Sentiment Analysis on Candidates'' Images: the United States Presidential Election in 2016

• Main Authors: Horng-Shiunn Su, 蘇宏訓
• Other Authors: 曹修源
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/08277981316974758905

碩士 === 國立中興大學 === 行銷學系所 === 105 === The research focuses on the candidate analysis in the United States presidential election of 2016. For the research, the researcher establishes the Characteristics Keyword Lexicon on the basis of sentiment analysis, which is valid for evaluating candidate''s characteristics. The CKL includes five dimensions to evaluating candidate images that aids conducting in-depth study into the criticisms and advices made by the voters on candidates'' Facebook fan page. Following the study, the researcher applies sentiment analysis to calculate candidates'' text sentiment scores on the five dimensions out of voters'' comments. The scores enable the researcher to determine candidates'' performances on each dimension, and allow the researcher to conduct further analysis of the differences between the two candidates. Based on the differences between them, the researcher manages to develop a general model to evaluate candidate image. Applying the general model is of use to data collecting for conventional election poll, which provides the poll with better capability of predicting the precise states of the election, as well as more abundant information for the candidate''s campaign team to organize its strategies.