The slides for my ALA Midwinter presentation on subject maps (which I described in a previous post) are now online. (Yes, I’m playing around with BePress’s Selected Works.) You can also find links to the presentation, a white paper, and demos, from our library lab’s subject maps page.

I’m afraid it’s only the slides (no notes) but I’ll be happy to answer questions or take suggestions. I’m hoping to spend a fair bit of time this semester on subject maps applications for our catalog and digital library architecture, and I’m very interested in seeing how much we can do with them.

During the presentation, someone asked “What do you do when you have more than one subject ontology/thesaurus?” I addressed this question in a talk I gave at last spring’s DLF forum (slides here, though some pictures show up dark). In short, I see four basic strategies one can take. (Warning: What follows has more technical abbrevations and shorthand than my usual posts here, but I’ve provided some explanatory links, and can go into more detail and explanation in later posts if there’s interest.) The four strategies are:

• Throw all the terms in together, and hope for the best. Sounds like a recipe for chaos, but it is cheap, and may work when users look for both kinds of terms. Ultimately, you’d like to relate the terms somehow (and automated tools may be able to help you find popular terms that are isolated from other terms, so you can relate them to other terms).
• Normalize to a preferred ontology. This may be a good way to go, for instance, when you have a big, controlled ontology you want to use, and a small, less controlled one that doesn’t get much use. The terms in the less controlled ontology can be added as aliases for terms in the controlled ontology, or can be rewritten to fit in with the main ontology.
  
• Make multiple subject maps, link them in appropriate places. This may be appropriate when your maps tend to support different research communities (e.g. MeSH for medical research vs. LCSH for general scholarship), or different kinds of search activities (e.g. folksonomies for current peer awareness vs. LCSH for back-literature searches). In some cases, such as for MeSH and LCSH, existing crosswalks exist that can be used for links between maps.
  
• Build a multi-ontology subject map. This can be challenging, but may be appropriate where you have different collections described through different ontologies that people are likely to want to explore in tandem; for example, a set of books on a particular era of history described in LCSH along with a set of photographs of the same era described in TGM. It can get a bit tricky when the two ontologies have different names for the same concept, though, or identical names that refer to two different concepts.

I hope to try out some of these strategies as we develop subject maps at Penn. (Our Franklin catalog has many subject entries in LCSH and MeSH, for example; and we also have uncontrolled tags that PennTags users assign to items in our catalog. I’d like to see if we can let users browse in useful ways across all three topical spaces.) I’ll post here about any interesting new visible developments and demos.