Deepak Nathani
Results
2
comments of
Deepak Nathani
As you can see in this paper (https://www.aclweb.org/anthology/2020.acl-main.489.pdf), we are dealing with two different issues here. 1. Test data leakage 2. Wrong evaluation strategy for the decoder part. Above mentioned...
The reported results (with ConvE as decoder) are with randomized evaluation strategy and they are independent of the evaluation strategy. The task formulation is still the same but the evaluation...