Researchers have introduced HUMAID-NER, a novel dataset designed for extracting structured information from disaster-related tweets. This dataset, built upon the HumAID benchmark, contains 60,000 English disaster tweets annotated with ten operationally relevant entity types, resulting in approximately 175,000 labeled entity spans. The annotation process involved a hybrid pipeline combining a spaCy transformer model, domain-specific patterns, and regular expressions. Additionally, a joint multitask learning framework utilizing a RoBERTa-large encoder was proposed to simultaneously perform named entity recognition and humanitarian event classification, achieving strong performance metrics. AI
IMPACT This dataset and methodology could improve the speed and accuracy of information extraction during humanitarian crises.
RANK_REASON The cluster describes a new academic dataset and associated methodology for named entity recognition and event classification in disaster tweets.
Read on Hugging Face Daily Papers →
- alphaXiv
- arXiv
- CatalyzeX
- DagsHub
- EntityRuler
- Gotit.pub
- Hugging Face
- Humaid Fayez Humaid Khalfan
- HUMAID-NER
- RoBERTa-large
- ScienceCast
- SpaCy
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
