README.md

metadata

language:
  - en
  - ar
  - fr
  - es
  - zh
  - ru
  - hi
  - mul
license: cc-by-4.0
tags:
  - tabular-classification
  - demographic
  - linguistics
  - names
  - geography
  - society
  - script-detection
pretty_name: Global Human Names Dataset (Enriched)
size_categories:
  - 1M<n<10M

Global Human Names Dataset (Enriched)

Dataset Description

The Global Human Names Dataset is a massive, highly structured demographic dataset containing 9.5 million rows of geographic, gender-based, and script-analyzed name distributions. It maps over 1.7 million unique human names (first names and last names) across more than 100 countries worldwide.

This dataset is designed for sociolinguistic research, machine learning (e.g., gender prediction, cultural origin inference, script detection), and demographic analysis. It aggregates data from multiple massive public sources, including statistical representations of over 533 million individuals globally.

Dataset Summary

• Total Records: 9,504,865
• Unique First Names: ~727,555
• Unique Last Names: ~983,824
• Unique Countries Represented: 104 (ISO 3166-1 alpha-2 format)
• Timeframe: Contemporary / Modern

Data Structure

Data Fields

The dataset is provided in a single CSV file with the following columns:

• Original_Name (string): The name in its original or romanized format.
• Gender (string): The statistically dominant gender for the name.
  • M: Male
  • F: Female
  • U: Unisex or Unknown/Not specified (Common for Last Names).
• Name_Type (string): Classification of the name.
  • First_Name
  • Last_Name
• Country_Code (string): The 2-letter ISO country code (e.g., US, YE, SA, FR) where the name is prominent.
• Source (string): The originating dataset or source from which the record was extracted (e.g., names-dataset-533M).
• Name_Script (string): [NEW] The writing system (Alphabet/Script) detected using Unicode block analysis (e.g., Latin, Arabic, Cyrillic, CJK, Devanagari).

(Note: Because a single name can appear in multiple countries, the dataset contains multiple rows per name. For example, the name "Ali" might have a row for "YE", a row for "EG", and a row for "AE").

Sample Row

Original_Name	Gender	Name_Type	Country_Code	Source	Name_Script
Ahmed	M	First_Name	YE	names-dataset-533M	Latin
ماجدولين	F	First_Name	MA	names-dataset-533M	Arabic

Statistics & Insights

Based on exploratory data analysis (EDA) of the 9.5 million rows:

• Script Distribution: Latin dominates globally (86.17%), but Arabic holds a massive second place (8.70% / ~827,000 occurrences), followed by Cyrillic (2.51%) and CJK (1.08%).
• Cultural Naming Patterns: In the CJK (Chinese/Japanese/Korean) script, a staggering 90.4% of entries are First Names. In contrast, Cyrillic names are predominantly Last Names (64.8%).
• Most Diverse Countries: The UAE (AE) and Iraq (IQ) rank as the most globally diverse countries in this dataset, containing names written in all 10 major world scripts.
• Gender Distribution: Unknown/Unisex (58.7%), Male (25.7%), Female (15.6%). Note: The high percentage of Unisex is due to Last Names, which are inherently gender-neutral.
• Most Common Starting Letter: Names starting with the letter 'M' are the most common globally (8.53%).

Usage

Pandas

You can easily load and filter the dataset using Pandas:

import pandas as pd

# Load the dataset
df = pd.read_csv("hf://datasets/aborasheed/Global-Human-Names/global_names_with_scripts.csv")

# Filter for female Arabic-script names in Saudi Arabia (SA)
saudi_arabic_females = df[(df['Country_Code'] == 'SA') & (df['Gender'] == 'F') & (df['Name_Script'] == 'Arabic')]
print(saudi_arabic_females.head())

Hugging Face Datasets

from datasets import load_dataset

dataset = load_dataset("aborasheed/Global-Human-Names")

Ethical Considerations & Privacy

• Anonymization: This dataset contains no personally identifiable information (PII). There are no phone numbers, email addresses, physical addresses, or full names tied to specific individuals.
• Aggregation: The data consists purely of aggregated statistical counts and probabilities of names appearing in certain regions.
• Bias & Representation: While massive, the dataset may still exhibit geographic biases based on internet penetration and social media usage in different countries. Some regions may be overrepresented (e.g., Western Europe, North America) while others may be underrepresented.

Credits & Sources

This dataset was compiled and organized by [aborasheed]. The underlying statistical data utilizes the open-source names-dataset library by Philippe Remy, which inferred geographic and gender distributions from large-scale anonymized public datasets. Unicode script detection was applied post-extraction to enrich the linguistic depth of the dataset.