The ongoing COVID-19 pandemic creates tremendous negative impacts on our health systems, businesses, and society. Therefore, monitoring the spread of the ongoing pandemic is an essential but challenging task. In this work, we first describe an annotated COVID-19 Twitter dataset that we provide to the research community to tackle this task. It allows identifying actual and potential COVID-19 patients as well as groups of potential COVID-19 positive contacts using social network sites. Second, we show that it is possible to detect COVID-positive users on the Twitter platform and estimate the officially reported COVID-19 infections in the U.S. per state by leveraging state-of-the-art Natural Language Processing (NLP) techniques. Moreover, our results reveal a high spatial and temporal correlation with the reported data, indicating a good fit for estimating the cumulated and time series' trend and a promising foundation for decision support and monitoring the pandemic.
Lowin, Maximilian and Kellner, Domenic, "From Twitter to COVID-19: Using NLP to Predict COVID-19 Infections" (2021). PACIS 2021 Proceedings. 147.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.