Question about the metric numbers on relation extraction?

[speedcell4]

I can not found the numbers in the original papers, could you please show me where did you get them?
https://nlpprogress.com/english/relationship_extraction.html

e.g. for the New York Times Corpus dataset, you report the RESIDE model can reach 73.6 on P@10% and 59.5 on P@30%. But I can not found those numbers in the original paper. Could you please show me from which figure you get those numbers?

[sebastianruder]

As far as I'm aware, the numbers were read from Figure 3a in the paper.

[speedcell4]

why do you consider to use this kind of metric? why not the P@100, P@200, P@300, and PR-AUC which are directly reported in the original papers?

[sebastianruder]

Those metrics were added by someone who considered them to be useful. Feel free to add any additional metrics that you are missing. :)