Analyzing sentiment bipolarity phenomenon in Social Networks using Python

Abstract

This study attempts to investigate the effect of sentiment interactions between users of a social network and diffusion of ideas. In addition, the study focused on identifying and analyzing bipolarity and controversy within a network in Twitter. Controversy is defined to be the phenomenon where users within a network are responding with opposing ideas to another’s user post. Data mining techniques implemented for constructing a network consisting of 62,799 single users and 64,693 connections between them. The network was representing all Tweets, Replies and Retweets posted in Twitter under hashtag “#Trump” for a given time in August 2020. From there, sentiment analysis was conducted for all posts by dividing users in two different polars, that is pro and against Trump candidacy, as well as dividing connections in controvert and assenting replies. A graphical visualization analysis followed, to draw relationships between influencing activity and polarity in our system. The Insights from graph analysis implied the strong relation between controversy in the network and diffusion of information. It appeared that almost 20% of network communications occurred after a controvert reply against another post was made. In addition, 3,442 homogenous communities with strong ties appeared in the network, each of them representing a group of users with the same sentiment towards Trump candidacy. It was also revealed that strong connected communities belonging to different polars are more probably to form communications between them showing controversy. Another interesting finding reported from graph analysis was the tendency of users polarized pro Trump candidacy to show homophony between their communications, while the other polar showed more controversy within users’ opinions.