Académique Documents
Professionnel Documents
Culture Documents
3. METHODOLOGY
This study follows the process of sentiment analysis shown in
Figure 1. Figure 2. Sample words with their neighboring words in
the vector space
3.1 Data Collection
Typhoon Yolanda related tweets for training and testing were 3.3 Data Annotation
gathered using the keywords ‘#YolandaPH’, ‘#BangonPH’, and From the total retrieved tweets, . a random sample of 3900 tweets
‘#BangonPilipinas’ starting from November 1, 2013 to January 1, were selected for manual labelling. The annotated data serve as the
2014. A total of 92,040 tweets were successfully obtained from the gold standard and ground truth of sentiments for the classifier.
given time period. Tweets were categorized in one of the three classes: positive,
negative, or neutral. Positive tweets include prayers, sympathy for
3.2 Data Preprocessing the victims, and encouragement; negative tweets are those
The dataset that will be used should be cleaned to avoid inclusion connoting stress, pity, and anger; while the neutral class contains
of noise and unnecessary information during training. Duplicate
news reports, announcements, observations by the people, plans, bidirectional Recurrent Neural Networks and their yielded
and tweets with no distinguishable sentiments. accuracy for fine-grained and binary classification.
.
[5] David CC, Ong JC, Legara EFT (2016) Tweeting [14] Philippines Typhoon Facts and Figures. (2015, August 05).
Supertyphoon Haiyan: Evolving Functions of Twitter during Retrieved November 21, 2017, from
and after a Disaster Event. PLoS ONE 11(3): e0150190. https://www.dec.org.uk/articles/facts-and-figures
[6] Dewey, Caitlin. 2015. “Twitter Might Make Tweets Longer. [15] Quick facts: What you need to know about Super Typhoon
Here’s Why They’re 140 Characters to Begin with.” Haiyan. (2013, November 16). Retrieved November 21, 2017,
Washington Post, October 1. from https://reliefweb.int/report/philippines/quick-facts-
https://www.washingtonpost.com/news/the-intersect/ what-you-need-know-about-super-typhoon-haiyan
wp/2015/10/01/twitter-might-make-tweets-longer-heres- [16] R. PH spends most time online and on social media – report.
why-theyre-140-characters-to-beginwith/ Retrieved November 22, 2017, from
[7] Dos Santos, C., & Gatti, M. Deep Convolutional Neural https://www.rappler.com/technology/features/159720-ph-
Networks for Sentiment Analysis of Short Texts. Brazilian spends-most-time-online-and-on-social-media-report
Research Lab. [17] Sentiment Analysis: How Does It Work? Why Should We Use
[8] Freeman, Mark. 2011. “Fire, Wind and Water: Social It? (2016, August 16). Retrieved November 24, 2017, from
Networks in Natural Disasters.” Journal of Cases on https://www.brandwatch.com/blog/understanding-sentiment-
Information Technology 13 (2): 69–79. analysis/
[9] Graves, A. (2012). Supervised Sequence Labelling with [18] Super Typhoon Yolanda is strongest storm ever to make
Recurrent Neural Networks (Ser. 385). Springer-Verlag Berlin landfall in recorded history. Retrieved November 21, 2017,
Heidelberg. DOI:10.1007/978-3-642-24797-2 from
[10] Montecillo, P. (2012). Philippines has 9.5M twitter users, http://www.gmanetwork.com/news/scitech/science/334571/s
ranks 10th. Retrieved from uper-typhoon-yolanda-is-strongest-storm-ever-to-make-
http://technology.inquirer.net/15189/philippines-has9-5m- landfall-in-recorded-history/story/
twitter-users-ranks-10th [19] Twitter by the Numbers: Stats, Demographics & Fun Facts.
[11] Mozetič I, Grčar M, Smailović J (2016) Multilingual Twitter Retrieved November 21, 2017, from
Sentiment Classification: The Role of Human Annotators. https://www.omnicoreagency.com/twitter-statistics/
[12] Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya [20] Waters, R. D., Burnett, E., Lamm, A., & Lucas, J. (2009).
Sutskever, and Ruslan Salakhutdinov. Dropout:A simple way Engaging stakeholders through social networking: How
to prevent neural networks from overfitting. JMLR, 2014. nonprofit organizations are using Facebook. Public Relations
Review
[13] Pang, B., & Lee, L. (2002). Opinion Mining and Sentiment
Analysis. Now Publisher Inc.