A Hierarchical Location Prediction Neural Network for Twitter User Geolocation

TL;DR

This paper introduces a hierarchical neural network model for Twitter user geolocation that predicts country and city levels sequentially, utilizing character-aware embeddings to handle noisy data, achieving state-of-the-art accuracy and reduced error distances.

Contribution

The paper presents a novel hierarchical neural network architecture that incorporates country-level guidance and character-aware embeddings for improved geolocation accuracy.

Findings

Achieves state-of-the-art results on three benchmarks.

Reduces mean error distance significantly.

Effectively handles noisy tweet data with character-aware embeddings.

Abstract

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.