How did you evaluate on SCIERC in section 5.2.1?

[feiba54]

First thanks for your brilliant work!
In your paper(section 5.2.1) you trained DYGIE on both your dataset and SCIERC dataset. Since SCIERC do not have cross-sentence relations, I was wandering how did you model cross-sentence relations by DYGIE, or you just ignored them?

[successar]

Hi Yes, You are correct -- DyGIE couldn't model the cross sentence relations so we ignore them (but they are included as part of evaluation, since one of the goals of SciREX was to model cross sentence / cross paragraph relations).