"Actionable Help" in Crises: A Novel Dataset and Resource-Efficient Models for Identifying Request and Offer Social Media Posts

TL;DR

Crises generate massive, noisy social-media streams, making it hard to identify explicit help requests and offers in real time. The authors introduce CrisisHelpOffer, a 101k-tweet dataset built from an ensemble of generative LLM labels and human validation to focus on actionable content, and they develop crisis-domain mini-transformers distilled from a crisis-aware teacher to achieve high accuracy with far smaller size and dramatically faster inference. Across 13 crisis classification tasks, the mini models often surpass base encoder architectures and existing distilled baselines, delivering substantial speedups (up to 18.6x) while maintaining strong performance. A case study on the COVID-19 crisis reveals regional disparities in help-seeking and assistance-offering and demonstrates the practical utility of the dataset and compact models for resource-constrained crisis response. Overall, the work provides a scalable, deployable framework for real-time crisis informatics that improves actionable content detection and supports targeted relief efforts.

Abstract

During crises, social media serves as a crucial coordination tool, but the vast influx of posts--from "actionable" requests and offers to generic content like emotional support, behavioural guidance, or outdated information--complicates effective classification. Although generative LLMs (Large Language Models) can address this issue with few-shot classification, their high computational demands limit real-time crisis response. While fine-tuning encoder-only models (e.g., BERT) is a popular choice, these models still exhibit higher inference times in resource-constrained environments. Moreover, although distilled variants (e.g., DistilBERT) exist, they are not tailored for the crisis domain. To address these challenges, we make two key contributions. First, we present CrisisHelpOffer, a novel dataset of 101k tweets collaboratively labelled by generative LLMs and validated by humans, specifically designed to distinguish actionable content from noise. Second, we introduce the first crisis-specific mini models optimized for deployment in resource-constrained settings. Across 13 crisis classification tasks, our mini models surpass BERT (also outperform or match the performance of RoBERTa, MPNet, and BERTweet), offering higher accuracy with significantly smaller sizes and faster speeds. The Medium model is 47% smaller with 3.8% higher accuracy at 3.5x speed, the Small model is 68% smaller with a 1.8% accuracy gain at 7.7x speed, and the Tiny model, 83% smaller, matches BERT's accuracy at 18.6x speed. All models outperform existing distilled variants, setting new benchmarks. Finally, as a case study, we analyze social media posts from a global crisis to explore help-seeking and assistance-offering behaviours in selected developing and developed countries.

Figures (3)

• Figure 1: Loss curves ($y$-axis is in log scale) for S$_{\text{i}}^{\text{G}}$ (T: S$_\text{T}^\text{G}$, S: S$_\text{S}^\text{G}$ and M: S$_\text{M}^\text{G}$). During training, the teacher’s token embeddings were mean-pooled. The difference in training was in how students were updated: (i) mean-pooling of tokens and (ii) $<$CLS$>$ token. All variants started converging after 75% of the total steps.
• Figure 2: Distribution of request and offers tweets across resource types: Food, Money, Shelter, and Volunteer.
• Figure 3: R/O distributions across selected countries.