Skip to main content

Quick Start

Make sure you have installed Alligator and have MongoDB and LAMAPI running before proceeding.

Python API​

Minimal example​

Alligator will automatically classify columns and run the full pipeline:

from alligator import Alligator

gator = Alligator(
    input_csv="tables/my_table.csv",
    num_workers=4,
    num_ml_workers=2,
    worker_batch_size=64,
    candidate_retrieval_limit=16,
    mongo_uri="mongodb://localhost:27017/",
)
gator.run()

With Wikidata type constraints​

Constrain candidate search per NE column to specific Wikidata entity types for higher precision:

from alligator import Alligator

# Map column index (as string) → list of Wikidata QIDs
column_types = {
    "0": ["Q11424"],   # film
    "3": ["Q483394"],  # music genre
    "7": ["Q5"],       # human
    "8": ["Q5"],       # human
}

gator = Alligator(
    input_csv="tables/imdb_top_100.csv",
    column_types=column_types,
    num_workers=4,
    num_ml_workers=2,
    worker_batch_size=64,
    candidate_retrieval_limit=16,
    mongo_uri="mongodb://localhost:27017/",
)
gator.run()

CLI​

python3 -m alligator.cli \
  --gator.input_csv tables/my_table.csv \
  --gator.num_workers 4 \
  --gator.num_ml_workers 2 \
  --gator.worker_batch_size 64 \
  --gator.candidate_retrieval_limit 16 \
  --gator.mongo_uri mongodb://localhost:27017/

See the full CLI Reference for all available options.

What Happens Next​

After run() completes:

  • Annotations are stored in MongoDB (alligator_db database by default)
  • If save_output_to_csv=True, a CSV file is written alongside the input
  • CEA / CTA / CPA results are accessible via the MongoDB input_data collection

For more details on the output format, see the Pipeline Architecture page.