Skip to main content

Lightcast Skills: Taxonomy, Methodology and Skills v10

This article covers how the skills taxonomy is structured, what changes in version 10, and how skills are classified in English and non-English (Global) job postings and profiles.

About the Lightcast Skills Taxonomy

Lightcast Skills is a taxonomy of over 34,000 skills, built from hundreds of millions of job postings, resumes and online profiles. Lightcast defines a skill as any attribute an employer may use to describe the requirements for a job, or a candidate may use to stand out from other candidates.

Skills act as a common language for the labor market. They give granular detail, can be compared across countries and languages, and show where demand is heading.

New skills are added to the taxonomy by our in-house taxonomists, who research emerging and outdated skills and review customer requests.

Structure

The taxonomy has three tiers:

  • Category: broad groups of skills by industry or functional specialism. From v10 this includes Artificial Intelligence and Machine Learning.

  • Subcategory: groups of similar skills used to perform a particular aspect of a job, for example ‘Large Language Models’.

  • Skill: the individual skill, a unique concept describing for example ‘Deep Learning’ or ‘Python (Programming Language)’.

Skill types

As well as the taxonomy hierarchy, we organize skills into different types:

  • Common Skills, which have wide purchase across the labor market, such as Communication or Problem Solving.

  • Specialized Skills, such as ‘Java (Programming Language)’ or ‘Financial Analysis’.

  • Certifications and Licenses, such as ‘CompTIA Security+’.

Spoken Languages and Software skills are also included and can be filtered. See Skills for more detail.

Skill descriptions

Most skills have a description, written from our taxonomists’ research with the help of AI tools. Sources including Wikipedia and Wiktionary are also used as references.

Skills Methodology and Classification

In the Lightcast Skills Taxonomy, each skill has one official name. Our English language models also recognize aliases, acronyms, abbreviations and historic names when classifying skills.

  1. The posting or profile text is segmented and tokenized, removing extra characters such as punctuation and line breaks.

  2. The model scans the text for word sequences that indicate a skill.

  3. The surrounding context decides which skill is meant. For example, "AWS" could mean American Welding Society or Amazon Web Services.

  4. A confidence score is assigned, and only predictions above our accuracy thresholds are shown.

Lightcast Skills Taxonomy is also classified and available in a range of other languages, fully explained in Global Skills.

Updates and release cadence

Lightcast Skills is refreshed every 4 weeks. Each refresh can add new skills, remove or consolidate old ones, and update coding and terminology.

  • New skills are added every month, and we welcome input from industry experts.

  • Consolidation is ongoing. Nearly 500 skills were consolidated over several years before v10, and smaller consolidations will continue.

  • Descriptions arrive one release after a new skill's name, so expect a one month delay before a new skill has a description.

  • All changes, including the full v10 consolidation list, are published in the changelog.

What's new in Skills v10

Version 10, released soon, is the 10th major release of the Lightcast Skills Taxonomy and brings two major immediate changes to the user experience, while preparing the ground for further innovations in 2027.

1. A new Artificial Intelligence and Machine Learning category

AI related skills were previously spread across several categories. From v10 they sit in one category with 12 dedicated subcategories, so you can find and analyze AI skills directly in Analyst or the API. New AI skills will be added to this category as they emerge.

The subcategories are:

  • AI Copilots and Chatbots

  • AI Strategy, Governance and Ethics

  • AI and ML Infrastructure

  • Applied Generative AI

  • Autonomous Agents and Reinforcement Learning

  • Computer Vision and Image Processing

  • Deep Learning

  • Large Language Models

  • Machine Learning

  • Natural Language Processing

  • Speech and Voice Technology

  • Symbolic AI and Optimization

2. Consolidation of overlapping skills

We have merged over 1,000 skills into existing skills so that each skill represents one distinct concept. For example, Deep Learning Methods is now part of Deep Learning. You will now find all postings under a single skill, rather than needing to know about both.

Overlapping skills rarely appeared in the same posting, so the average number of skills per posting or profile will remain consistent. The data will give clearer, more consistent signals. The largest consolidations are listed below, and the full list will be published in the Skills Taxonomy Changelog

Consolidated skill

Now part of

Restaurant Management

Restaurant Operation

Virtualization

Virtual Computing

Power Distribution

Electric Power Distribution

Dental Hygiene

Oral Hygiene

Dementia Care

Dementia

Construction Project Management

Construction Management

Nursing Home Care

Nursing Homes

Laboratory Techniques

Laboratory Procedures

Animal Welfare

Animal Care

Safety Codes

Safety Standards

Farm Equipment

Agricultural Equipment

Patient Coordination

Care Coordination

Water Resources

Water Resource Development

Patient Care Management

Care Management

Child Safeguarding

Child Protection

Family Counseling

Family Therapy

Robotic Systems

Robotics

Chinese Language

Mandarin Chinese

Basic Writing

Writing

Related articles

Feedback and skills suggestions are welcome please email [email protected]
​

Did this answer your question?