Detecting Actionable Requests and Offers on Social Media During Crises Using LLMs

2025-04-22

AI Summary

During natural disasters, social media generates vast volumes of time-sensitive求助 and aid requests (e.g.,物资, personnel, or action-oriented), yet existing methods struggle to accurately identify and prioritize them under data-scarce conditions. This paper introduces a three-dimensional fine-grained classification framework for crisis information and proposes Query-Steered Few-shot (QSF) Learning—a novel paradigm integrating LLM-based prompt engineering, embedding-augmented retrieval, and hierarchical classification to jointly model operational feasibility and urgency. Evaluated on real-world crisis datasets, our approach significantly outperforms standard few-shot and zero-shot baselines, achieving high-precision request-supply matching and urgency-driven dynamic prioritization. The method delivers deployable, actionable intelligence for humanitarian response operations.

Technology Category

• Search and Optimization: Learning to Search
• Natural Language Processing: Information Extraction
• Data Mining & Knowledge Management: Intelligent Query Processing

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Natural disasters often result in a surge of social media activity, including requests for assistance, offers of help, sentiments, and general updates. To enable humanitarian organizations to respond more efficiently, we propose a fine-grained hierarchical taxonomy to systematically organize crisis-related information about requests and offers into three critical dimensions: supplies, emergency personnel, and actions. Leveraging the capabilities of Large Language Models (LLMs), we introduce Query-Specific Few-shot Learning (QSF Learning) that retrieves class-specific labeled examples from an embedding database to enhance the model's performance in detecting and classifying posts. Beyond classification, we assess the actionability of messages to prioritize posts requiring immediate attention. Extensive experiments demonstrate that our approach outperforms baseline prompting strategies, effectively identifying and prioritizing actionable requests and offers.

Problem

• Detect crisis-related requests and offers on social media
• Classify posts into supplies, personnel, and actions dimensions
• Prioritize actionable messages for immediate humanitarian response

Innovation

• Fine-grained hierarchical taxonomy for crisis information
• Query-Specific Few-shot Learning with LLMs
• Actionability assessment for prioritizing critical posts