After about two decades, the American Numismatic Society has finally turned the lights off in its FileMaker Pro database. The last two departments, Medieval and Modern European were migrated into CollectiveAccess about one month ago, although not in an entirely completed form--the places, people, organizations, dynasties, and denominations still require further normalization by curator, David Yoon, over the next year before these entities are merged into the relational database system. Nevertheless, FileMaker Pro cataloging has ceased, marking a major step in a once-in-a-generation task of normalizing and migrating systems.
The ANS database system was a bespoke DOS-based system pioneered in the 1980s called PRIME. The ANS was among the first major numismatic collections to implement a database for cataloging, although it was never designed as a relational database (a technology which did exist at the time). The database was ostensibly a flat "spreadsheet" data entry system with more than 100 fields, most of which were uncontrolled free-text fields. Before the World Wide Web, curators never envisioned their catalog would become public, and quality and completeness were inconsistent across departments as curators developed their own idiosyncratic data entry processes.
At some point during the 2000s, this PRIME system was picked up and moved into FileMaker Pro, with little to no intermediary normalization. Indeed, many MSDOS special characters (non-ASCII and non-UTF-8) for tabs or other such breaks were retained within FileMaker. Since its inception in 2011, many thousands of lines of PHP code were necessary to transform the ANS' cataloging data into something generally usable on the web with Mantis. The code became more and more complex as we sought to reconcile type or hoard reference patterns to URIs in OCRE, Hellenistic Royal Coinages, or Coinhoards.org. The FileMaker->Mantis publication process was a house of cards.
In 2015, we decided that migrating from FileMaker Pro was a necessity and decided upon the open source collection management software, CollectiveAccess. CollectiveAccess was built on PHP and MySQL, with a fully customizable metadata entry system that could accommodate the significant complexity our curators require. Over several years, we customized and tested a numismatic data entry system for CollectiveAccess, but it wasn't until January 2023 that we officially began the data cleanup and migration process. Significant time and effort was invested over two years in reconciling entities to URIs in Linked Open Data controlled vocabulary systems--not only Nomisma.org, but also Wikidata and Geonames. Most ancient places in CollectiveAccess align to Nomisma and most modern ones align to Geonames. I am not certain of the exact percentage, but there is significant overlap between the rulers, kingdoms, artists, dynasties, issuers, etc. in our numismatic collection and Wikidata. This reconciliation will enable better quality and more complete query of our collection.
I myself worked on Medals and Decorations, as well as Greek, Roman, and Byzantine. Sami Norling, now of the Smithsonian Institution, helped tremendously with North and Latin American, Islamic, and South Asian. David Yoon completed East Asian several months ago and continues to iron out loose ends in Medieval and Modern.
Once these data have been migrated into a proper relational database system, the difficult and time-consuming work of normalization and reconciliation has already been completed. Going from PRIME to FileMaker Pro was kicking the can down the road and building a sustainable curatorial database. If we migrate from CollectiveAccess to another system in another 10 or 20 years, it will be much easier to migrate rigorously curated relational data than the free text fields that made up the ANS database for the last 40 years. This is a significant milestone in the ANS's technological history.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.