Introduction
You have a shield in front of you — a gravestone, a signet ring, a bookplate, a stained-glass window — and you want to know whose it is. Enquerre identifies it from what you can see: describe the shield, even partially, even with doubts, and the engine checks it against some 135,000 blazons gathered from the classic armorials — Rietstap, d’Hozier, Siebmacher, Mailhol, Rolland — and the French and Dutch Wikipedias. A computational armorial, leafed through by description rather than by name.
The name comes from the arms à enquerre: those that deliberately break the rule of tinctures, inviting the viewer to enquire into their history.
How it works
Describe the shield
On the Identify page, build the description field by field — field tincture, ordinaries, charges — or type a blazon directly in French. No need to be complete or perfect: every detail is a clue, not a constraint.
Mark your doubts
The “Uncertainties” panel tells the engine what is fragile: “No tinctures” for an engraving or a seal, counts that are hard to read, forgotten positions, a charge perhaps misidentified — a lion taken for a dog. Those traits are relaxed; the others stay firm.
Read the shortlist
Candidates arrive grouped by match quality, with chips justifying each placement: whole shield, field by field, Renesse attestation, region. Open a record to weigh the original blazon against your memory.
Two real examples
An engraved bookplate, so no colours: a chevron, two stars above it, a hunting horn below.
- 1. Describe — a chevron (ordinary), two stars “en chef”, a huchet (the hunting horn) “en pointe”. Positions matter.
- 2. Doubt — no tinctures known: leave them all on “Any / unknown”; the search will compare structure only.
- 3. Read — the arms of Ciel (Saône-et-Loire) appear in the shortlist.
A shield you know to be Breton: on a blue field, a black anchor and a silver building — tower or castle?
- 1. Describe — field azure; an anchor sable; a tower argent.
- 2. Doubt — add “château” as another possible reading (“+ or another charge”) and filter on Brittany: the filter lifts Breton arms without discarding those whose region is unknown.
- 3. Read — Damgan (Morbihan) appears in the shortlist, region chip attached.
These two examples belong to our test suite: any change to the engine must keep passing them before it ships.
What is inside
Seven corpora, merged into one searchable database:
| Source | Contents | Coats of arms |
|---|---|---|
| Rietstap | Armorial général (1884–1887) — families from all over Europe | ≈ 80,000 |
| Wikipedia (fr) | Municipal and family arms, with sourced blazons | ≈ 34,000 |
| Rolland | Supplément à l’Armorial général, vols. 1–2 | ≈ 8,400 |
| Siebmacher | The great German armorial (selected volumes) | ≈ 7,500 |
| Mailhol | Dictionnaire historique et héraldique de la noblesse française | ≈ 2,900 |
| Wikipedia (nl) | Dutch municipalities, provinces and water boards | ≈ 1,800 |
| D’Hozier | Armorial général de France (1696) | ≈ 1,500 |
Coverage is densest for France, the Benelux and the German lands, and thins out beyond them. Rounded figures — exact state of the corpus →
Known limitations
- Blazons are entered in French. The interface speaks English, but the engine does not (yet) understand a blazon written in English. Describe the shield with the builder — its menus feed the engine directly — or type the blazon in French.
- Parsing is not always complete. Nearly three quarters of the Wikipedia blazons are fully parsed; Rietstap’s abridged entries are mostly parsed partially — tinctures and charges spotted, structure incomplete — and surface less precisely.
- Some OCR noise remains. The Rietstap entries come from optical character recognition: typos survive in some texts.
- The drawings are interpretations. The shields shown are generated automatically from the blazon — diagrams, not authoritative reproductions. Only the text is binding; compare with the Wikimedia image where one exists.
- A shortlist of candidates, not a certificate. Unrelated families sometimes bear identical arms. A result is a serious lead, to be checked in the cited source — not a guaranteed attribution.
The right arms are missing?
Tell us by email: what you were looking for, what you entered. Every report becomes a permanent test case — the query, mistakes included, joins the regression suite, and any change to the ranking must pass it before it ships.
Write to: contact@enquerre.app