BERT and SpanBERT for Coreference Resolution

[sayanb-7c6]

🌟 New model addition

Model description

This is a recent approach for co-reference resolution using BERT, implemented from the papers BERT for Coreference Resolution: Baselines and Analysis and SpanBERT: Improving Pre-training by Representing and Predicting Spans, which is the current state of the art on OntoNotes (79.6 F1). It uses tensorflow 1.14.0.

Reason why this is interesting is it achieves strong improvements on the OntoNotes (+3.9 F1) and GAP (+11.5 F1) benchmarks. Also, I think it would be a nice addition to huggingface library, as it has only the neuralcoref as the coreference resolution module.

Open source status

• [x] the model implementation is available: (https://github.com/mandarjoshi90/coref)
• [x] the model weights are available: (https://github.com/facebookresearch/SpanBERT)
• [x] who are the authors: (@mandarjoshi90, @jkkummerfeld, @wenyudu)

[krrishdholakia]

Commenting for visibility - is this available now ? can't seem to find it, would love to use this for a question-answering project i'm working on!

[eugeneware]

I'd also like to see this model incorporated into the core list of supported models. I did note that you can download it from the community models here though: https://huggingface.co/SpanBERT/spanbert-base-cased

[iliemihai]

Are there any translations of the above repository (https://github.com/mandarjoshi90/coref) into the awesome HuggingFace API ? That would be very cool to test :D !

[hlahkar]

I would like to work on this...but will need some guidance