A Hierarchical Location Prediction Neural Network for Twitter User Geolocation
This addresses the need for accurate location estimation in online services, offering a domain-specific improvement for social media geolocation.
The paper tackles the problem of Twitter user geolocation by proposing a hierarchical neural network that predicts home country first to guide city-level prediction, achieving state-of-the-art results on three benchmarks with improved accuracy and reduced mean error distance.
Accurate estimation of user location is important for many online services. Previous neural network based methods largely ignore the hierarchical structure among locations. In this paper, we propose a hierarchical location prediction neural network for Twitter user geolocation. Our model first predicts the home country for a user, then uses the country result to guide the city-level prediction. In addition, we employ a character-aware word embedding layer to overcome the noisy information in tweets. With the feature fusion layer, our model can accommodate various feature combinations and achieves state-of-the-art results over three commonly used benchmarks under different feature settings. It not only improves the prediction accuracy but also greatly reduces the mean error distance.