Using Topic Modelling to make sense of typhoon-related tweets

The Philippines is a country that is often plagued by typhoons. In times of this disaster, many people turn to social media such as Twitter for information, making it useful. We take advantage of data present in tweets to get some insights, through discovered topics, on how they reflect the behavior of Filipinos during typhoons. Thus, we present a framework that uses Biterm Topic Modelling (BTM) to make sense of typhoon-related tweets. We focused on tweets collected from February 2013 to November 2014 of Twitter users from Metro Manila. Data preprocessing was applied to remove noisy and irrelevant data, like stop words and punctuations. We then conducted experiments using BTM for topic modelling and open coding for evaluating the results. Results revealed different Filipino behaviors during a typhoon such as determination to rise up after the typhoon, voicing out concerns, and using word play. Future work could experiment on selecting the appropriate number of words per topic model.