Franticuk/job-titles

Comprehensive Job Titles Dataset

A high-quality, deduplicated dataset of 65,248 unique job titles compiled from authoritative sources including ESCO (European Skills, Competences, Qualifications and Occupations), O*NET (Occupational Information Network), and OSCA (Occupational Skills and Competencies Australia).

Dataset Description

This dataset provides a comprehensive collection of job titles that have been carefully processed to remove duplicates and near-duplicates using semantic similarity matching. It serves as a valuable resource for:

• Job matching and recommendation systems
• Resume parsing and analysis
• Labor market research
• Career counseling applications
• HR technology development
• Natural language processing tasks related to employment

Dataset Structure

The dataset is provided in Parquet format with a single column:

• job_title (string): The standardized job title

Example entries:

.NET Developer
2D Animation Artist
Accounting Clerk
Administrative Assistant
Agricultural Engineer
AI Research Scientist
Business Analyst
Chef
Data Scientist

Sources

The dataset combines job titles from three major occupational classification systems:

1. ESCO v1.2.0 (European Commission)

   • ~33,000 occupations with multilingual support
   • Includes preferred labels, alternative labels, and hidden labels
   • Structured according to ISCO-08 classification

2. O*NET Database v29.3 (U.S. Department of Labor)

   • ~1,000 detailed occupational descriptions
   • Comprehensive taxonomy of U.S. occupations
   • Includes detailed job characteristics and requirements

3. OSCA (Australian Government)

   • Australian occupational classifications
   • Principal titles, alternative titles, and specializations

Processing Pipeline

1. Extraction

Job titles were extracted from multiple source files:

• ESCO: Preferred labels and alternative labels from occupations_en.csv
• O*NET: Occupation titles from Occupation Data.txt
• OSCA: Principal titles and alternative titles from Excel files

2. Deduplication

A sophisticated deduplication process was applied:

• Embedding Model: sentence-transformers/all-mpnet-base-v2
• Similarity Threshold: 0.85 (cosine similarity)
• Strategy: Length-based blocking for efficiency
• Preference: Shorter titles retained (typically more general/common)

The deduplication process identified semantically similar job titles such as:

• "Software Developer" and "Software Engineer"
• "Administrative Assistant" and "Admin Assistant"
• "Customer Service Representative" and "Customer Service Rep"

3. Quality Control

• Removed exact duplicates (case-insensitive)
• Filtered out malformed entries
• Standardized formatting and capitalization
• Preserved diversity while eliminating redundancy

Statistics

• Total unique job titles: 65,248
• Original titles before deduplication: ~100,000+
• Reduction rate: ~35% (semantic duplicates removed)
• File size: 756.7 KB (Parquet format with Snappy compression)

Use Cases

1. Job Search and Matching

import pandas as pd

# Load the dataset
df = pd.read_parquet('jobs.parquet')

# Search for data-related jobs
data_jobs = df[df['job_title'].str.contains('Data', case=False)]

2. Building Job Title Embeddings

from sentence_transformers import SentenceTransformer

model = SentenceTransformer('all-mpnet-base-v2')
job_titles = df['job_title'].tolist()
embeddings = model.encode(job_titles)

3. Job Title Standardization

Use this dataset as a reference for standardizing job titles in your organization or application.

Limitations

• Language: English only (though source data includes multilingual options)
• Geographic bias: Stronger coverage of European, U.S., and Australian job markets
• Temporal: Reflects job titles as of 2025; emerging roles may not be included
• Granularity: Some highly specific or niche job titles may have been merged during deduplication

License

This dataset combines data from multiple sources, each with their own licensing:

• ESCO: European Union Public License (EUPL)
• O*NET: Public domain (U.S. Government work)
• OSCA: Creative Commons Attribution 3.0 Australia

Please review the original source licenses for commercial use.

Citation

If you use this dataset in your research or applications, please cite:

@dataset{jobs_dataset_2025,
  author = {Greg Priday},
  title = {Comprehensive Job Titles Dataset},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/gpriday/jobs}
}

Acknowledgments

This dataset builds upon the excellent work of:

• European Commission (ESCO)
• U.S. Department of Labor (O*NET)
• Australian Government (OSCA)

Contact

For questions, suggestions, or contributions, please open an issue on the dataset repository.