An Analysis of Political Sentiment and the Roleof Fake Accounts among Twitter ElectionMessages
The goal behind this research was to analyse the sentiment of Twitter users during the 2019 General Elections in India using Twitter messages from various users regarding their sentiment towards BJP and Congress, the two biggest national parties in India. The study also investigates the presence and role of fake accounts and the effect they have on the analysis. Using Convolutional Neural Networks, sentiment analysis was performed to analyse the opinion of online users and to classify them based on the sentiment of the tweets. The study also used a Bot detection framework, Botometer, to detect the presence of fake accounts among these users. The actual election results were used as corroborating evidence for this study. The results showed an equal percentage of attention as well as sentiment towards both the party and it was contradictory to the actual election results. The Botometer framework was successful in detecting bots, but the presence of these accounts did not have much of an elect on the analysis. The study has led to gaining insight into the relationship that has helped in understanding if twitter political sentiment can be used as a valid indicator of the actual sentiment of the voters and derive suggestions for further research.
Keywords: Sentiment Analysis, Bot Detection, Convolutional Neural, Network, Botometer, Twitter data
Sentiment140 dataset: http://help.sentiment140.com/for-students GloVe: https://nlp.stanford.edu/projects/glove/