Twitter User Geolocation based on Heterogeneous Network and Label Propagation

Abstract

Twitter user geolocation is applied into various applications, such as local messages recommendation and event location recognition. Existing methods do not usually utilize the potential positional correlation between heterogeneous information, resulting in low user geolocation accuracy. Therefore, this paper proposes a user geolocation method based on heterogeneous network and label propagation. First, a heterogeneous network is built based on heterogeneous information mined from user tweets. Then, by analyzing the factors that impact the efficiency and accuracy of label propagation, redundant nodes and edges in the initialized heterogeneous network are removed. Finally, the locations of unlabeled users are geolocated through label propagation in the simplified heterogeneous network. Experiments on the publicly available Twitter benchmark dataset show that proposed method accurately locate 57% of users within 161 km of their locations, a significant improvement over several existing methods.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 3 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.