README.md

metadata

license: afl-3.0
language:
  - en
tags:
  - entity disambiguation
  - disambiguation
  - ned
  - GENRE
  - BLINK
pretty_name: Entity Disambiguation
task_categories:
  - zero-shot-classification
  - text-classification

Entity Disambiguation datasets as provided in the GENRE repo. The dataset can be used to train and evaluate entity disambiguators.

The datasets can be imported easily as follows:

from datasets import load_dataset

ds = load_dataset("boragokbakan/entity_disambiguation", "aida")

Available dataset names are:

• blink
• ace2004
• aida
• aquaint
• blink
• clueweb
• msnbc
• wiki

Note: As the BLINK training set is very large in size (~10GB), it is advised to set streaming=True when calling load_dataset.