Authors: Marlene M. De Leon, Ma. Regina E. Estuar
Twitter is used by a significant number of people as a form of posting their current thoughts, feelings, and behaviors to their online social community. The Philippines is not left behind as evidence shows that Filipinos have ranked 8th in the world for having the most number of users .
With citizens using tweets during disasters and other calamities, the study recognizes the opportunity to determine sentiments and emotions that are embedded in the tweets. This study is an attempt in arriving at a bilingual sentiment model for determining emotions during disaster. The model is a two-step sentiment analysis process which classifies bilingual English-Filipino tweets according to subjectivity, then subjective tweets are classified according to sentiment polarity.
Latent constructs were identified using principal component analysis and associated to emotions. Disasters cause different kinds of emotions from people who were both directly affected or merely observers. Depending on the emotion, a tweet may be associated with people who are directly affected by disaster. Location metadata that may come with each tweet may be used to determine the location and magnitude of the disaster. Results show that 17 emotions emerged from the tweets analyzed in the study. 10 of those are considered negative emotions while the remaining 7 are positive.
Keywords: disaster, affect analysis, sentiment analysis, sub- jectivity analysis, emotions