Enriching Loss Functions for Link Prediction with Domain and Range Constraints

The authors propose enriching loss functions for link prediction by treating different negatives differently based on their semantic validity. This approach leads to better results in terms of MRR, Hits@10 values, and Sem@K metric.