Provenance
ភាសាKhmer for “language”
Every item traces back to its source.
Provenance is the part of a corpus that is hardest to add afterwards, so we build it in from the first capture. This page sets out what travels with the data, how we label the figures we publish, and how we count.
What travels with every item
The same four things, in every delivery, whatever the language or the kind of material.
Item record
What the item is, who holds it, its language and date, and when it was captured. People, institutions and places are recorded as typed, consistently spelled entities, so records join to one another across a corpus.
Capture record
How and when each file was captured, with its technical parameters. Documented rather than assumed.
Named agreement
The agreement under which the material was captured, naming the holder and the scope of use granted. Documented, not asserted.
Masters retained
Archival masters are kept permanently, so better derivatives can be issued years after delivery.
How we label figures
Every figure we publish carries one of three marks, so you can tell what we measured from what we were told.
- Verified
- Recomputed by us from the primary artifact.
- Reported
- Published by the institution. Self-reported and unaudited.
- Estimate
- Derived by us under the stated assumptions.
Institutional holdings are self-reported and unaudited, and we label them that way rather than rounding them into something cleaner. Each figure is published with its source, a link to the page it was read from, and the date we read it.
How we count
A collection holds items, not tokens. Getting from one to the other takes adjustments, and we publish each one rather than quoting a headline figure.
- 1
Start from what is held
We begin with the holder’s own counts and say which unit each one uses: titles, volumes or physical items. Where a holder publishes a floor rather than an exact count, we say so.
- 2
Remove the duplicates
A collection holds multiple copies, editions and reprints of the same work. We apply a distinct-title ratio, publish the ratio and the band around it, and publish the query that settles it whichever way it goes.
- 3
Name the tokenizer
The same text yields very different counts across tokenizers. We state the fertility we quote at and recompute against your tokenizer on request, because a token price without a named tokenizer is not a price.
See it applied
Each catalog publishes its own count and the source behind every figure.