Discovering Informative Data from Catastrophe Tweets Using Sentiment Analysis
M. Priadarsini, J. Akilandeswari, S. P. Priyadharshini, R. Janani Devi, V. Abi Priya · 2023
Most of the social media platforms especially Twitter aims at connecting the users and helps in communicating their thoughts and feelings. During times of crisis, whether they involve immediate calamities or longer-term occurrences like floods and earthquakes, social media messages could be a vital origin of information. There is a probability that a tweet will contain both situational and non-situational information. Majority of previous efforts concentrated on recognizing tweets that contained a particular piece of information, such as situational, relevant, or actionable information. It can be difficult to distinguish between tweets that contain situational and non-situational information. Situational tweets include details that the concerned authorities may find useful, such as the count of impacted people in a certain area or contact information. Non-situational tweets can include sentimental information and personal viewpoints that may not be helpful to the authorities in question. Even after classifying the tweets, it is more important to give the information in short and efficient way. So that the tweets have to be summarized for quicker consumption without even losing a vital information. In data mining, clustering analysis is used to group similar data points together. In this paper, we are proposing Hierarsort framework that clusters and summarizes the data. The summarization is performed by pulling tweets out of different clusters.