This paper introduces a novel method for leveraging Open-Source Intelligence (OSINT) in disaster management by analyzing social media data, notably from microblogging sites. Utilizing text embedding, event detection, and knowledge graphs, our technique converts Twitter/X messages into actionable insights, filtering out irrelevant content. Addressing the challenge of large-scale social media data’s volume and hate speech, we employ an approach for sub-event detection by utilizing techniques like text embedding combined with clustering to categorize messages, simplifying data analysis, and enhancing pattern detection. Our approach surpasses traditional methods by automating data filtering and cleansing, reducing manual effort, ensuring data quality, and facilitating efficient resource utilization. It aids in identifying event-related clusters, tracking their development, and generating knowledge graphs for swift, accurate disaster management. This paper details our clustering, cleansing, and processing strategies, setting the stage for future discussions on their application in comprehensive situation report generation.