with high probability of being fake by debunking tweets by users and presents event clusters gathering higher attention. Our system currently functions in two languages: English and Japanese. It shows event clusters, 65\% of which are actually fake. In future studies, it will be applied to other languages and will be published with a large fake news dataset.