# An Encyclopaedia of Bookbinding and the Printing Trades — entry dataset

**5,460 dictionary entries** on bookbinding, book conservation, papermaking and the
printing trades, transcribed from **nine works published between 1683 and 1982** and
normalised into one schema.

Landing page: <https://hibrkraft.com/resources/encyclopedia/>
This directory: <https://hibrkraft.com/data/encyclopedia/>
Machine descriptors: [`croissant.json`](croissant.json) (MLCommons Croissant 1.0) ·
[`datapackage.json`](datapackage.json) (Frictionless Data Package)

---

## Read this first: rights are mixed, and the field says which

| tier | entries | what it is |
|---|---:|---|
| `public-domain` | **1,599** | Eight works published 1683–1906, from Project Gutenberg. Unambiguous. |
| `us-government-flagged` | **3,861** | Roberts & Etherington, *Bookbinding and the Conservation of Books* (1982) — a United States Government publication (Library of Congress) issued openly in electronic edition by Conservation OnLine. |

The 1982 material is reproduced on the understanding that a US Government publication
carries no copyright. **A rightsholder who believes otherwise is asked to write, and
those entries will be withdrawn.** Contact: admin@hibrkraft.com

Every record carries `rights_tier` and the full `source.rights` string. Nothing in this
dataset asks you to guess. If you need an unambiguous position, use
[`entries-public-domain.jsonl`](entries-public-domain.jsonl), which is the 1,599-entry
subset and nothing else.

The **compilation** — the normalisation, the schema, the resolved citations, the
apparatus — is released under
[CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/). That covers the work of
assembling this; it cannot and does not speak for the source texts above.

---

## Files

| file | bytes | what |
|---|---:|---|
| `entries.jsonl` | 8.5 MB | All 5,460 entries, one JSON object per line. **Start here.** |
| `entries.jsonl.gz` | 1.7 MB | The same file, gzipped. The server does not compress on the fly. |
| `entries-public-domain.jsonl` | 3.0 MB | The 1,599 `public-domain` entries only. |
| `bibliography.json` | 60 KB | The 373 works the entries cite by number. |
| `sources.json` | 3.5 KB | The nine source works: siglum, author, year, entry count, rights. |
| `bookbinding-encyclopedia-v2.json` | 7.5 MB | The compiled source file, exactly as it came out of the build. |
| `croissant.json` | 15 KB | MLCommons Croissant 1.0 descriptor. |
| `datapackage.json` | 10 KB | Frictionless Data Package descriptor. |

**JSONL, not the 7.5 MB JSON.** The source file is one JSON object wrapping a
5,460-element array, so reaching the first entry means parsing all 7.5 MB. JSONL costs
the same bytes and streams. It is the format a retrieval or embedding pipeline actually
wants.

---

## Record schema

```json
{
  "id": "dt0058",
  "term": "\"agenda\" format",
  "norm": "\"agenda\" format",
  "sense": 1,
  "kind": "definition",
  "letter": "A",
  "letter_raw": "\"",
  "volume": 1,
  "variant": null,
  "dates": null,
  "text": "A narrow book format, i.e., one where the height of the book is …",
  "chars": 496,
  "is_article": false,
  "figures": [],
  "source": {
    "siglum": "RE",
    "id": "cool",
    "author": "Roberts, Matt T.; Etherington, Don",
    "work": "Bookbinding and the Conservation of Books: A Dictionary of Descriptive Terminology",
    "year": 1982,
    "rights": "US Government publication (Library of Congress), published openly …"
  },
  "rights_tier": "us-government-flagged",
  "reference_numbers": [347],
  "citations": ["Middleton, Bernard C. A history of English craft bookbinding technique. …"],
  "landing_page": "https://hibrkraft.com/resources/encyclopedia/"
}
```

| field | notes |
|---|---|
| `id` | Stable across rebuilds. Use it as the primary key. |
| `term` / `norm` | `term` is as printed. `norm` is folded for joining; it is **not unique** — see below. |
| `sense` | Disambiguates where one source defines a term twice. |
| `letter` | Folded to A–Z. `letter_raw` keeps the source's own first character, which for 12 entries is `"`, `À`, `È` or `É`. |
| `volume` | Which printed volume carries that letter: 1 (A–F), 2 (G–O), 3 (P–Z). |
| `text` | Paragraphs joined by a blank line. Split on `\n\n` to recover them — exact for all 5,460 records. |
| `is_article` | True for the 68 long-form entries that run as articles rather than definitions. |
| `reference_numbers` | The source edition's own numbering into the bibliography. |
| `citations` | Those numbers **already resolved** to full citation strings, so you do not have to carry `bibliography.json` and join. 2,656 entries have at least one; 4,630 references in total. |

### `norm` is not a key, on purpose

**414 terms are defined more than once**, accounting for 654 records, because more than
one source defined them. This is the most interesting property of the dataset and it is
not a defect:

| term | definitions | years |
|---|---:|---|
| `register` | 8 | 1841 · 1856 · 1866 · 1889 · 1890 · 1901 · 1906 · 1982 |
| `signature` | 8 | 1841 · 1856 · 1866 · 1889 · 1890 · 1901 · 1906 · 1982 |
| `bands`, `finishing`, `forwarding`, `gathering`, `inset`, `overcasting` | 6 each | 1841–1982 |
| `pull`, `break` | 3 each | **1683 · … · 1982 — 299 years apart** |

So this is a *diachronic* record of how a craft vocabulary was written down across three
centuries, not a flat glossary. Group by `norm` and sort by `source.year` to see a term
change meaning. Deduplicate by `norm` only if you specifically want one definition per
term, and if you do, say which source you kept.

---

## The nine sources

| siglum | author | work | year | entries |
|---|---|---|---:|---:|
| MOX | Moxon, Joseph | *Mechanick Exercises, Volume 2* | 1683 | 35 |
| DAP | Savage, William | *A Dictionary of the Art of Printing* | 1841 | 753 |
| MAB | Nicholson, James B. | *A Manual of the Art of Bookbinding* | 1856 | 106 |
| AMP | MacKellar, Thomas | *The American Printer: A Manual of Typography* | 1866 | 215 |
| BIB | Sabin, Joseph | *Bib-li-op-e-gis-tic* | 1889 | 39 |
| ABB | Zaehnsdorf, Joseph W. | *The Art of Bookbinding: A Practical Treatise* | 1890 | 137 |
| BCB | Cockerell, Douglas | *Bookbinding, and the Care of Books* | 1901 | 64 |
| NBL | Bailey, Arthur L. | *Notes on Bookbinding for Libraries* | 1906 | 250 |
| RE | Roberts, Matt T.; Etherington, Don | *Bookbinding and the Conservation of Books* | 1982 | 3,861 |

Text is reprinted as printed. No wording, spelling or punctuation has been altered.

---

## The dataset and the books are not the same size

The three printed volumes carry **4,790** entries; this dataset carries **5,460**. The
difference is the 654 duplicate-term records above plus a small remainder: where several
sources define one term, the books print one and the dataset keeps all of them.

The books are at
<https://hibrkraft.com/resources/encyclopedia/> — 1,046 pages, free, and each PDF carries
this same structured record in its own XMP metadata.

---

## Citing it

> Anwar, Ibrahim, comp. *An Encyclopaedia of Bookbinding and the Printing Trades: entry
> dataset*, version 2. Bogor: PT Hibrkraft Kreasi Indonesia, 2026.
> <https://hibrkraft.com/resources/encyclopedia/>

When you quote a **definition**, cite the source work and its year — they are in the
record — not this dataset. The 1841 definition of `register` is William Savage's, and it
should travel with his name on it.

ORCID: [0009-0006-0425-4923](https://orcid.org/0009-0006-0425-4923)