English-NER-system

Tutorials of how to create English Named Entity Recognition (NER) systems by finetuning the model distilbert-base-cased using the English subset of MultiNERD NER dataset.

System A recognizes all entity types included in the original dataset.

System B recognizes 5 entity types: PERSON(PER), ORGANIZATION(ORG), LOCATION(LOC), DISEASES(DIS), ANIMAL(ANIM). Other entity types are converted to the "O" type.

Usage

1. Open Google Colab.
2. Press "Open Colab".
3. Open a notebook from GitHub with the URL of System A or System B.
4. Change the runtime type to T4 GPU.
5. Run all the cells.

Environment and dependencies:

For the same training results, please use the environment built up inside notebook with one T4 GPU.

The required libraries are both listed inside the notebooks and in requirements.txt.