Frank Cervone from Northwestern spoke about implementing federated searching. He also spoke about this last year (link to notes)

He admits to trying to make it Google-like (the horror!), trying to help interdisciplinary researchers as the databases tend toward silos of information, and trying to remove the reliance on vendor specific interfaces.

Now that federated interfaces are in place only 10% of users will use both the federated/simplified, only 5% will use the database-specific interface. Of the 5%, many are librarians.

User expectations

Search strategies:

Keywords

Author’s last name

Students almost never use the refine features

Faculty only use certain ones

Context Sensitive Linking

From a citation list – takes you to full text (via SFX?)

You can have it check consortium catalogs and request ILL

What happens when you do metasearch- there are so many variables

You don’t always get the same results back, sometimes just “a solution exists” sometimes the full record with link out to full text

The main protocol that is used is still Z39.50.

• Although this protocol is very complex and allows a ton of different fields, most database providers only allow 7.
• Requires a lot of back and forth
• Returns only small sets of results (makes de-duping difficult)
• Investigating XML -- A flavor of this is SRU/SRW

He lists 7 vendors of the service right now.

How do you pick a service?

• How does it support external authentication? LDAP, Active Directory
• Custom interfaces for specific populations? (he hasn’t done a lot of customization because user feedback hasn’t really supported the work required for it) How customizable is it in general?
• Can you go directly into results or do you have to stop at a screen and see a list of databases? Can you merge results from the various databases? How does relevance ranking work (on vendor said everything is alphabetical – no choices)
• Can individuals save interface settings (number of results, favorite databases)

Rollout, Acceptance – most of the resistance comes from within the library, it’s not the be-all end-all, but users accept this and don’t expect perfection.

¶ 10:30 AM| |cites (technorati) |