A good digital library interface is one that supports its users information seeking behavior. Think about a physical library; its “interface” – stacks, catalogs, and the Dewey decimal system – aren’t too friendly in some ways. A person looking for information on a subject (if they can’t find a librarian) has to search a catalog, write down call numbers, look at a map of the library and then check the shelves. If the book isn’t there, there’s no easy way to say, “show me some good substitutes”, or to check the availability of some other related book, whose call number isn’t known.
The power of digital library interfaces is that they build on top of digital catalogs that can make all kinds of information instantly accessible. Their role is to allow people to look up and get to the information that they are actually seeking.
So the next question: how do people actually seek information in a digital environment? Many researchers who study this field think that they follow a cycle: formulate a query or a question, figure out an action that gives an answer, review the results, and then reformulate.
An example of this cycle is the way we might use a search engine like Google. Say you’re visiting San Francisco and want to know where to get a good lunch – that would be your question. You might type “lunch near Mission district” into the search box and click “Search” – that would be the action. After looking at the results and reading a few, you might decide you need to re-word your question “sandwiches near Mission district, SF”, or you might find an interesting looking restaurant and then search for its name to get reviews – that would be the reformulation step, and so on.
Another principle is that recognition is easier than remembering; it’s usually easier to recognize something when presented with its name, picture, or other description than it is to think up a description of that very same thing. This is why hierarchical browsing structures can be so useful – they present users with places to start looking and also ways to successively narrow down and filter out the set of items they’re looking at.
If I’m looking for information on a 17th Century building my professor mentioned in an architecture class but don’t quite remember its exact name or who designed it, I might still be able to find it by browsing images within a hierarchy, such as the one shown above from the UC Berkeley Architecture Slides Then I can search within those results for anything else I might remember. Additionally, if I saw pictures, a list of architects, places, or styles, I might be able to immediately recognize it, even though I would have never remembered it on my own.
A good library interface must therefore support search and browsing in a way that allows a user to overview a set of items, filter the set according to some dimensions, and “zoom in” on an item that’s interesting for more details.
Here’s a list of some digital libraries whose interfaces have great support for all these kinds of actions (more on why I think they’re great here):
- The VirgoBeta catalog interface at the University of Virginia Libraries
- SearchWorks at the Stanford University Libraries
- The Historical State Collection at North Carolina State University
- Lens Search at the University of Chicago Libraries
- OneSearch at the Fondren Library at Rice University
- The Santa Cruz Public Library
Except for WorldCat’s, they’re all implemented using “next generation catalog” software, the most widely-known of which are Project Blacklight (which powers VirgoBeta, Searchworks and The Historical State), and AquaBrowser (which powers LensSearch shown above, OneSearch, and the Santa Cruz Public Library among others). WorldCat not only developed their own front end but even published a usability report [PDF] on their findings during the process.
Using this software usually involves importing an entire library catalog into a database, which can take more or less work depending on whether such a digital catalog already exists in your library. While it may still seem like a gargantuan task, library patrons will immediately start to feel the benefits of faceted browsing and collection overviews that are fully integrated with filtering and searching.
Aditi S. Muralidharan is second-year Ph.D. student with Professor Marti Hearst in the Department of Computer Science at UC Berkeley. Her interests are in the general field of text mining and user interfaces – extracting information about entities, relationships, and trends from text. This involves natural language processing, machine learning, data mining, and information retrieval. She is especially interested in building good user interfaces for search, browsing, and learning, with a focus on the process of iteratively and selectively expanding a searcher’s knowledge of a subject.
You don’t have to be just a Libtechtalk groupie. Did you know that this blog is looking for guest authors? Contact ctomlinson at towson.edu to find out how you too can write about your favorite technologies and how they might be used in academic libraries.