The Janium capture app is for cataloging with the item in front of you, in the stacks, in the reading room or next to the shelf, from a phone or a tablet and also from the browser of a desktop computer. The person cataloging photographs the item and saves the submission, and Janium Collect turns each submission into a catalog record without anyone having to retype what was noted.
It is most useful for material that should not be moved or that is described by walking through the collection, such as a fonds on its original shelving, small pieces in a display case or an inventory taken aisle by aisle. At a desk with a pile of pending books it works the same way: the cataloger photographs each book and moves on to the next.
Two ways to work
The app has two modes, and the difference between them is capture speed.
Photographs only. The operator photographs each book and saves the submission without filling out any form. Collect creates the bibliographic record from the photos, and the holdings (library, location and copy) are created later in the Janium cataloging module, once the record is loaded. In this mode an operator processes about 500 books a day.
With holdings. Besides the photos, the operator fills out a form in the app with the institution’s codes: the material type, the library or collection, the location and the inventory number. With those codes, Collect creates the record and its holdings in a single step. The form takes more time per item, and throughput drops to about 100 records a day per operator.
The first mode suits the case where the urgent task is to get a backlog of books without records into the catalog, and the second suits the case where each copy has to have its location recorded from the start, as in an inventory taken piece by piece.
The photographs
The photos follow a fixed order. For a book, the cover goes first, the copyright page next and the table of contents last, and each step accepts several photos when the item calls for it. Collect keeps that order when it receives the submission, which lets it read the item the way a person turning the pages would.
What Collect does with each submission
Every few minutes, Collect picks up from the app the submissions that have not been processed yet. The app is one of the sources through which material enters Collect, along with folders, the cloud and email, as explained in Where material enters Collect. Each submission goes to the institution it belongs to and to the cataloging profile for that type of material, and its photos are merged into a single multipage document, in the order they were captured and with the orientation corrected.
The language model reads the photos, identifies the item and completes the record with what it knows about it. When the submission includes a form, the typed values reach the model as context and help it identify the item. Each value is marked according to where it comes from, as explained in Collect marks AI-made metadata.
In the mode with holdings, some values do not go through the model. The material type is set in the record exactly as it was chosen in the app, and the library, location and copy codes are written directly into field 852 of the record, which is where MARC stores holdings. This is the mechanism described in Data injection, applied to what the operator entered in front of the item.
What goes back to the app
Once the record is at its destination and its quality score reaches the configured minimum, Collect marks the submission as processed in the app. The person cataloging sees in their own list which items are done and which are still pending, without asking through another channel.
A record that stayed below that minimum is not marked. It keeps appearing as pending in the app, and if the score is very low it also carries a review note in the record itself, so the person who knows the item can review it or photograph it again if the shot was the problem.
Limits of the capture app
The photos. The model reads what the photo shows, and a blurred shot, glare or poor lighting limits what it can identify. The codes from the form arrive complete even if the photo is bad, but the title, the author or the publisher depend on being legible in the image.
Holdings in the fast mode. With photographs only, the bibliographic record arrives without holdings, and the book has no location in the catalog until someone creates them in the Janium cataloging module.
Review. The app shortens capture and avoids transcribing afterward, but it does not replace review by someone who knows the collection: the score and the pending mark show where to start.
To continue the conversation
If your institution has a backlog of uncataloged books, collections that are better described by walking through them or items that should not be taken to a desk, write to us at info@janium.com. We would like to know what material you would catalog with the app and whether you care more about speed or about having each copy’s location recorded from the start.