A periodicals collection brings together serial publications, such as a newspaper that came out every day for decades, a monthly magazine with hundreds of issues, or the supplements and special editions that break the sequence. Each issue says little on its own, and what makes it useful for research is the title, the date, the number and the place it occupies in the series. Janium Collect catalogs this material from the inventory the collection already has, from the scans or from both at once.
Almost no periodicals collection starts from zero or has everything in the same state. One part is usually in a spreadsheet made years ago, with title, year and holdings, and another exists only as microfilm scans or PDFs of loose pages, with no description beyond the file name.
From the inventory
When there is a list, Collect takes each row and turns it into a record. The language model receives the row with the collection’s profile and writes it up in the institution’s format: it expands abbreviations, places each value in its field and can add what it knows about the publication, such as its frequency or the context of an era, marked as coming from the model. The earlier work of whoever made the list is used in full and nobody retypes it. How the columns are prepared and what happens with empty cells is covered in Cataloging from lists and spreadsheets.
From the scans
When there are only scans, each page is converted to an image and a vision model reads it, so the title, the date, the edition number and the place come from what the page itself prints. Scan quality governs here: yellowed paper, smeared ink or low-contrast microfilm give partial readings, and the same newspaper scanned under two different approaches gives two different results.
When there is both an inventory and a scan
Often both sources exist for the same issue, and they do not always say the same thing. Collect lets the institution decide which one carries more weight in each collection. In one, the inventory is the reference: its value is kept and, if the printed page says something else, the difference is noted in the record for someone to resolve. In another, where the inventory is known to carry copying errors, the model can take the title, the date or the number from the page and correct what the list had. The institution chooses the policy for each collection before processing.
ISAD-G with the hierarchy or MARC21 as a serial
The format is also decided before processing, because Collect extracts directly into the chosen format.
In ISAD-G, the periodicals collection is described as an archive, with the fonds, the sections, the series by publication and each issue as an item. Collect builds the tree from the location each record carries, calculates the date range of each series from the dates of its issues and moves up to the higher level the values that nearly all the children share, such as the creator, the language or the geographic coverage. The threshold is high (nine out of ten) so that what holds for only part of a series is not attributed to the whole series. The full mechanism is in From loose documents to Fonds and Series.
In MARC21 or UNIMARC, each title is described as a serial, with its ISSN, its frequency and the numbering the collection covers. This is the format to use when the periodicals collection lives inside a library’s catalog. The details are in How to generate MARC21 and UNIMARC records with AI.
Where to review
Review starts with the records that carry a review note, which are those with the lowest quality scores. In a periodicals collection, records that come from a well-kept list usually call for a spot check, and those that come from poor scans call for a look at the date and the number.
The institution can also ask for some fields, such as the date, to be checked against the text of the page. With that rule active, a value the model proposed that the page does not support lowers the score and is flagged. The rule is activated per institution and per field, because in some collections an inferred date is acceptable and in others it is not.
Limits in a periodicals collection
Legibility. In old newspapers, what a poor scan most often lacks is the date and the number. If the list exists, it fills them in; if it does not, what the model proposes for those fields arrives marked as AI-made, and someone who knows the collection confirms it.
The source structure. The ISAD-G hierarchy is built from the location of each issue in the collection’s classification scheme. Without that location, Collect produces the issue-level records and cannot build the levels above them.
The chosen format. A batch extracted in ISAD-G is not converted to MARC21 afterward. If the collection needs both, it is processed with both profiles.
The size of the collection. The savings are large when there are many issues of few titles, with legible mastheads or already inventoried. In a small, heterogeneous collection, with no lists and with irregular scans, it is best to measure first with a sample.
To continue the conversation
If your periodicals collection has one part in lists and another only in scans, and you want to bring both into the catalog with the structure of the collection, write to us at info@janium.com. We would like to see how your material is inventoried today and in what format you describe it.