TY  - JOUR
T1  - Evaluating Urgency Levels of Emergency Alerts through Sentiment Analysis
AU  - Ju, Sang-Lim 
AU  - Kang, Hyunjoo 
AU  - Oh, Seung-Hee 
JO  - The Journal of Korean Institute of Communications and Information Sciences
PY  - 2024
DA  - 2024/1/1
DO  - 10.7840/kics.2024.49.2.226
KW  - Emergency Alerts
KW  - Natural Language Processing
KW  - Sentiment Analysis
KW  - Urgency Level evaluation
AB  - Emergency alerts are crucial means for promptly and accurately conveying disaster information in urgent situations. However, as the type and frequency of emergency alerts delivered to the public increase, there has been a rise in public fatigue and a decline in trust. To evaluate the appropriateness of the emergency alert service, this study proposes models that evaluate urgency levels for emergency alerts based on sentiment analysis. The models are developed using four representative natural language processing algorithms.
Furthermore, criteria and methodology for evaluating urgency levels are presented. In the experimental results, among the four algorithms, the urgency level evaluation model based on the Bidirectional Encoder Representations from Transformers algorithm showed the best learning performance with an accuracy of 98%.
And through the four learned algorithm-based urgency level evaluation models, the urgency of the emergency alerts issued in 2022 was classified and evaluated.