This library helps with the generation of fingerprints for entity data. A fingerprint in this context is understood as a simplified entity identifier, derived from it's name or address and used for cross-referencing of entity across different datasets.
import fingerprints
fp = fingerprints.generate('Mr. Sherlock Holmes')
assert fp == 'holmes sherlock'
fp = fingerprints.generate('Siemens Aktiengesellschaft')
assert fp == 'ag siemens'
fp = fingerprints.generate('New York, New York')
assert fp == 'new york'A significant part of what fingerprints does it to recognize company legal form
names. For example, fingerprints will be able to simplify Общество с ограниченной ответственностью to ООО, or Aktiengesellschaft to AG. The required database
is based on two different sources:
- A Google Spreadsheet created by OCCRP.
- The ISO 20275: Entity Legal Forms Code List
Wikipedia also maintains an index of types of business entity.
- Clustering in Depth, part of the OpenRefine documentation discussing how to create collisions in data clustering.
- probablepeople, parser for western names made by the brilliant folks at datamade.us.
- The study Developing a Legal Form Classification and Extraction Approach For Company Entity Matching by Kruse et al. (2021) investigates four approaches for identifying and classifying legal forms in company names.
- List of Legal Forms from AnaCredit dataset by ECB (one of the Annexes).
- Transformer-based Entity Legal Form Classification by Arimond et al. (2023).