Dataset Card for "LENU - Legal Entity Name Understanding"

---

in collaboration with

---

Dataset Summary

This dataset contains legal entity names from the Global LEI System in which each entity is assigned with a unique Legal Entity Identifier (LEI) code (ISO Standard 17441) along with their corresponding Entity Legal Form (ELF) Codes (ISO Standard 20275), which specifies the legal form of each entity.

The dataset has been created as part of a collaboration of the Global Legal Entity Identifier Foundation (GLEIF) and Sociovestix Labs with the goal to explore how Machine Learning can support in detecting the ELF Code solely based on an entity's legal name and legal jurisdiction. See also the open source python library LENU, which supports in this task.

The data is created from LEI data downloaded from GLEIF's public website (Date: 2022-11-01 00:00), where it is accessible free of charge. It is divided into subsets for a selection of legal jurisdictions, whereas each Jurisdiction has its own set of ELF Codes. The ELF Code reference list can be downloaded here.

Languages

The data contains several major Jurisdictions (e.g. US-DE (US Delaware), JP (Japan), DE (Germany) and others). Legal entity names usually follow certain language patterns, depending on which jurisdiction they are located in. Thus, we apply models that are pre-trained on the corresponding language.

Dataset Structure

Data Instances

The data contains of the LEI, the corresponding legal name and ELF Code.

{
  'LEI': '254900OMZ079O2SDWA75',
  'Entity.LegalName': 'Park Reseda Mortgage LLC',
  'Entity.LegalForm.EntityLegalFormCode': 0 
}

Data Fields

This is just a subset of available fields in the LEI system. All fields are described in detail in GLEIF's LEI Common Data Format (CDF).

• LEI: The Legal Entity Identifier Code. Uniquely identifies a Legal Entity.
• Entity.LegalName: The official name of the legal entity as registered in the LEI system.
• Entity.LegalForm.EntityLegalFormCode: class encoded column which contains the Entity Legal Form Code

Data Splits

We have divided each jurisdiction's subset into stratified train (70%), validation (10%) and test (20%) splits. ELF Codes that appear less than three times in a Jurisdiction have been removed.

Licensing Information

This dataset, which is based on LEI data, is available under Creative Commons (CC0) license. See gleif.org/en/about/open-data.