Detecting Reported Side Effects of COVID-19 Vaccines From Arabic Twitter (X) Data

Abstract

Vaccines might potentially cause side effects as any other drugs, which needs to be investigated and analyzed to identify the public safety concerns. The massive vaccination rollout against COVID-19 provoked discussion among people through social media platforms. Twitter (X), a popular social media platform, plays a significant role in disseminating information about COVID-19 vaccines and monitoring people’s reports regarding vaccination side effects. The aim of this study is to mine Twitter (X) to identify self-reported side effects related to COVID-19 vaccines in Arabic language, compare their distribution among six vaccine types, and construct Arabic lexicon of symptoms. We collected the tweets posts in Arabic language after the distribution of COVID-19 vaccines, then we developed a workflow for identifying self-report symptoms using biterm topic modeling (BTM) and support vector machine (SVM) to extract the symptoms then cluster them in groups based on their co-occurrence. A total of 51 symptoms were extracted from 65,387 tweets that were reported 148,324 times. We performed a more in-depth analysis to investigate the symptoms that tend to occur simultaneously. The results show that the symptoms that more likely to occur together may indicate to a particular connection. The findings suggested that the social media conversation can provide a comprehensive depiction of symptoms that may complement what identified in clinical studies.

Keywords

• Arabic language
• biterm topic modeling (BTM)
• COVID-19 vaccine
• machine learning
• NLP
• side effects