Abdullah-afify/egyptian-names

Egyptian Names & Onomastic Intelligence Dataset

From 15.88M+ Raw National Records to an Empirical Onomastic and Linguistic Engine

This repository hosts the complete, multi-phase statistical and linguistic dataset powering `egy-names`, the production onomastic intelligence engine for contemporary Egyptian naming traditions.

Dataset Pipeline Overview

Egyptian names follow an unbroken patronymic lineage chain ($Personal + Father + Grandfather + Ancestor + Family/Tribe$). This dataset provides the full transformation pipeline across all developmental phases:

[ Phase 0: 15.88M+ Raw Full Name Chains (30M+ Records) ]
              │
              ▼
[ Phase 1: 63.5M+ Segmented Patronymic Tokens & Positions ]
              │
              ▼
[ Phase 2: 43.3K Unique Raw Token Frequencies ]
              │
              ▼
[ Phase 3: 23.4K+ Orthographic & Spelling Corrections ]
              │
              ▼
[ Phase 4: 44.6K Master Annotated Canonical Names ]
(Gender + Religion + Generational Slot Probabilities + Tashkeel + Meanings + Transliterations)

Full Data vs. Single Names Breakdown

Understanding Population Records vs. Onomastic Lexicon

In an Egyptian population of ~16–30 Million records, names repeat extensively across generations:

• —Highly common given names like محمد, أحمد, محمود, علي, فاطمة, مريم occur millions of times.
• —Family surnames like الشرقاوي, السيد, إبراهيم occur tens of thousands of times.
• —When every patronymic chain is decomposed and deduplicated, the complete onomastic vocabulary of Egypt consists of 44,626 unique canonical lemmas, capturing >99.9% of all contemporary and historical Egyptian personal and family names.

Quick Start with Hugging Face datasets

1. Load Final Canonical Master Dataset (Default — 44,626 Names)

from datasets import load_dataset

# Load 44,626 annotated canonical names
dataset = load_dataset("Abdullah-afify/egyptian-names")
print(dataset["train"][0])

2. Load Raw Unprocessed Full Names (Phase 0 — 1.54M Sample / 15.88M Corpus)

raw_dataset = load_dataset("Abdullah-afify/egyptian-names", "phase0_raw")
print(f"Total raw names: {len(raw_dataset['train']):,}")
# Example: 'احمد باسم لينان محمد سليمان'

3. Load Segmented Patronymic Chains (Phase 1)

chains_dataset = load_dataset("Abdullah-afify/egyptian-names", "phase1_segmented")
# Inspect token-by-token genealogical position
print(chains_dataset["train"][:3])

4. Load Unique Token Frequency Analysis (Phase 2)

freq_dataset = load_dataset("Abdullah-afify/egyptian-names", "phase2_frequencies")
# Top names by frequency across the national corpus
print(freq_dataset["train"][:5])

5. Load Spelling & Typo Correction Rules (Phase 3 — 23,457 Rules)

corr_dataset = load_dataset("Abdullah-afify/egyptian-names", "phase3_corrections")
# Map misspelling to canonical form: 'احمد مصطفا' -> 'أحمد مصطفى'

Configurations & Data Schemas

1. final_canonical (Default — 44,626 Rows)

The master 14-dimensional enriched onomastic dictionary:

Related model

The `egy-names-fallback-classifier` model is trained directly on this dataset’s canonical catalog, and is used as the offline ML fallback for names outside this catalog inside the `egy-names` library.

Research Citations & Licensing

This dataset is released under the MIT License and is free for academic, commercial, and research use.

@dataset{afify2026egyptian_names,
  author       = {Abdullah Afify},
  title        = {Egyptian Names & Onomastic Intelligence Dataset},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/Abdullah-afify/egyptian-names}}
}