thunlp/Few-NERD
A benchmark dataset and training code for few-shot Named Entity Recognition using BERT-based taggers.
Not currently ranked — collecting fresh signals.
star history
Few-NERD is a dataset released with an ACL 2021 paper designed for few-shot NER evaluation, supporting both few-shot and supervised training settings. The repository includes episode-sampled data, BERT tagger training scripts, and evaluation code for entity typing under limited-label scenarios. Researchers can download pre-sampled episodes or raw data and train models using the provided bash scripts.
Frequently asked
- What is thunlp/Few-NERD?
- A benchmark dataset and training code for few-shot Named Entity Recognition using BERT-based taggers.
- Is Few-NERD open source?
- Yes — thunlp/Few-NERD is open source, released under the Apache-2.0 license.
- What language is Few-NERD written in?
- thunlp/Few-NERD is primarily written in Python.
- How popular is Few-NERD?
- thunlp/Few-NERD has 400 stars on GitHub.
- Where can I find Few-NERD?
- thunlp/Few-NERD is on GitHub at https://github.com/thunlp/Few-NERD.