I needed a way to match phrases using a fuzzy approach to prepare an annotated dataset to train a model. After searching in forums etc., I realized that I was not the only one looking for this to work, so I share my solution here.

Clone the repo:

git clone https://github.com/jackmen/fuzzy_spacy.git

Create a virtual environment and activate:

python3 -m venv your_env_name
cd your_env_name
source bin/activate

Install requiremets:

cd /your_path/fuzzy_spacy/
pip3 install -r requirements.txt

Run jupyter notebook

jupyter notebook

Follow the steps presented in the jupyter notebook. To run the PhuzzyMatcher on a batch of documents, check the spacy documentation (https://spacy.io).

The code execution is not fast, but was sufficient for my use-case to annotate documents to train a NER model.