Natural Language Processing for Adverse Event Detection: a Case Study on COVID-19 Vaccines
Adverse events are among the main causes of hospitalization and death in the world. Pharmacovigilance plays therefore a crucial role in monitoring the drugs released in the market to minimize harm to the patients. In this talk, I will present our Adverse Event Detection system, which was ranked at the top of the Social Media Mining for Health (SMM4H) shared task leaderboard and won the IBM best short paper at the Workshop on Pharma and Healthcare Intelligence (W3PHIAI).
3 Key Takeaways:
*Adverse events are the cause of 5% to 8% of hospitalization.
*The FDA Adverse Event Reporting System has received over 16 million reports until 2018, 9 millions of which are classified as severe and 1.5 of which are related to the death of the patient
*Because the number of reports increased by 400% between 2007 and 2017, it is necessary to automate their processing and Natural Language Processing offers a way to do so.
Enrico Santus is a senior data scientist at Bayer. After his PhD at the Hong Kong Polytechnic University, Enrico joined the group of Regina Barzilay at CSAIL, MIT. His academic career includes affiliations with the King's College of London, the University of Pisa, the University of Stuttgart, the Nara Institute of Technology and Harvard. His work touches topics such as NLP in Oncology, Cardiology and Palliative Care. Enrico has also worked on Epidemiology, Fake News Detection, Sentiment Analysis and Lexical Semantics. As of today, Enrico has published over 50 papers, with over 637 citations. He collaborated to the creation of The Prayer (artist: Diemut Strebe), a mouth-shaped robot that pronounces original prayers, generated with Artificial Intelligence, exposed at the Centre Pompidou, in Paris. He was also involved in the creation of Safe Paths, the MIT tracing app.