Unveiling Hidden Patterns to Find Social Relevance

Twitter is both a useful social networking device and an incredible marketing tool. However, it is also a venue for dangerous stalkers and a sub-world of internet users that most people would intuitively avoid if seeing them in real life. It would improve Twitter's safety to have filters available which would allow users to select an audience for their status updates without being forced into changing their profiles to a private setting. The hypothetical filter, or model, studied in this paper was based on two particular attributes: activity correlations and vocabulary similarities between users and followers. If implemented, this model would restrict the availability of status updates to an automatically generated group of socially relevant followers. The result of this study shows that both of the attributes can be used to define social relevance, however, it was found that activity patterns have better predictive capabilities than correlating vocabulary usage between users and followers.