Claims in User Generated Multi-Modal Content: Implications for Fact Checking and Machine Learning

Speakers: 

Anushree Gupta is a PhD scholar at the Department of Liberal Arts (IIT Hyderabad). She is interested in exploring technologies as cultural artifacts, and their intersections with labour and place-making.

Tarunima Prabhakar is the research lead at Tattle Civic Tech. Her work looks at the implications of machine learning and prediction algorithms for democracies and development imperatives.

 

Abstract:
Snapchat, Tiktok and other short video platforms have put the spotlight on user generated and multi-modal content in driving online narratives. However, machine learning models, be it for claim detection or content moderation, have disproportionately focused on textual content in English. Datasets are also skewed towards sources where provenance is explicit such as ‘news’ websites, presidential speeches and television news. For majority of user-generated meme like content, provenance is unclear. In this talk, we will present our work from annotating claims in multi-modal, user generated content from an Indian social media platform and its implications for fact checking and automated approaches to misinformation response. This talk will draw from two papers: Check Mate: Prioritizing User Generated Multi-Media Content for Fact-Checking (ICWSM 2021), and A Contextual Examination of Factuality on Social Media (4S 2021).

 

The Zoom link for the webinar will be shared on our mailinglist: you can subscribe or simply send an email to scott.hale [at] oii.ox.ac.uk asking for the link for this talk.