Datasets:

HadjerHaninebgt7878
/

ELNER-DZ

license: cc-by-4.0
task_categories:
  - text-classification
  - token-classification
  - feature-extraction
language:
  - ar
  - fr
  - en
tags:
  - darija
  - dz
  - arabizi
  - entity-linking
  - el
  - ner
  - named-entity-recognation
  - wikidata
  - 2m
  - algerian
  - algeria
  - dialect
size_categories:
  - 1M<n<10M

ELNER-DZ: Algerian Arabic Dataset for Named Entity Recognition and Entity Linking

This dataset, titled ELNER-DZ, was created by Bouguettoucha Hadjer Hanine and Djouablia Ilhem as part of our Master’s thesis . It is the first large-scale dataset designed for Named Entity Recognition (NER) and Entity Linking (EL) in Algerian Arabic Dialect (Darija), including both Arabic script and Arabizi (Latin-script).

This dataset contains over 2 million dialectal sentences labeled with more than 1.9 million named entities and linked to Wikidata QIDs.

🧾 Dataset Summary

Name: ELNER-DZ
Languages: Arabic (ar{For Modern Standard Arabic}, arq{For dialectal Arabic}), Arabizi (Latin), French (fr), English (en)
Script: Arabic script and Latin script (Arabizi)
Format: JSON (single file data.json)
Annotations:
- Named Entity spans (start, end)
- Entity labels (PER, LOC, ORG, PROD, EVENT, LANG, LAW, MISC, DATE, NORP)
- Wikidata IDs
- Normalized canonical forms

📁 File Structure

The dataset is provided as a single JSON file:

data.json: A list of annotated examples. Each entry includes:
- id: Unique ID of the sentence
- text: Sentence string (Darija in Arabic or Arabizi)
- entities: List of dictionaries with:
  - start, end: Character offsets of the entity span
  - label: NER label (PER, LOC, ORG, etc.)
  - normalized: Canonical surface form
  - wikidata_id: Corresponding QID from Wikidata

✨ Example

{
  "id": 188,
  "text": "3reft wa7ed lperson khadem f Yassir",
  "entities": [
    {
      "start": 29,
      "end": 35,
      "label": "ORG",
      "wikidata_id": "Q117156470",
      "normalized": "Yassir"
    }
  ]
}

🏷️ Entity Types

The dataset includes the following entity types (NER labels):

PER: Person
LOC: Location
ORG: Organization
PROD: Product
LAW: Legal concepts or texts
LANG: Language
EVENT: Events
DATE: Dates
NORP: Nationality, Religious, or Political groups
MISC: Miscellaneous
SPORT: Sports and Competitions
SYMPTOM, DISEASE: Medical categories

🧪 Tasks Supported

This dataset can be used for:

Named Entity Recognition (NER)
Entity Linking (EL) with Wikidata
Dialectal NLP in Algerian Arabic
Multilingual and multiscript NER modeling
Pretraining or fine-tuning low-resource models

🧰 How to Use

Install 🤗 Datasets and load the JSON file:

from datasets import load_dataset

dataset = load_dataset("HadjerHaninebgt7878/ELNER-DZ", data_files="data.json", split="train")
print(dataset[0])

🔍 Dataset Details

Source: Social media, forums, dialogues, e-commerce, Wikidata SPARQL
Annotation:
- Manual and semi-automatic annotation
- Entity normalization
- Linking to Wikidata (QIDs)

👩‍💻 Authors

Bouguettoucha Hadjer Hanine
Djouablia Ilhem

📄 License

This dataset is licensed under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

You are free to share, copy, redistribute, adapt, and build upon the material for any purpose, even commercially, as long as you give appropriate credit.

🔗 View full license

📚 Citation

Please cite this dataset as:

@dataset{elnerdz2025,
  author       = {Bouguettoucha, Hadjer Hanine and Djouablia, Ilhem},
  title        = {ELNER-DZ: Algerian Arabic Dataset for Named Entity Recognition and Entity Linking},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/datasets/HadjerHaninebgt7878/ELNER-DZ}},
  note         = {Dataset created as part of Master’s Thesis }
}