Sentiment analysis in reviews

• Main Author: Abdul Rasyid Sapuan
• Other Authors: Quek Hiok Chai
• Format: Final Year Project
• Language: English
• Published: 2017
• Subjects: DRNTU::Social sciences::Psychology::Affection and emotion; DRNTU::Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: http://hdl.handle.net/10356/70500
• Institution: Nanyang Technological University

In this study, we investigate the feasibility of using self-ratings and autonomic EEG activity in classifying sentiment elicited from individuals watching truthful and deceptive videos. The alpha and beta waves of EEG activity are isolated and tested to understand the role of subconscious and conscious brain activities in discriminating between truthful and deceptive videos. Furthermore, three cortical functions: attention, emotion and memory are studied to observe the significance of the different brain processes contributing to the act of forming sentiment. The study concludes with the outlook that conditioning impedes the performance of subjects while unconditioned, inexperienced subjects performed better. Also, memory access is a factor in conditioned subjects that consistently provides their best classification results while emotion contributes similarly to the performance of unconditioned subjects. We note that EEG activity performs satisfactorily with an average accuracy of 75.4% in discriminating between truth and deceit for the inexperienced, unconditioned subjects. This then increases to 78.8% when performing the same task on a separate set of stimuli, suggesting that subjects learn from an unconditioned exposure to truth and deception.