The problem of training pair set (sup_ent_ids).

[pp413]

Did no one find that the file "sup_ent_ids" was wrong? The two entity ids in each entity pair of sup_ent_ids are same.

[sunzequn]

Hi,

Thanks for your interest in our work.

This is not a mistake. The file "sup_ent_ids" contains IDs for seed entity alignment as training data. Each pair of aligned entities uses the same ID to merge their representations. In this way, two aligned entities are actually one identify in the model such that we do not need an additional alignment learning loss to minimize their distance.