weizhepei/CasRel
A BERT-based cascade framework for extracting subject-relation-object triples from text, published at ACL 2020.
Not currently ranked — collecting fresh signals.
star history
CasRel implements a novel binary tagging approach where relations are modeled as functions mapping subjects to objects, rather than discrete classification labels. The framework first identifies all possible subjects in a sentence, then applies relation-specific taggers to simultaneously extract corresponding objects. Built on Keras and BERT, it supports multiple relation extraction benchmarks including NYT, WebNLG, and ACE04 datasets.
Frequently asked
- What is weizhepei/CasRel?
- A BERT-based cascade framework for extracting subject-relation-object triples from text, published at ACL 2020.
- Is CasRel open source?
- Yes — weizhepei/CasRel is open source, released under the MIT license.
- What language is CasRel written in?
- weizhepei/CasRel is primarily written in Python.
- How popular is CasRel?
- weizhepei/CasRel has 814 stars on GitHub.
- Where can I find CasRel?
- weizhepei/CasRel is on GitHub at https://github.com/weizhepei/CasRel.