An enhanced approach in lexicon-based sentiment analysis for social issues / Yasir Mehmood

• Main Author: Yasir, Mehmood
• Format: Thesis
• Published: 2018
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://studentsrepo.um.edu.my/10812/1/Yasir.pdf; http://studentsrepo.um.edu.my/10812/2/Yasir_Mehmood.pdf; http://studentsrepo.um.edu.my/10812/

Millions of users use social media to publically share their opinion and sentiment on different aspects of life. In decision making or option selection process, it is very important to know what others are thinking. In the last more than one and half decades, sentiment analysis has transformed into a very attractive research area due to the extended need to extract opinion and sentiment from the huge opinionated text data. In this context, mostly research has been conducted on the product and services. Nevertheless, sentiment analysis of social issues is different than sentiment analysis of product and services. Also minimal literature is available for the sentiment analysis of social issues. The purpose of this research is to enhance the lexicon-based sentiment analysis for social issues. Two datasets of custom data collected randomly from tweets over the issue of illegal immigration were used in the experiment of proposed technique. Same datasets manually labeled by the industry experts were analyzed by ten online sentiment analysis tools to check the effectiveness of proposed solution by using benchmark evaluation metrics precision, recall, F measure and accuracy. The proposed enhanced approach not only outperformed with overall accuracy of 81.4 and 82.3 as compared to the highest accuracy of 72.9 and 74.6 among ten online tools for both datasets respectively, but also classified each class of positive, negative and neutral with highest F measure values.