MLibrary 2.0 : Kristin Antelman

Kristin Antelman
“Next Generation Catalogs.” NCSU Librarian. See NCSU Library Catalog. Associate Director of Digital Libraries.

Never quite got the OPAC right. We were very wrong for quite some time. Worse than card catalog, esp. with browsing by subject.
Now: experimenting and questioning traditional frameworks.
Experiments: SUNY index an XML extract; Seattle Public Library; Scriblio
Calhoun: UC Report: questions frameworks, including (but not limited to) LCSH
In NCSU: TWO search boxes?? We’re not ready to let go of authority. Bottom search box uses authority indexes; top one does keyword search. In past: results in keywords: last in first out (roughtly, new books first). HAD to use title as default.
Sample: search art history. 13,000 hits. BUT, LCSH classifications show up at top to help narrow books. Oooh, it answers the questions “where do I go in the stacks for …” With Art History, through, spread all over the place. BUT you can use the browse features to narrow your search, and thus narrowing the call number options at the top.
“Faceted Navigation” works well against MARC metadata – even with wildly broad searches. Allows for relevancy ranking.
EXAMPLE: search java programming. Many books not available – limit to what’s available. ALSO – sort my MOST POPULAR using circulation statistics. Social networking aspects.
“Did you mean” – about 2% - 5% of searches use “Did You Mean” feature.
Can drill down through LCSH, and then use it as RSS feed.
This is what they consider LIBRARY 1.1. Browser search box. Phone search.
Presented data of searching: significant use of these tools.
Limitations example: revolutionary war. 870 results. Missing subdividisions: 3000 items. If you could send that correct sting, you could do that, then drill down. So, there’s still no connection between their natural language and the lcsh.
Faceted navigation “disguises” this problem.
Other experiments:
Phoenix Public Library (http://www.phoenixpubliclibrary.org/ )
State University Libraries of Florida ( http://catalog.fcla.edu/ )
UVa, project blacklight (http://aleph.lib.virginia.edu/blacklight/ )
GA Tech: Communicat ( http://rsinger.library.gatech.edu/papers/thecommunicat.ppt )
Library Thing ( http://librarything.com/ )
Google Book Search ( http://books.google.com/ )
To a system / web person, title and author are not much in the way of an identifier. A number is!
Need 3 pieces of an item: title / isbn / author for a book.
Leads to FRBR.
How do we get our catalogs on the web? Shouldn’t just expose our contents to the web – as soon as many libraries do that, it’s overload.
Answer? Netwroking: be a part of the cload. Practicalities? Faceted navigation: FAST from LCSH.
A Catalog should: recognize clusters of knowledge, popularity, lineage of publications and authors, authoritativeness of sources…
Ben Birschbaeu (sp?) – “People’s Card Catalogs”
Legacies – a powerful cultural barrier (MARC. Current cataloging practices)
Community values – we don’t want to just jettison those (authority, etc)
HOW TO BREAK AWAY: could do much more if our vocabularies were open and extensible. Individual libraries can do a lot.

Q&A

Q: Semantic web – what is it?
A: Can search on meaning rather than just terms. Our OPACS have tried this using this search and narrow, other features in the NCSU catalog.

Q: Licensed content. How do we apply these things to those?
A: Talked about it – no good solution so far from clustering results in metadata in a federated search because of the different fields you get form different vendors. As long as we don’t have control over that data, our hands are tied. It’s sad we got rid of that control a long time ago.

Q: Find something in their search results – is there a “find similar”?
A: Yes – recommender system. UCal explored this. We haven’t done it yet (staffing), but it’s a good diea. How much of a recommender system can you build from these datas? What about other user behavior? We’re not tracking that information. Privacy concerns. Also: aggregating. What’s the popularity in the whole world?

COMMENT: 1/3 of keyword searches match LC subject headings. How can we convince OCLC to let us work with the LCSH? … … … Where’ve we been? The user is the focus.