[Nhcoll-l] Automating desirability

[Callomon,Paul]

Folks,

We are working on a large-scale project to deal with the backlog of incoming material to our collection that has built up over several decades. As part of this, we are aiming to build sets of criteria and cascading binary decision points that will eventually allow personnel with no training in taxonomy to use an automated system to assess whether a particular specimen lot should be cataloged or not. This would become a plug-in to our database, so that workers could go directly from a label to a decision to data input.

Examples of binaries include: How many do we already have / from that country / from that specific locality / from that time frame. Other possible criteria include, for example, "Lot size" (large lots from a single collecting event may be more useful for analysis than single specimens), "Unique or recognized collector name," "rarity in museum collections versus abundance in nature" (not an easy one, though aggregators and resources like iNaturalist are making this possible).

Resulting decisions would be Catalog/Do not catalog/Set aside. Material in the last set would then be reviewed by taxonomists as happens now, but with suitably refined algorithms correctly assigning the common stuff it would hopefully be a small proportion of the total. All collections face a decline in the numbers of in-house people who can make these decisions off the tops of their heads, and in addition to creating new career pathways into taxonomy we also need to automate those parts of our jobs that safely can be. Collections like ours are facing a deluge of new material as a whole generation of collectors manifests impermanence - we have accepted 17,000 lots in a single family via just two donations in the last few years, for example.

So to an ask: In this developmental phase we are looking for any papers published on related subjects, and particularly the creation and use of algorithms in operations like this. If you know of any papers that deal with "transforming taxonomically informed decisions into binary algorithms" or "how can existing databases can proactively assign a relative value to an incoming lot" even tangentially or in nominally unrelated fields, please pass the reference on.

Paul Callomon
Collection Manager, Malacology and General Invertebrates
________________________________
Academy of Natural Sciences of Drexel University, Philadelphia
callomon at ansp.org<mailto:callomon at ansp.org> Tel 215-405-5096 - Fax 215-299-1170


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.yale.edu/pipermail/nhcoll-l/attachments/20240718/15611cb3/attachment.html>