Regarding details of the model used for the TACRED experiments shown in section 7.2 of the paper

[anushkasw]

Hello,

I found your paper very interesting and wanted to apply it to my work. I wasn't able to completely understand the methodology used for calculating the relation similarity on the TACRED dataset mentioned in section 7.2 of the paper.

I wanted to know if you trained a separate fact distribution model on the TACRED dataset? And if so, how did you generate the pretrained embeddings?

Thanks,
Anushka