Therefore, to address this challenge, we propose novel relation-aware line graph neural networks for cross-lingual entity alignment (RALG). More specifically, we first propose to learn relation ...