Reproducing results in papers

[antonyscerri]

Hi

Could you comment on which version of this repository or the associated supervised-oie (and the benchmark included in that) can be used to reproduce results in your papers (including "Creating a Large Benchmark for Open Information Extraction"). There appear to be two sets of code which differ, various versions of the corpora and its also unclear which splits may have been used. So far I have found several different PR Curve plots and been able (with fixes to code) to generate my own but they are all different.

I'm also curious as to what might have been the latest state wrt to any performance of each system tested. I've not even considered any updates to those and the outputs included in this repo wrt their current state, this might also explain some differences.

Thanks

Tony