Geolocated Event Detection using Graph Mining Approach on Real-Time Multimodal Data
Medha Mishra, Sadhana Tiwari, Ritesh Chandra, Sonali Agarwal, Sanjay Kumar Sonbhadra · 2024
Geolocated Event Detection is one of the most prevalent and novel research topic in social media event detection but presently very few researchers have addressed this problem. Thus, the aim of this research is to tackle the problem of natural disaster event detection with the help of real-time social media data in order to successfully incorporate the sentiments of the affected people so that the developed model must be able to train better with the aid of real-life scenarios. Currently, the focused natural disasters for this work are, earthquakes, floods and cyclones keeping the Indian Geography in mind, as Indian subcontinent is mostly affected by these calamities. Due to this, the real-time data required for training of the model was sufficiently available. Therefore, the prime aim of this work is to detect these events by mining real-time twitter data in the form of user tweets and images. Thus, these collected data is then linked to form graphs by finding patterns or relationships among them. This research is basically considering two different modalities of data i.e., text and image data in order to attain realistic results. This study will help in detecting natural disasters, such as earthquakes, floods etc. along with their location coordinates with the help of real-time user tweets and alerting the users in close proximity to the deducted location. This will benefit not only the people but also alert the concerned authorities so that help and other necessities can be provided well in time to the affected people and evacuations can be made before being too late, and thus the damages thus involved in any such event are kept to minimum.