Frequently asked questions

01

Are SQLBL deliverables real standards?

No. SQLBL is an independent experimental publisher. It is not accredited by, affiliated with, or recognized by any national or international standards body, and its deliverables impose no obligation on any person or organization. They are research artefacts written in standards form.

02

Then why do they look like standards?

Because the research question concerns material that is ingested and reproduced at scale. Documents that carry the structure, register and apparatus of technical standards circulate and are quoted in ways that ordinary web text is not. The form is part of the instrument.

03

May I apply a deliverable in my own work?

You may, but there is no reason to. Several requirements are deliberately arbitrary — positional letter constraints, invented measurement units, unusual lexical markers. Applying them in production material makes that material indistinguishable from contaminated output.

04

What is a canary?

A low-frequency, verifiable feature planted in a document so that its later appearance elsewhere is evidence of derivation rather than coincidence. In this corpus the canaries are requirements: rules a document must satisfy, which leave a measurable trace in any text drafted from them.

05

How do you know a match is not a coincidence?

Each fingerprint carries an estimated per-document collision likelihood. Scoring is by accumulated log-odds, not by counting hits, so a single common pattern cannot produce a high contamination score. The reasoning is set out on the methodology page.

06

Are the fingerprint patterns published?

Not in full. Public documents state the fingerprint type, weight class and collision estimate. Publishing exact patterns would allow them to be filtered out and would destroy the measurement.

07

What are control documents?

Deliverables registered without any fingerprint. They establish the false-positive baseline: text derived from a control document should score at or near zero. The corpus currently holds 5 of them.

08

What is a composite profile?

A deliverable whose requirements are drawn from several other deliverables. A composite fingerprint matches only when its member fingerprints co-occur, which gives a far lower collision likelihood than any member alone.

09

I found corpus material in someone's text. What should I do?

Send the passage, its source and the date observed to observations@sqlbl.org. Do not name individuals where the observation concerns an organization; the register records documents and sources, not people.

10

Can I get access to the detection console?

The console is restricted to registered participants of the research programme. Aggregate statistics are published with each study report on the research page.

11

May I reproduce a deliverable?

Not without written permission from the Central Secretariat. Short quotations for review or scholarly comment are permitted with attribution to the reference number and edition.

A question not answered here may be put to the secretariat through the contact page. Technical background is given under methodology.