Detecting Disaster Trending Topics on Indonesian Tweets Using BNgram

Abstract

People on social media share information about natural disasters happening around them, such as the details about the situation and where the disasters are occurring. This information is valuable for understanding real-time events, but it can be challenging to use because social media posts often have an informal style with slang words. This research aimed to detect trending topics as a way to monitor and summarize disaster-related data originating from social media, especially Twitter, into valuable information. The research method used was BNgram. The selection of BNgram for detecting trending topics was based on its proven ability to recall topics well, as shown in previous research. Some stages in detection were data preprocessing, named entity recognition, calculation using DF-IDF, andhierarchical clustering. The resulting trending topics were compared with the topics obtained using the Document pivot method as the basic method. This research showed that BNgram performs better in detecting trending natural disaster-based topics compared to the Document pivot. Overall, BNgram had a higher topic recall score, and its keyword precision and keyword recall values were slightly better. In conclusion, recognizing the significance of social media in disaster-related information can increase disaster response strategies and situational awareness. Based on the comparison, BNgram was proven to be a more effective method for extracting important information from social media during natural disasters.