Identifying Adverse Drug Events from Health Social Media Using Distant Supervision

Adverse drug events (ADEs) have been recognized as a significant healthcare problem worldwide. Prior studies have shown that health social media can be used to generate medical hypotheses and identify adverse drug events. Most studies adopted supervised learning approach for ADE detection in health social media, which requires human annotated data and is not scalable to large datasets. In this study, we develop an information extraction framework to identify novel adverse drug events using distant supervised learning, which leverages existing knowledge of adverse drug events and requires no labeled text. Our proposed framework achieves competent performance in identifying adverse drug events without expensive human annotation.