Open research · release 2026-07-v2
Resume Parsing Lab
See the document, the extracted text, and the failure — side by side.
A versioned synthetic corpus run through the same PDF and DOCX text extractor used by ResumeGet uploads. Every fixture, raw output, result label, method, limitation, and checksum is public.
- PDF and DOCX files
- 20
- Synthetic layouts
- 10
- Risk patterns
- 11
- Published
- Jul 26, 2026
Research boundary
What this release does and does not show.
Tests one real pipeline
Each file runs through ResumeGet's production PDF and DOCX text extraction function. Results are observed, not estimated.
Uses synthetic data only
Names, employers, institutions, contact details, URLs, and achievements are fictional and reserved for testing.
Does not certify an ATS
The lab does not reproduce, rank, or guarantee the behavior of any third-party hiring system or employer configuration.
Observed output
All results, including the failures.
A result checks ten disclosed markers and their order. “Preserved” is not a universal compatibility claim; it means only that this extractor found those markers in sequence for this release.
14
Preserved
1
Reordered
5
Content loss
0
Unreadable
Showing 20 results
baseline-single-column.pdf
Baseline single column
A conventional one-column resume with contact details in the main document flow.
Fixture titles, risk patterns, and raw files remain in English so every localized page points to the same technical evidence.
- Markers found
- 10/10
- Characters
- 1,355
- Risk patterns
- single column baseline
Observed issues
No disclosed marker was missing or out of order in this observation.
Result method
Four bounded labels, no mystery score.
01
Preserved
Every disclosed marker appears in the expected order.
02
Reordered
Every disclosed marker appears, but at least one is outside the expected reading order.
03
Content loss
Extraction completes, but at least one disclosed marker is missing.
04
Unreadable
The production extractor rejects the file or returns fewer than 120 characters.
Release files
Inspect or reproduce every claim.
The release is available in human-readable and machine-readable formats. SHA-256 checksums cover every published file except the checksum list itself.
For career educators
Run a 45-minute parsing workshop without collecting participant resumes.
The teaching kit uses paired synthetic fixtures to compare visual hierarchy with extracted reading order. It includes accessible text equivalents and a canonical URL beside the QR code.
- Timed agenda and learning objectives
- Paired rendered-file versus raw-text exercise
- Facilitator, privacy, and accessibility notes
- Four fixture pairs ready for a group session
Reuse and citation
Use the material. Cite the release, not a promise.
The written materials and synthetic fixture content are available under CC BY 4.0. Plain-text attribution is enough; a hyperlink, reciprocal link, registration, payment, endorsement, or favorable result is never required.
ResumeGet Resume Parsing Lab, release 2026-07-v2, https://resume-get.com/research/resume-parsing-labRead reuse terms
Related paths