Closed yangjingyi closed 2 years ago
Hi! This repo used to be this replica, but as the commits were squashed together before uploading, you can't find it on github. If you specifically need my reimplementation, I could find a local commit and send it to you. If you just need a pytorch version of the model by Joshi et al., I suggest you take a look at this repo: https://github.com/emorynlp/coref-hoi (SpanBERT + Attended Antecedent is the one).
Thank you for your quick reply! I tried your suggested repo but I only get about avg F1 78.1 on test dataset. Even when I use Provided trained models, I cannot get the correct result. Did you get the reported result? Maybe there are some mistake of my data?
My guess is that you are probably better off asking this question there :)
I have not run the SpanBERT + AA model, but the results of EE and no-HOI models were close to the ones reported in the paper.
Ok. Thanks for answering. I will close this issue:)
Hi,
Which code did you use in the reimplementation of Joshi et al. pytorch version? Thanks