Beyond Words Alone: Disentangling Distributed Evidence for Visually Grounded Cultural Translation