Augmenting Google Search in Ranking Twitter Users
Aleksey V. Panasyuk, Edmund Szu-Li Yu, Kishan G. Mehrotra · 2019
Content recommendation, and other applications, may require a ranking of popular accounts relevant to a specific geographic area. Ranking requires well-formed city communities that are representative of the geographic area. This task is complicated due to Twitter users' inherently noisy locations. In this paper, we propose a novel method for establishing city-level communities by exploiting ties to known local celebrity accounts discovered via automatic Google searches. A modified TF-IDF measure is used to rank celebrity users that are influential in one city community and not others. This approach allows for a targeted collection that is orders of magnitude smaller than comparable approaches, but still achieves better performance.