Closed jasperhyp closed 1 year ago
Hi,
For the design of KE loss function, we refer to the loss function in the KEPLER [1] and TransE, and we think that TransE is an easy and effective KE method.
[1] KEPLER: A Unified Model for Knowledge Embedding and Pre-trained Language Representation, 2021
Thank you! Very helpful.
Regarding your KE loss function, could you kindly provide some intuitions on why this specific loss function was chosen (given there are so many metric learning losses on KG)? A few relevant pieces of literature that you referenced would be appreciated.