Hossain, T., Logan IV, R. L., Ugarte, A., Matsubara, Y., Singh, S., & Young, S. (2020). Detecting covid-19 misinformation on social media. https://openreview.net/forum?id=tRDw11nID9 

University of Michigan - Ann Arbor

Natural Language Processing (NLP) models were not particularly effective in detecting misinformation in Tweets, specifically the list of common Covid-19 misconceptions that Wikipedia provides

Detecting COVID-19 Misinformation on Social Media

- Studied 4,790 covid-19-related Tweets from March to April of 2020 
- Used Wikipedia to determine 86 of the common misconceptions regarding covid-19
- Used BERTScore to calculate the similarity of Tweets to common misconceptions
- 100 of the most similar Tweets were then studied in detail
- Categorized each of the 100 Twitter misconceptions as misinformative, informative, or irrelevant. Informative means the post was simply trying to correct the false information presented in other posts

Learn Before

Related