Information Extraction From Tweets for Disaster Damage Assessment Using Rule-based NLP
Surbhi Soni, Swati Nanda Gupta, Rama Devi, K. K. Hazra · 2024
A natural disaster is an unanticipated incident that affects society. Disasters such as earthquakes, cyclones, floods, tsunamis, landslides, etc., are detrimental to the world. For a society to be prepared to save as many lives as possible the disasters must be monitored and evaluated. The government disaster management personnel generally do the damage assessment in post-disaster scenarios by using a damage assessment form that includes a set of information related to the disaster. They fill up the form manually by roaming the disaster-affected area and enquiring the victims and rescue workers. However, this process of manually collecting disaster information is time-consuming and laborious. To solve this problem, in this work, we propose an automatic approach to fill the damage assessment form using social media data, i.e. tweets that have been posted during or after a disaster event. To do that, Firstly, we identify the classes in which the tweets can be divided based on the damage assessment form; next, we preprocess and classify the collected tweets into different categories; and finally, we propose a rule-based information extraction strategy utilizing NLP to answer the questionnaires related to the damage assessment form. The evaluation of our approach reveals that it can outperform the baseline approach to answer the questionnaires.