Showing 1–1 of 1 results for author: Dhankar, A

UofA-Truth at Factify 2022 : Transformer And Transfer Learning Based Multi-Modal Fact-Checking

Authors: Abhishek Dhankar, Osmar R. Zaïane, Francois Bolduc

Abstract: Identifying fake news is a very difficult task, especially when considering the multiple modes of conveying information through text, image, video and/or audio. We attempted to tackle the problem of automated misinformation/disinformation detection in multi-modal news sources (including text and images) through our simple, yet effective, approach in the FACTIFY shared task at De-Factify@AAAI2022.… ▽ More Identifying fake news is a very difficult task, especially when considering the multiple modes of conveying information through text, image, video and/or audio. We attempted to tackle the problem of automated misinformation/disinformation detection in multi-modal news sources (including text and images) through our simple, yet effective, approach in the FACTIFY shared task at De-Factify@AAAI2022. Our model produced an F1-weighted score of 74.807%, which was the fourth best out of all the submissions. In this paper we will explain our approach to undertake the shared task. △ Less

Submitted 28 January, 2022; originally announced March 2022.