---
language:
- en

configs:
  - config_name: dataset
    data_files:
      - split: train
        path: data/train-00001-of-00001.jsonl
      - split: validation
        path: data/dev-00001-of-00001.jsonl
    default: true
  - config_name: dictionary
    data_files:
      - split: kb
        path: dictionary/dictionary-00001-of-00001.jsonl

size_categories:
- 100K<n<1M
---

# Dataset Card for ZELDA
ZELDA is a benchmark for Entity Disambiguation.
As there is no development split in ZELDA, we split the dataset using the first 90% for training and the remaining 10% for development.
For the entity dictionary, we use Wikipedia page ids and Wikidata descriptions processed by [Rücker and Akbik, 2025](https://github.com/flairNLP/VerbalizED).

## Dataset Description
- **Repository:** [https://github.com/flairNLP/zelda](https://github.com/flairNLP/zelda)
- **Public:** True
- **Source:** [Kensho Derived Wikimedia Dataset](https://www.kaggle.com/datasets/kenshoresearch/kensho-derived-wikimedia-data)
- **Paper:** [ZELDA: A Comprehensive Benchmark for Supervised Entity Disambiguation](https://aclanthology.org/2023.eacl-main.151/)
- **Number of Entities** 821401


### Citation Information
```
@inproceedings{milich2023zelda,
  title={{ZELDA}: A Comprehensive Benchmark for Supervised Entity Disambiguation},
  author={Milich, Marcel and Akbik, Alan},
  booktitle={{EACL} 2023,  The 17th Conference of the European Chapter of the Association for Computational Linguistics},
  year={2023}
}
```