metadata

license: mit
task_categories:
  - token-classification
language:
  - de
tags:
  - relation extraction
  - entity extraction
  - legal
  - tax law
pretty_name: Key Figures from german Tax Acts
size_categories:
  - n<1K

Dataset Card for Dataset Name

Dataset Description

This dataset contains Key Figures with their properties from german tax acts. The dataset is annotated by tax experts and consists of 85 annotated paragraphs from 14 different German tax acts with 157 annotated tax key figures. The annotation was performed based on a developed universally applicable annotation schema and a semantic model for key figures and their properties in legal texts. More details about the schema and the semantic model can be found in our Paper. There we also provide a knowledge graph populated from these annotated paragraphs.

Repository: https://github.com/danielsteinigen/nlp-legal-texts
Paper: https://ceur-ws.org/Vol-3441/paper7.pdf

Supported Tasks and Leaderboards

Entity Extraction
Relation Extraction

Languages

German

Dataset Structure

Data Fields

id: a unique ID of the data sample
law: the abbreviation of the law from which this parapgraph originates
paragraph: the title of the paragraph
text: the actual text string of the paragraph
entities: list of entity objects
relation: list of relation objects

Entities:

id: a unique ID of the entity
start: start character offset of the entity
end: end character offset of the entity
entity: label/name of the entity
subclass: entity subclass, if there is one, else NULL
text: text string of the entity

Relations:

id: a unique ID of the realation
head: ID of the head entity
tail: ID of the tail entity
relation: label/name of the relation

Dataset Creation

More details about the annotation process can be found in our Paper.