An Analytical Framework for Analyzing Tweets for Disaster Management: Case Study of Turkey Earthquake 2023

Citizens utilize Twitter to communicate and post useful information amid disasters. The timely analysis of the first-hand information can be valuable for enhancing disaster management. This paper presents a new analytical framework for analyzing tweets (text and images) shared during disasters to (1) identify topics discussed during a disaster from tweet text using BERTopic along with their spatiotemporal analysis (2) to assess the severity of damage from images centered on a domain adaptation (DA) approach which uses EfficientNetB4 model and an iterative self-training algorithm. Using the Turkey Earthquake 2023 as a case study, a geo-located collection of tweets shared during the earthquake is analyzed, demonstrating the framework’s usefulness in analyzing tweets related to a real-world disaster event. The proposed framework can assist to identify critical areas that require immediate attention, assess the severity of the damage, and understand the evolving situation in real-time without the requirement for labeled data.