NARA introduces a unified self-supervised method for learning relational, context-dependent representations of heterogeneous vector geoentities that improves performance on building classification, traffic prediction, and POI recommendation.
Rethink geographical generalizability with unsupervised self-attention model ensemble: A case study of openstreetmap missing building detection in africa
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
NARA: Anchor-Conditioned Relation-Aware Contextualization of Heterogeneous Geoentities
NARA introduces a unified self-supervised method for learning relational, context-dependent representations of heterogeneous vector geoentities that improves performance on building classification, traffic prediction, and POI recommendation.