Utilizing Twitter Data for Identification of Need and Availability in Disaster Relief

Main Article Content

Dr. Kavita Asnani
Nikita Salkar
Mr.
Mugdha Khatavkar
Delstan D'Souza

Abstract

 


Social media in today's world has substantial impact in several domains. Disasters caused by natural hazards witness active communication on social media platforms. In such situations, the popular microblogging forums like Twitter have been resourceful in providing information related to relief operations. Tweets about disaster related to need and availability of emergency resources have been referred to as need-tweets and availability-tweets respectively. Automatic identification of such need-tweets and availability-tweets could be of good aid for timely action in relief operations. Our research exploits posts from Twitter to investigate the feasibility of using machine learning techniques of clustering and deep learning to assist in identification of need and availability of resources during crisis.
We performed experiments using Nepal Earthquake dataset. The results obtained by clustering algorithms yielded clusters that were specious with the occurrence of noise. We also performed experiments using recurrent neural network with long short term memory and compared the results with two baseline techniques using need and availability tweets. The proposed recurrent neural network achieved the precision of 0.772 and F-score of 0.600 for both need as well as availability tweets, which were higher than the baseline techniques.

 

Article Details

How to Cite
Asnani, D. K., Salkar, N. ., Manerikar, S. ., Khatavkar, M., & D’Souza, D. (2021). Utilizing Twitter Data for Identification of Need and Availability in Disaster Relief. INFOCOMP Journal of Computer Science, 20(2). Retrieved from https://infocomp.dcc.ufla.br/index.php/infocomp/article/view/1615
Section
Machine Learning and Computational Intelligence