AffinityFinder: A System for Deriving Hidden Affinity Relationships on Twitter Utilizing Sentiment Analysis

Twitter is one of the largest and most popular social networking sites. While it has many interesting features, Twitter has no direct way to determine the relationship status between users. For example, unlike other social networks (e.g., Facebook), Twitter does not have any features for marking users as friends. In this paper, we present AffinityFinder, a system for automatically inferring potential friendship relationships (in terms of affinity) amongst Twitter users. The system collects and analyzes tweets to derive relationship scores that reflect affinity degrees amongst Twitter users. We implemented our tool using the TextBlob Python text processing library and the MongoDB database. Our evaluation shows that the system is able to derive potential friendship relationships with high accuracy. This system could provide useful data both to users and companies.