Authors : Nur Atiqah Sia Abdullah, Nurul Iman Shaari and Abd Rasid Abd Rahman
Abstract: This study reviews sentiment analysis approaches specifically used for political research and social media data. The comparison is based on classifier, social media type, algorithm, data review and polarity classes. In this study, systematic literature review is used to explore the sentiment analysis approaches used in classifying social media data. The approaches include supervised machine learning, unsupervised learning, lexicon-based and hybridization approaches. The reviewed literatures involve data from social media such as Twitter, Email, Youtube and websites. All the approaches are evaluated and compared based on classifier, type of social media, data review and polarity classes. Based on the comparison, most of the researches use hybrid approach to classify the social media data. The algorithms in hybrid approaches are combination of lexicon-based and supervised machine learning. Most of the social media data used in these researches are extracted from Twitter. Lexicon of dictionary based and support vector machine are used for classifying political related tweets. There are also literatures involve Malay posts in social media. The past research uses social media, blog and Facebook as data. Then the sentiment analysis approaches are based on support vector machine and lexicon-based. The polarity classes involve only positive and negative or happy, unhappy and emotionless. As a conclusion, the hybrid approach of lexicon dictionary based and support vector machine is the best hybridization approach to classify the sentiment for the Malay political tweets.
Nur Atiqah Sia Abdullah, Nurul Iman Shaari and Abd Rasid Abd Rahman, 2017. Review on Sentiment Analysis Approaches for Social Media Data. Journal of Engineering and Applied Sciences, 12: 462-467.