Network　representation　learning　represents　nodes　in　networks　as　low-dimension　vectors　which　has　been　attracting　increasing　attention　recently　due　to　its　effectiveness　in　network　analysis　tasks　such　as　classification　and　link　prediction.　In　this　paper,　our　focus　is　on　content　curation　social　networks　(CCSNs).　There　are　more　than　one　user　relation　subnetworks　formed　by　different　user　relations　in　a　social　media　network.　However,　most　of　existing　representation　learning　algorithms　usually　study　only　one　subnetwork　which　cannot　study　users　from　different　views.　On　the　other　hand,　most　of　the　existing　approaches　are　designed　for　universal　networks　which　do　not　consider　the　unique　characteristics　of　different　networks.　We　propose　a　multi-perspective　network　representation　based　on　human　curation　(MNHC)　model,　which　aims　to　infer　network　representations　across　multiple　relations　in　CCSNs.　The　network　representation　model　utilizes　the　unique　structure　of　networks　and　human　curation　signals　in　CCSNs　which　combines　two　user　relation　subnetworks　and　human　curation　signals　for　informative　network　representations.　Experiment　results　show　that　the　model　could　obtain　good　performance　in　both　classification　and　link　prediction　tasks.　©　2019　ACM.