A numismatic collection almost always arrives already inventoried: each banknote, each coin and each share certificate takes one row of a spreadsheet that someone built piece by piece, with the issuer, the series, the number, the date, the denomination and the place where it is kept. Janium Collect catalogs from that inventory. It reads the sheet row by row and produces one record per piece in ISAD-G, or in whatever format the institution uses, without anyone retyping what is already captured.
This is how a collection of Mexican banknotes from the 19th and 20th centuries was cataloged, from a CSV file with sixteen columns and without a single image of the pieces.
What each row contains
A row in that inventory says, for example, that the piece is a 5-peso banknote from the Banco de Aguascalientes, series C, dated July 1, 1910, and kept on shelf G. These are values the specialist determined by examining the piece, and the reference code, the denomination, the currency, the series, the number, the issuer, the date and the location pass into the record exactly as they appear in the sheet. How the columns are mapped to fields and what happens with empty cells is explained in Cataloging from lists and spreadsheets.
What the model builds from those values
With the row and the collection’s profile, the language model writes up the record. The title follows a fixed template (denomination, currency, issuer, series, number and date), so that all the banknotes in the collection are titled in the same form and can be sorted and compared. The date that the inventory noted as “1910 07 01” is written as “July 1, 1910”, and the extent is formed from the number of pieces and the denomination. The fields that hold the same value for the whole collection, such as the level of description or the access conditions, are set by the profile so that they do not vary from one record to another.
The context of the issue
The inventory does not say what kind of bank issued the banknote, in what period it circulated or what was happening to the currency at the time, and that context is what Collect adds to the scope and content from the model’s knowledge. In the banknote collection, the record for an issue from the years of Victoriano Huerta explains that the mass issue of notes without metallic backing contributed to the inflationary crisis of that period, something that no column of the sheet contained. Today that field is marked in the record’s provenance as generated text, in a category distinct from that of the fields that came from the row, so that whoever reviews knows which part of the record the model wrote.
From the list to the series of each issuer
In ISAD-G, the collection is described as a fonds with a subfonds of banknotes, a series for each issuing bank and each banknote as an item. For each row to fall into its series, the inventory needs a column with the issuer’s code; without it, Collect produces the item-level records but cannot group them by issuer. In the banknote collection that column was added to the sheet before processing, with a short code for each bank, and with it each piece ended up in the series of its issuer.
Coins, bonds and shares
Coins, bonds and shares are cataloged the same way, each material with its own profile. For coins, the inventory contains the mint, the metal and the date range, and the code of each piece is formed from the mint. For bonds and shares, each certificate is described as a document and not as an item, because it is a paper with text and signatures as well as a collection object.
Where to review
Each record is scored against the archival profile, which requires the code, the title, a date with its year, the level of description and a scope and content with enough text. A record that does not reach the level the institution asks for carries a review note, and whoever reviews the collection starts with those records. The score and the review notes are explained in Why trust a catalog made with AI.
Limits in a numismatic collection
The starting inventory. Collect works with what the sheet says, so if the inventory does not distinguish a variant or noted a date wrongly, the record repeats that value until someone corrects the sheet. The values that take a specialist and that the inventory does not contain, such as a numismatic catalog reference or the condition grade, stay out of the record as long as nobody captures them.
The historical context. The context paragraph comes from the model’s knowledge and is not checked against a source, which is why whoever knows the history of the collection should read it first, especially in periods with many issues that resemble one another.
The image of the piece. When cataloging from the sheet, each row produces a text-only record. If the collection has been photographed and what can be read on the piece should enter the record, the process starts from the images and the inventory comes in as context, matched by file name, as in a periodicals collection with an inventory and scans.
To continue the conversation
If your institution has a collection of banknotes, coins, bonds or shares that is already inventoried and you want to bring it into the catalog without capturing it again, write to us at info@janium.com. We would like to see which columns your inventory has and how you describe each piece today.