ClaimViz: Visual Analytics for Identifying and Verifying Factual Claims



Md Main Uddin Rony, Enamul Hoque, Naeemul Hassan

 External link (DOI) 

 View presentation:2020-10-29T17:00:00ZGMT-0600Change your timezone on the schedule page
2020-10-29T17:00:00Z

Exemplar figure — ClaimViz interface supports faceted exploration of a debate transcript based on discussion topics (A), speakers (B) and claims’ check-worthiness (E). The Minimap at the middle (D) visualizes how the potentially check-worthy sentences are distributed across different topics and claim types (e.g. numerical) and allows the user to quickly locate a check-worthy claim made by a speaker in the transcript view (F). The user can bookmark any potential claims for performing the verification task (C).

Fast forward

Direct link to video on YouTube: https://youtu.be/iZdakZN9U2M

Keywords

Human-centered computing-Visualization-Visualization design and evaluation methods

Abstract

Verifying a factual claim made by public figures, aka fact-checking, is a common task of the journalists in the newsrooms. One critical challenge that fact-checkers face is- they have to swift through a large amount of text to find claims that are check-worthy. While there exist some computational methods for automating the factchecking process, little research has been done on how a system should combine such techniques with visualizations to assist factcheckers. ClaimViz is a visual analytic system that integrates natural language processing and machine learning methods with interactive visualizations to facilitate the fact-checking process. The design of ClaimViz is based on analyzing the requirements of real factcheckers and our case studies demonstrate how the system can help users to effectively spot and verify claims.