Community Augmented Rapid-response to Events (CARE) Integrated Crisis Communication System

Salley, Christin; Xie, Jiajia; Bermeo, Mathias; Bishnoi, Maithly; Ehrenhalt, Amanda; Mohammadi, Neda; Taylor, John E · 2024 · ROSA P / Georgia. Dept. of Transporation. Office of Performance-Based Management and Research

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This research addresses the challenge of integrating social media data (Twitter/X) with community awareness applications (Waze) to enhance crisis communication and incident detection for transportation agencies. The study was motivated by the limitations of current Department of Transportation (DOT) systems, which rely heavily on Waze for real-time traffic data but lack the ability to ingest social media data for broader situational awareness. While Waze provides precise geolocation, it lacks contextual detail and sentiment; conversely, Twitter/X offers rich context but suffers from sparse geotagging and high noise levels. The project aims to develop a "Community Augmented Rapid-response to Events" (CARE) system that fuses these data streams to improve the accuracy, promptness, and equity of emergency response. The methodology employed a multi-pronged approach involving stakeholder engagement, software assessment, and machine learning development. First, the researchers conducted interviews with multiple state DOTs to understand current practices and barriers to social media integration. Second, they developed a weighted competency matrix to evaluate existing event detection software, identifying gaps in automated data fusion capabilities. Third, the team designed and tested two unique machine learning frameworks. The first framework utilizes transfer learning, topic modeling (Latent Dirichlet Allocation), and natural language processing to fuse historical social media data with Waze alerts, employing the Wells-DuBois Protocol to mitigate algorithmic bias. The second framework introduces an online confirmation-augmented probabilistic topic model that integrates variational lower bounds with a linear reward function to handle sparse, real-time data. The findings indicate that while RITIS is widely used for event detection, it does not currently ingest social media data, and existing commercial software lacks functionality for automating this fusion. The proposed machine learning models demonstrated significant efficacy. The transfer learning model achieved an average accuracy of approximately 0.83 in classifying historical data and Hurricane Ian events, successfully pairing Twitter/X posts with Waze alerts to enhance location and context understanding. The second model improved topic coherence and data labeling metrics, showcasing enhanced interpretability and precision in real-time analysis. Case studies on hurricanes and tropical storms validated the models' ability to provide heightened situational awareness and support adaptive decision-making. The significance of this work lies in establishing a foundation for equitable, real-time crisis detection that leverages the complementary strengths of social media and crowdsourced navigation data. By addressing biases and improving data fusion, the CARE system offers a pathway for DOTs to reduce response times, improve resource allocation, and better manage evolving crisis conditions. The study concludes that integrating these digital citizen communication channels into existing Advanced Traffic Management Systems can significantly bolster emergency preparedness and operational efficiency.

Key finding

Integrating Twitter/X data with Waze data through specialized machine learning frameworks significantly enhances the accuracy and promptness of crisis event detection and improves resource allocation for transportation authorities.

Methodology

mixed_methods

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• naturalistic crash near crash