Navigating without a map: discovery in a world beyond subject access
What do you do when your library has a unique collection, like ETDs -- but it unfortunately does not have high-quality subject metadata? How can you make that content discoverable and (dare I say) engaging?
In this talk, we'll show a machine learning approach that situates a collection within a conceptual space. We'll then explore that space, using a variety of visual and textual approaches, and see how it can be the back end for possibilities as diverse as...
- Visualizing a whole collection at a glance, and seeing stories in it
- A recommendation engine (if you like X, try reading Y!)
- A way to find people with shared interests
- A quick-and-dirty literature review
