Cuet_Data_Wizards at #SMM4H-HeaRD 2026: Multilingual ADE Detection and Influenza Vaccine Effectiveness Estimation from Social Media

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Natural Language Processing · Depth: Advanced, quick

Summary

The Cuet_Data_Wizards team developed systems for two #SMM4H-HeaRD 2026 shared tasks, demonstrating robust transformer adaptation for health-related social media classification. For Task 1, multilingual adverse drug event (ADE) detection, they fine-tuned XLM-RoBERTa-large with weighted cross-entropy loss, augmenting low-resource settings with CADEC data and machine translation for zero-shot Persian. This achieved a macro F1 score of 0.582, surpassing the shared task average of 0.547. For Task 3, influenza vaccine effectiveness estimation, the team fine-tuned twitter-roberta-large to classify vaccination status and flu-test results from X posts. Their system achieved micro F1 scores of 0.845 for vaccination status and 0.883 for flu-test classification, with further improvements from focal loss and test-time augmentation.

Key takeaway

For NLP Engineers developing health-related social media analysis systems, consider fine-tuning robust transformer models like XLM-RoBERTa-large or twitter-roberta-large. You should prioritize data augmentation strategies, such as machine translation for low-resource languages, and experiment with techniques like weighted cross-entropy loss, focal loss, and test-time augmentation to enhance classification performance, especially for adverse drug event detection and vaccine effectiveness estimation.

Key insights

Robust transformer models effectively adapt for multilingual health-related social media classification, even with limited data.

Principles

• Weighted cross-entropy loss improves classification.
• Data augmentation aids low-resource language tasks.
• Post-evaluation techniques can boost performance.

Method

Fine-tune XLM-RoBERTa-large or twitter-roberta-large with weighted cross-entropy loss. Augment low-resource data via machine translation and external datasets. Apply focal loss, test-time augmentation, and head-tail truncation for refinement.

In practice

• Use XLM-RoBERTa-large for multilingual text.
• Apply twitter-roberta-large for X post analysis.
• Employ focal loss for class imbalance.

Topics

• Adverse Drug Event Detection
• Influenza Vaccine Effectiveness
• Social Media Mining
• Transformer Models
• XLM-RoBERTa
• twitter-roberta
• Multilingual NLP

Best for: Research Scientist, AI Scientist, NLP Engineer, Machine Learning Engineer

Related on AIssential

• mdok-style at SemEval-2026 Task 9: Finetuning LLMs for Multilingual Polarization Detection — Finetuning mid-sized LLMs with QLoRA and augmented data improves multilingual online polarization detection.
• Experimental Evaluation of Topic Modeling Methods for Categorizing Irregularities in Health-related news — LSA topic modeling effectively categorizes health news irregularities, enhancing public health audit efficiency.
• Detecting HIV-Related Stigma in Clinical Narratives Using Large Language Models — LLMs can effectively detect HIV-related stigma in clinical notes, with encoder models outperforming generative models in zero-shot settings.
• An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect — A hybrid framework combining transformer embeddings with classical classifiers enables effective rumour detection in low-resource Algerian dialect.
• From Logistic Regression to GPT-2: Building a Complete Spam Detection & Sentiment Analysis Pipeline — For imbalanced text classification, prioritize F1-score and Precision-Recall AUC over accuracy and ROC-AUC.
• Data Augmentation for Named Entity Recognition in Domain-Specific Scenarios in Portuguese — Data augmentation significantly improves NER performance in low-resource, domain-specific Portuguese datasets.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.