The notebook shows how to improve the performance of the CRF algorithm about named entity recognition, using expert.ai edge NL API as feature generation step.
The task has been trained and measured with reference to CoNLL 2003 dataset; the notebook downloads the dataset from this https://github.com/nluninja/nlp_datasets.

it requires sklearn-crfsuite and edge-nlapi libraries.

conda env create -f environment.yml

conda activate crf-ner

Edge NL API is a local runtime for the expert.ai Edge NL API. For this task, I use edge NL Api as feature generation step relying on the metadata generated by the NLP engine. Create your project in expert.ai studio, and then deploy the edge instance selecting _Studio-->Deploy. Then run

runmeWindows.bat

or
runmeLinux.sh

according to your environment

Edge NL API is free to use; it only requires an account creation at https://developer.expert.ai for being able to use it.