See what you can build with JAAT

From job postings to workforce information with open tools

Open tools for mapping descriptions of work. Find the tasks that connect occupations, compare taxonomies, and inspect the data and code.

Explore occupations Build with JAAT on GitHub Try JAAT.app

WORKFORCE INFORMATION AUTOMATION
WITH OPEN TOOLS

Tasks connect work within and across occupations

40–54% higher recovery of observed occupational transitions

Occupations that share tasks offer a map of related work. Using tasks found in job postings improves recovery of observed occupational transitions over the established workforce information product (O*NET).

The comparison uses equally long recommendation lists in two resume and career-history benchmarks.

Explore occupational connections

Published RN rankings. Task commonality invites investigation of licensing, credentials, experience and other conditions of a move.

How we represent work matters

Explore task networks for Registered Nurses and other occupations. See which tasks occur together within jobs and how shared tasks connect occupations.

Explore the RN task network

See the broader task profile →

Registered Nurses

Six released task examples. Lines join tasks observed in the same postings; thicker lines mean more shared postings. The outer ring marks an O*NET-linked task. Labels are shortened.

This illustrates tasks performed together. The relatedness model compares task-frequency profiles across occupations.

External validation: compare with observed transitions

External validation

How well do the models recover observed occupational transitions? Compare recommendations with observed transitions in resume and career-history data. Additional survey comparisons are available on the benchmark page.

Explore benchmark results

● Posting tasks● Published O*NET

Mean capture / recoveryOrigins with greater recovery

SST

15.7%

11.2%

EMSI

27.4%

17.8%

Equal-length comparisons across all evaluated origins. EMSI measures ranked-destination recovery; other rows capture transition weight. These are historical comparisons, not estimates of hiring effects.

Tasks and work activities: compare the occupational assignments

Explore existing connections. Build your own with the data and open tools.

Choose an occupation to explore Try JAAT.app Build with JAAT on GitHub

Extensions: build on a common language of work

Incident-linked tasks

Follow OSHA incident narratives to 610 task IDs, then explore their occupational prevalence alongside fatality and injury rates.

Download 610 tasks (CSV) · Evidence & code

Military titles

Tasks extracted from military job descriptions connect military titles to civilian occupation-task profiles; compare the recommendations with established crosswalks.

STEM and green tasks

Group tasks into reusable bundles and explore where they appear across occupations. Track DEI keywords, green tasks and STEM tasks in federal postings over time.

STEM tasks (CSV) · Green tasks (CSV) · Review files

Profiles

Compare the work and requirements advertised in federal postings by occupation or agency.

Federal posting trends

Track DEI keywords, green tasks and STEM tasks per 1,000 active federal postings, including the January 2025–March 2026 comparison.

Data and code

Posting-task model and released scores · Reproduce the benchmark results · Compare alternative representations · Excluded review pairs

NLx posting records are not redistributed. Public reproduction resources identify which steps use released scores, illustrative inputs or restricted data.

Get Building

Open tools let institutions and researchers build workforce information from their own descriptions of work. Use JAAT to create comparable data, extend existing classifications, and test new applications.

Explore JAAT on GitHub →

Try JAAT.app · Download the starter notebooks

Model weights and public datasets · Demonstration notebooks · Aggregate-data deposit

Python · JAAT import output

>>> import JAAT

+--------------------------------------------------------------+
|       ___           ___           ___           ___          |
|      /\  \         /\  \         /\  \         /\  \         |
|      \:\  \       /::\  \       /::\  \        \:\  \        |
|       \:\  \     /:/\:\  \     /:/\:\  \        \:\  \       |
|   ___  \:\  \   /::\~\:\  \   /::\~\:\  \       /::\  \      |
|  /\  \  \:\__\ /:/\:\ \:\__\ /:/\:\ \:\__\     /:/\:\__\     |
|  \:\  \ /:/  / \/__\:\/:/  / \/__\:\/:/  /    /:/  \/__/     |
|   \:\  /:/  /       \::/  /       \::/  /    /:/  /          |
|    \:\/:/  /        /:/  /        /:/  /     \/__/           |
|     \::/  /        /:/  /        /:/  /                      |
|      \/__/         \/__/         \/__/                       |
|                                                              |
|                    Welcome to JAAT v1.1.3                    |
| © 2025-2026 Loyola University Chicago. All rights reserved.  |
|       An NLP-powered toolkit for job ad text analysis.       |
+--------------------------------------------------------------+

   ✅ Loading JAAT...... Ready.

USAJOBS worked example: coded postings → occupation-task profiles → relatedness scores → a pipeline and validation figure. View the figure and results. Input requirements.

Starter notebooks: follow each step

JAAT records to occupation-task profiles — requires the coded postings described in the README.
Profiles to relatedness scores — includes aggregate inputs.
Relatedness scores to the worked figure — includes aggregate inputs.

The notebooks begin with saved JAAT codes. They do not rerun extraction from raw posting text.

Reproduce the manuscript results

National model: code, released scores and inputs. Rebuilding profiles requires authorized NLx inputs.
Benchmark calculations in Python and Stata · Summary results (CSV)
RN figure data and code
Alternative representations · ·

Review decisions and methods

Page index

Results, demonstrations and resources described in the paper and SI Appendix.

Results

Related occupations: Explore occupations, shared tasks and related-occupation recommendations.
Benchmark comparisons: Compare mean recovery, occupation counts, list lengths and filters.
Tasks, activities and skills: Compare the representations described in the SI.
Manuscript figure and inputs: Inspect the RN example and download its figure resources.

Demonstrations

Incident-linked tasks: Follow OSHA narratives to the 610-task bundle and occupational comparisons.
Injury comparison in SI Table S13: Compare task, activity and work-context measures on common samples.
Military titles: Connect military job descriptions to civilian occupations.
STEM and green tasks: Inspect and download the reusable task lists.
Federal occupation profiles: Explore attributes of federal postings by occupation on the Evidence dashboard.
Federal agency profiles: Explore attributes of federal postings by agency on the Evidence dashboard.
Federal posting trends: Follow DEI keywords, green tasks and STEM tasks over time.

Code, data and methods

USAJOBS notebooks, database and results: Repeat the workflow using public federal-posting data.
Methods and resource guide: Find the code, source data and construction methods.
JAAT toolkit: Use the open Python toolkit.
Models and datasets: Find model weights and public datasets.
Aggregate-data deposit: Access the versioned data deposit.

Website layout and explanatory text were prepared with generative AI (Codex). Data sources, model methods and code.

The National Labor Exchange (NLx) Data Trust bears no responsibility for the analyses or interpretations of the data presented here. The opinions expressed herein, including any implications for policy, are those of the authors and not of the NLx Data Trust members or others.