WISER

WISER (Weak and Indirect Supervision for Entity Recognition) is a library that trains sequence-to-sequence models using programmatic weak supervision. Users provide rules / heuristics (in the form of labeling functions) to train neural networks for sequence labeling tasks such as named entity recognition.

Installation

WISER requires Python 3.7. To use WISER, please run

git clone https://github.com/BatsResearch/wiser.git
cd wiser
pip3 install -r requirements.txt

If you are using conda environment, you can alternatively run

git clone https://github.com/BatsResearch/wiser.git
cd wiser
conda install --file requirements.txt

Debugging

• If you run into the error of TypeError: Params.pop: key must be a supertype of <class 'inspect._empty'> but is <class 'str'>, you can fix the issue by running pip3 install overrides==4.1.2.