# Detecting Adverse Drug Events in Social Media: A Brief Literature Review

## TL;DR

This paper reviews 100 studies on using AI to detect drug side effects from social media posts, highlighting common methods, challenges like language bias, and future directions.

## Contribution

The first comprehensive review of 100 peer-reviewed studies on NLP for ADE detection in social media, organized by task types and methodological trends.

## Key findings

- Transformer-based models like BERT dominate ADE detection in social media.
- Twitter is the primary data source, with most studies focusing on English.
- Multilingual and code-mixed content remain underexplored challenges.

## Abstract

Adverse drug events (ADEs) remain a significant burden to public health and a persistent challenge for pharmacovigilance. The proliferation of patient-generated discourse on social media offers a complementary, real-time signal for ADE surveillance. This article provides a concise yet comprehensive review of recent natural language processing (NLP) research on identifying ADEs in social media text. We systematically reviewed 100 peer-reviewed studies (2017–2025) on NLP/AI for detecting or analysing ADEs in social media. Searches in Google Scholar targeted English-language journal and conference papers; patents and protocols were excluded. Of 130 records screened, 6 were protocols and 24 were excluded because the full text could not be located or the item was a conference abstract lacking methodological detail (i.e., no description of approaches or experiments), yielding a final sample of 100 studies. One reviewer performed screening, with full-text eligibility verified by a second. We extracted objectives, data sources/languages, preprocessing and annotation practices, datasets, model families, evaluation metrics, and stated limitations. Studies were grouped into five task categories–classification, extraction, normalization, corpus creation, and broader analytical work–with evidence tables summarizing contributions, toolchains, datasets, and performance. Recurrent challenges include noisy/imbalanced data, multilingual and code-mixed content, and variability in annotation standards. Twitter remains the primary data source: 60% of studies analyse Twitter alone and a further 18% combine Twitter with other platforms (78% in total). English overwhelmingly dominates; only about 5% of studies draw on non-English sources (e.g., French, Chinese, Arabic). Standard pre-processing–URL removal, tokenisation, and lowercasing–is near-universal. Transformer-based models predominate, with BERT and its biomedical or “tweet” variants (e.g., RoBERTa, BioBERT, BERTweet) used in more than 60% of approaches. Persistent obstacles include severe class imbalance and ambiguous or implicit drug-event expressions. Although shared tasks such as SMM4H provide widely used benchmarks, comprehensive annotation guidelines remain uncommon (12% of papers). Recent work increasingly incorporates multimodal inputs and integrates structured biomedical knowledge, yet gaps persist in multilingual coverage, temporal/longitudinal modelling, and real-world deployment. To our knowledge, this is the first review to synthesise findings from a corpus of 100 peer-reviewed studies on ADE detection in social media using NLP. By organising the literature by task type and tracing methodological trends and limitations, it provides practical guidance for researchers and practitioners. The review also outlines actionable directions for future work, including model explainability, support for low-resource languages, and closer collaboration with regulatory authorities to enable real-world deployment.

In this paper, we look at how natural language processing (NLP)–computer methods that read and learn from text–can spot adverse drug events (ADEs) in social media. We reviewed 100 research papers and summarised their aims, approaches, models, datasets, novelty, and limitations. We grouped the work into five practical buckets: classification (deciding if a post mentions an ADE), extraction (pulling out the drug and the event), normalisation (mapping everyday terms to medical ones), corpus creation (building datasets), and broader analysis. We compile the latest techniques and resources in a synthesis table and highlight consistent patterns: Twitter is the most studied platform; simple pre-processing such as removing or replacing URLs is common; and transformer models–especially BERT and its biomedical variants–are widely used. We also flag recurring hurdles: imbalanced data, the difficulty of annotating posts (especially when drug-event links are implied rather than explicit), few papers with clear annotation guidelines, heavy reliance on shared-task datasets like Social Media Mining for Health (SMM4H), and a strong bias toward English with limited work in languages such as French or Chinese. By bringing these findings together–summarising, comparing, and contrasting–we fill a gap in the field and offer a clear, lay-friendly map of current methods, datasets, and benchmarks.