mandarjoshi90/coref
An implementation of BERT and SpanBERT transformer models for coreference resolution, achieving up to 79.6 F1 on OntoNotes.
Not currently ranked — collecting fresh signals.
star history
This repository provides code and pretrained models for coreference resolution using BERT and SpanBERT. It extends the e2e-coref architecture with transformer-based encoders, enabling span-based mention detection and clustering for identifying all expressions referring to the same entity in text. The models are fine-tuned on the OntoNotes dataset and can be used via pretrained downloads or trained from scratch with GPU resources.
Frequently asked
- What is mandarjoshi90/coref?
- An implementation of BERT and SpanBERT transformer models for coreference resolution, achieving up to 79.6 F1 on OntoNotes.
- Is coref open source?
- Yes — mandarjoshi90/coref is open source, released under the Apache-2.0 license.
- What language is coref written in?
- mandarjoshi90/coref is primarily written in Python.
- How popular is coref?
- mandarjoshi90/coref has 454 stars on GitHub.
- Where can I find coref?
- mandarjoshi90/coref is on GitHub at https://github.com/mandarjoshi90/coref.