Seminar: "Misinformed by Visualisations: Countering with Explainable Fact-Checking over Real-World Chart Images"
Abstract
Data visualizations (e.g. bar charts, pie charts, line graphs) are common in real-world data sources and are frequently used to summarize key information in a visual form. However, they are also misused to spread misinformation, with the aim to convince their audience towards certain agendas. Whilst fact verification has attracted substantial interest in the natural language processing community, verifying misinforming statements against data visualizations has so far been overlooked. This requires extracting information from data visualizations that use tightly integrated text and visual elements to represent information. In this work, we introduce ChartCheck, the first dataset for explainable fact-checking against data visualizations. We propose two baseline architectures to the community: vision-language models and chart-to-table architectures. Finally, we study reasoning types and visual attributes that pose a challenge to these models.
Invited Speaker: Mubashara Akhtar (King’s College London, UK)
Bio: Mubashara Akhtar is a final-year PhD student at King’s College London and a student researcher at Google DeepMind. Previously, she interned at Google DeepMind in Zurich and was a visiting student at the Cambridge NLP Group. She is also part of the organisation committee of the FEVER workshop and co-leading the responsible AI committee of Croissant, a MLCommons project on ML dataset documentation. Her research is on knowledge-grounded NLP, focusing on reasoning over text, tables, and images as modalities for automated fact-checking and question-answering. Her research interests further include understanding large language models’ reasoning capabilities, focusing on numeracy and factuality.
