hanine-bgt/ELNER-DZ

ELNER-DZ: A Dataset for Named Entity Recognition and Linking in Algerian Arabic Dialect (Darija) # ELNER-DZ: Algerian Arabic Dataset for Named Entity Recognition and Entity Linking > 📌 **Official DOI**: > 🤗 Also available on Hugging Face > 📥 Download Dataset (.rar from Zenodo) --- 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 MSA, `arq` for dialectal), Arabizi (Latin), French (`fr`), English (`en`) - **Script**: Arabic and Latin (Arabizi) - **Format**: JSON (compressed in `data.rar`) - **Annotations**: - Named Entity spans (start, end) - NER labels (PER, LOC, ORG, etc.) - Normalized forms - Wikidata QIDs --- ## 📁 File Structure - `data/data.rar` — Compressed archive containing `data.json` - `examples/loading_example.py` — Script to extract and load the dataset - `LICENSE` — CC-BY-4.0 - `dataset_card.md` — Hugging Face dataset summary --- ## ✨ Example Format ```json { "id": 188, "text": "3reft wa7ed lperson khadem f Yassir", "entities": [ { "start": 29, "end": 35, "label": "ORG", "wikidata_id": "Q117156470", "normalized": "Yassir" } ] } ```` --- ## 🏷️ Entity Types * `PER`: Person * `LOC`: Location * `ORG`: Organization * `PROD`: Product * `LAW`: Legal texts or rules * `LANG`: Language * `EVENT`: Events * `DATE`: Temporal expressions * `NORP`: Nationality/Religious/Political groups * `SPORT`: Sports & Competitions * `SYMPTOM`, `DISEASE`: Medical categories * `MISC`: Miscellaneous --- ## 🧪 Tasks Supported * Named Entity Recognition (NER) * Entity Linking (EL) with Wikidata * Dialectal NLP in Algerian Arabic * Code …