Twitter User Geolocation Using a Unified Text and Network Prediction Model

TL;DR

This paper introduces a novel label propagation method for Twitter user geolocation that combines network structure and text priors, achieving state-of-the-art accuracy by enhancing homophily and tractability.

Contribution

It presents a unified model that improves geolocation prediction by removing celebrity nodes and integrating text-based priors, advancing previous network-based approaches.

Findings

Achieves state-of-the-art geolocation accuracy on Twitter datasets.

Demonstrates the effectiveness of removing celebrity nodes.

Shows the benefit of incorporating text priors into network prediction.

Abstract

We propose a label propagation approach to geolocation prediction based on Modified Adsorption, with two enhancements:(1) the removal of "celebrity" nodes to increase location homophily and boost tractability, and (2) he incorporation of text-based geolocation priors for test users. Experiments over three Twitter benchmark datasets achieve state-of-the-art results, and demonstrate the effectiveness of the enhancements.