DepremKit: Human and ChatGPT Labeled Turkish Earthquake Dataset

Earthquakes are always a significant threat to people and countries, posing a natural disaster that demands immediate attention. The rapid identification of affected areas, determination of individuals and regions in need of assistance, and effective organization of aid teams play a crucial role in the development of decision support systems. Nowadays, social media users generate a vast amount of data rapidly. In the aftermath of the Kahramanmaras earthquake disaster in our country in 2023, affected individuals sought help through Twitter, and information about donations and needs was shared on this platform. This study presents a swift analysis of tweets posted after the earthquake disaster, aiming to contribute to providing efficient assistance to affected individuals. A dataset was created, labeled both manually with human tags and with the GPT-3.5 model, to address three distinct problem definitions and conduct experiments. The experimental results demonstrate the effective usability of the provided datasets and proposed models.