Entities get mapped to a metadata schema.

map.py

from transformers import pipeline

# Load a pretrained NER model for entity extraction
ner_pipeline = pipeline(
    "ner",
    model="dslim/bert-base-NER",
    aggregation_strategy="simple"
)

def extract_metadata_entities(ocr_text: str) -> dict:
    """Extract candidate people, places, and organizations
    from OCR-derived text for a digital library record."""
    entities = ner_pipeline(ocr_text)
    metadata = {"persons": [], "locations": [], "organizations": []}

    label_map = {
        "PER": "persons",
        "LOC": "locations",
        "ORG": "organizations",
    }

    for entity in entities:
        key = label_map.get(entity["entity_group"])
        if key and entity["score"] > 0.85:
            metadata[key].append(entity["word"])

    # Deduplicate while preserving order
    for key in metadata:
        metadata[key] = list(dict.fromkeys(metadata[key]))

    return metadata

sample_text = (
    "Correspondence between Marie Curie and the Sorbonne "
    "regarding laboratory funding, Paris, 1911."
)
print(extract_metadata_entities(sample_text))