storytracer/bhl_copyright_statuses

Biodiversity Heritage Library Copyright Statuses

This dataset contains all unique copyright statuses present in the items.txt.gz file of the Biodiversity Heritage Library open dataset on AWS Open Data. The unique copyright statuses were extracted, grouped and sorted by frequency using the following DuckDB query:

COPY (SELECT CopyrightStatus, COUNT(*) as Count FROM read_csv('https://bhl-open-data.s3.amazonaws.com/data/item.txt.gz') GROUP BY CopyrightStatus ORDER BY Count DESC) TO 'item_copyrightstatus_count.csv';

UV Dataset Classifier

Using the UV Dataset Classification Script, we use SmolLM3-3B (a fast 3B model with native thinking capabilities) to assign each copyright status one of two labels: in_copyright or public_domain. The descriptions for both labels have been adjusted to deal with idiosyncratic features of the BHL copyright status strings, specifically the use of the abbreviation NIC, short for "not in copyright".

Here is the full command that was run to classify the copyright statuses:

hf jobs uv run \
  --flavor l4x1 \
  --secrets HF_TOKEN \
  --timeout 1h \
  --with "vllm<0.12.0" \
  https://huggingface.co/datasets/uv-scripts/classification/raw/main/classify-dataset.py \
  --input-dataset "storytracer/bhl_copyright_statuses" \
  --column "CopyrightStatus" \
  --labels "public_domain,in_copyright" \
  --label-descriptions "public_domain:Work is explicitly public domain or has no known copyright restrictions or is marked with the abbreviation NIC (short for Not In Copyright);in_copyright:Work is protected by copyright or requires permission or has a named owner" \
  --output-dataset "storytracer/bhl_copyright_statuses_classified" \
  --enable-reasoning

Classified copyright statuses

The classified statuses are saved in the dataset storytracer/bhl_copyright_statuses_classified. The resulting dataset contains a reasoning column, which documents the reasoning of the model in classifying each status.