PoliticAlly: Finding political friends on twitter

Abstract

Twitter is fast becoming the most popular platform for spread of information in general and pertaining to political views in particular. Like any other social networking site, Twitter is a medium to socialize with people and particularly in political space, it is pertinent to gauge the public opinion from time to time. One of the key requirements is to find people with similar political opinions in order to consolidate and find new political friends (online sepoys). In this work, our objective is to assist political parties in addressing this issue through a recommendation system which recommends new people to an already registered party worker based on participation of this worker on Twitter in the trending hashtags. Our proposed system is based on a new metric that we call relatedness between any two users on twitter. This metric is derived from analysis of two sources namely link and content of the tweets. The link source is characterized by participation of party worker on Twitter in form of @Mentions and @RT (retweet) that he/she posts in trending hashtags. The content source is characterized by the words appearing in the tweets posted by party worker combined with the hashtags present in them. We construct a directed weighted graph and use the proposed relatedness metric as weight of edges in the graph, whereas the users are considered as nodes of the graph. A well known WalkTrap community detection algorithm is then used to identify clusters of people with similar views based on which recommendation is done. A prototype system called PoliticAlly is developed which provides a simple web interface for party workers to use and get friend recommendations.