Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLM-Assisted Crisis Management: Building Advanced LLM Platforms for Effective Emergency Response and Public Collaboration

Hakan T. Otal, M. Abdullah Canbaz

TL;DR

This work tackles rapid crisis management by leveraging open-source LLMs to process 911 calls and social-media signals, aiming to reduce information overload and improve decision support for responders. It introduces two frameworks—one to enhance 911 dispatch efficiency and another to enable public crowdsourcing and guidance during large-scale emergencies—using fine-tuned LLAMA2 and Mistral models trained on disaster-related text datasets. Through a comprehensive methodology (model selection, dataset curation, preprocessing, prompt engineering, and SFT with parameter-efficient training), it demonstrates that LLAMA2-70B achieves top discriminative performance while LLAMA2-13B offers a balanced, practical trade-off for real-time emergency use, with robust handling of multilingual data. The study also addresses privacy, bias, and data/model poisoning concerns, underscoring the need for ethical, human-centered deployment and future work in governance and resilient crisis-communication systems that can scale to urban environments.

Abstract

Emergencies and critical incidents often unfold rapidly, necessitating a swift and effective response. In this research, we introduce a novel approach to identify and classify emergency situations from social media posts and direct emergency messages using an open source Large Language Model, LLAMA2. The goal is to harness the power of natural language processing and machine learning to assist public safety telecommunicators and huge crowds during countrywide emergencies. Our research focuses on developing a language model that can understand users describe their situation in the 911 call, enabling LLAMA2 to analyze the content and offer relevant instructions to the telecommunicator, while also creating workflows to notify government agencies with the caller's information when necessary. Another benefit this language model provides is its ability to assist people during a significant emergency incident when the 911 system is overwhelmed, by assisting the users with simple instructions and informing authorities with their location and emergency information.

LLM-Assisted Crisis Management: Building Advanced LLM Platforms for Effective Emergency Response and Public Collaboration

TL;DR

This work tackles rapid crisis management by leveraging open-source LLMs to process 911 calls and social-media signals, aiming to reduce information overload and improve decision support for responders. It introduces two frameworks—one to enhance 911 dispatch efficiency and another to enable public crowdsourcing and guidance during large-scale emergencies—using fine-tuned LLAMA2 and Mistral models trained on disaster-related text datasets. Through a comprehensive methodology (model selection, dataset curation, preprocessing, prompt engineering, and SFT with parameter-efficient training), it demonstrates that LLAMA2-70B achieves top discriminative performance while LLAMA2-13B offers a balanced, practical trade-off for real-time emergency use, with robust handling of multilingual data. The study also addresses privacy, bias, and data/model poisoning concerns, underscoring the need for ethical, human-centered deployment and future work in governance and resilient crisis-communication systems that can scale to urban environments.

Abstract

Emergencies and critical incidents often unfold rapidly, necessitating a swift and effective response. In this research, we introduce a novel approach to identify and classify emergency situations from social media posts and direct emergency messages using an open source Large Language Model, LLAMA2. The goal is to harness the power of natural language processing and machine learning to assist public safety telecommunicators and huge crowds during countrywide emergencies. Our research focuses on developing a language model that can understand users describe their situation in the 911 call, enabling LLAMA2 to analyze the content and offer relevant instructions to the telecommunicator, while also creating workflows to notify government agencies with the caller's information when necessary. Another benefit this language model provides is its ability to assist people during a significant emergency incident when the 911 system is overwhelmed, by assisting the users with simple instructions and informing authorities with their location and emergency information.
Paper Structure (16 sections, 5 figures, 1 table)

This paper contains 16 sections, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Enhancing 911 Dispatch Efficiency with LLM Integration
  3. LLMs for Collective Crisis Resolution: Public Collaboration
  4. Methodology
  5. Model Selection
  6. Datasets
  7. Additional Resources
  8. Preprocessing & Prompt Engineering
  9. Supervised Fine-Tuning
  10. Hardware Requirements
  11. Experimental Results
  12. Potential Problems
  13. Addressing Technical Challenges: Data and Model Poisoning in AI Systems
  14. Data Poisoning: Threats and Mitigation Strategies
  15. Model Poisoning: Identification and Countermeasures
  16. ...and 1 more sections

Figures (5)

  • Figure 1: Framework Design: Enhancing 911 Dispatch Efficiency with LLM Integration
  • Figure 2: LLM-Assisted Public Collaboration System Design
  • Figure 3: Class Distribution in the Emergency-Disaster Messages Dataset
  • Figure 4: Training losses of different models over epochs
  • Figure 5: Validation losses of different models over epochs