Quick Start

This page walks through running CHNOBLi on the provided example data, so you can confirm your setup works before using your own documents.

Note

We highly recommend you run this on a GPU, as running this on a CPU is very slow due to the FlairNLP model and vector database components.

Try It with Example Data

1. Tag Example Documents

sh scripts/tag_example.sh

On Windows:

python main.py --tasks prep,tag --magazine_year_paths ./data/input_example/tjb/1955_030 --config_file configs/configurations_example.json

Output: data/output/tag/

3. Evaluate Results

sh scripts/eval_example.sh

On Windows:

python3 main.py --tasks eval --config_file ./configs/eval_config_example.json --eval_level ref

Output: data/output/eval_ref_top1/tjb/1955_030.jsonl

Next Steps

Once you’ve confirmed the pipeline runs correctly on the example data, see Using Your Own Data for how to bring in your own OCR output or custom tagging results.