Synthetic Onboarding — Employee Handbook Acknowledgment Data

What this document is

The handbook acknowledgment is the page a new hire signs to confirm they have received and read the employee handbook, the code of conduct, the anti-harassment policy, the IT acceptable use policy, the confidentiality agreement and the safety rules. It is a short attestation: a column of policy checkboxes, a printed name, a signature and a date. In a dispute it is the document an employer reaches for first.

Why generate synthetically

Every onboarding packet contains one, every HR-tech platform must recognise and file one, and almost nothing about it helps a model do so. It is mostly printed policy text with a handful of marks — which makes it the sparsest classification and extraction target in the packet, and the one where a classifier trained on dense forms most often fails. Generating it in volume is how you find out whether your document router can identify a page by its layout and its few marks rather than by field density.

What makes synthetic data useful

Each acknowledgment belongs to a coherent simulated new hire, carrying the same printed name and signature that appear across their other onboarding documents. The corpus supplies the executed state of the form — every policy acknowledged, signed and dated — which is the state an HR file is supposed to contain and the state an audit expects to find. Paired with the rest of the packet, it gives a document router a realistic mix of dense and sparse pages arriving together in one upload.

Training challenges

Very few marks on a page dominated by printed paragraphs is a hard classification problem: there is little to key on and a lot of text that resembles every other policy document in the packet. It arrives in the same upload as a multi-page federal eligibility form and a two-page beneficiary designation, and a pipeline that mislabels it files a legally significant attestation in the wrong place. The signature, printed name and date sit in a compact block at the foot of the page, so signature detection has to work against a small region rather than a dedicated signature panel. And the acknowledgment checkboxes are marked on every document in the corpus, which means a detector trained here sees the checked state only — a genuine limitation, not a feature.

Who uses this data

HR-tech onboarding suites that must file each packet document to the right slot, HRIS and document-management integrators building classification for mixed uploads, and compliance teams automating audits that ask whether every employee file contains a signed acknowledgment.

Document complexity profile

The shortest form in the catalog: six policy acknowledgment checkboxes plus a printed name, a signature and a date on one page, with one label-to-value relation per field. 1 arithmetic binding and 2 function calls, at maximum expression depth 2. Nearly all of the page is printed policy text carrying no extractable value, which inverts the usual difficulty — the challenge is recognising the document, not reading it.

Key stats from our synthetic corpus

Quantitative characteristics of the Onboarding — Employee Handbook Acknowledgment documents our generator produces.

How this document co-occurs with others

Rates at which identities in our corpus that produce a Onboarding — Employee Handbook Acknowledgment also produce other documents.

Rates above are corpus-derived: they were computed over the 641 employment-eligible identities inside a local synthetic corpus of 1,000 identities generated by SymageDocs' World Simulation Engine at seed 20260421 — the shipped definition gates onboarding forms on active employment. Field and relation counts and the always-checked acknowledgment state come from the shipped handbook acknowledgment definition in the SymageDocs form library. No real employee data was used at any stage.

Frequently asked questions

What data format do synthetic handbook acknowledgments include?

Each generated identity produces a rendered PDF plus a structured JSON annotation file with bounding boxes, field types, and ground-truth values for all 9 fields — six policy checkboxes plus a printed name, signature and date — with one label-to-value relation per field. COCO, YOLO, FUNSD, and BIO/NER exports come from the same job.

Are any of the acknowledgment boxes ever left unchecked?

No. Every policy box is marked on every generated document, because the corpus models the executed form an HR file is meant to hold. If your use case is detecting an incomplete acknowledgment, this corpus supplies only the complete class and you should treat that as a stated limitation.

Why generate such a simple document?

Because simple documents break document routers. A classifier tuned on dense forms keys on field density and label vocabulary, both of which this page lacks — and it arrives in the same upload as forms that have plenty of both.

Does it match the rest of the packet?

Yes. The printed name and signature belong to the same simulated employee who appears on the personal information sheet, the I-9, the W-4 and the direct deposit authorisation, so the packet reads as one person's file.

Can I use this data commercially?

Yes. Every name, signature and date is synthetic, contains no real personal data, and is licensed for commercial use including model training, benchmarking, and redistribution inside your own products.