Making Federated Search More Usable

Making Federated Searching More Usable
Frank Cervone
Assistant University Librarian for Information Technology
Northwestern University
The Education Institute
Tech Tuesdays: Talking with Techies #14
Agenda
• What are some of the key issues related to
federated searching that could affect
usability?
• What have we learned from usability
research that has been done already?
– What information should be included on the
result screen of a federated search?
– What should you consider when developing
specific subject area groupings?
– Where should the federated search box be placed
on a home page?
© 2006 Frank Cervone
A fundamental issue
• Federated search engine concept is still
new to most people
• Don’t have a mental model for it
– People understand searching Google
– They understand searching individual
databases
• Most people have no clue what
“metasearch” means
© 2006 Frank Cervone
While it may be “old hat” for some of us…
• Federated search is still relatively new
– For the majority, still in the early stages of
implementation or planning
• Federated search can be complicated
– Offering services as a seamlessly
integrated whole
– How does architecture of the software
influence service implementation?
© 2006 Frank Cervone
Component integration
• Tight vs. loose integration
– Integration with other services
• Advantages of tight integration
– “Big bang” implementation
• Advantages of looser integration
– Staged introduction of the services
© 2006 Frank Cervone
Key questions - Authentication and
authorization
• Support for external authentication
– LDAP, Active Directory
• How does it interact with an
authentication service to determine
authorization to services?
– Allow only students in grades 6-8
– Deny access from all campuses except
Main
© 2006 Frank Cervone
Key questions - Interface customization
• Customized searches for specific
populations?
– Undergrads, Med school, specific branches?
• Can user interfaces be customized at a
departmental, divisional, institutional level?
• How are “collections” of databases defined?
• Can users create their own collection of
databases?
– Can individual users save “preferences”?
– Number of items returned
– Preferred databases
© 2006 Frank Cervone
Issues related to usability
• Issues don’t differ much regardless of
model
• Federated search usability has many of
the same characteristics of usability
testing on the main web site of a
library
• Certainly, similar goals:
– Ensure that people are able to find and
use resources in the most efficient and
straightforward manner possible.
© 2006 Frank Cervone
Upfront work
• Customize OpenURL resolver’s
“target menu” page
• Some things to consider
–Limit number of sources to display
–Display alternative services
•Link to catalog
•Link to ILL
•Citation download
•Online chat
© 2006 Frank Cervone
© 2006 Frank Cervone
Our testing program
• Program of usability testing in
place for well over three years
• In that time, we have tested
various aspects of our web site
and our library catalog.
• Each iteration feeds into the next,
so we are able to incrementally
improve the interface over time
© 2006 Frank Cervone
Introducing federated searching
• We decided we wanted this to be a “named”
service
• But what to call it?
– Quick tour of the federated search engine, by
way of a usability testing question
– Asked them how they would describe what the
search engine is
• None of the test participants spontaneously
used a “standard term”
– metasearch, federated searching, megasearch
© 2006 Frank Cervone
So what happened?
• Described it in terms of our current resource
finder
– Often with extra superlative adjectives or
descriptors
• We knew from prior testing that names for
services did not, in general, stick with our
undergraduates
– “NUcat” and “Answers Online”
• We’re not sure why this is
– But we were able to see the blank stares when
we named them
© 2006 Frank Cervone
Preference testing
• Consequently, we did a preference test
• Tested out the working name for the
project
– What did they think of “Einstein”?
• Surprisingly, they liked it
• As one student described it
“Well, it’s not clear what it is, but now that I
know I’ll remember it. I mean, Google
doesn’t make a whole lot of sense as a
name either.”
© 2006 Frank Cervone
What is the point?
1. When in doubt, perform a preference test
and find out what resonates with people
2. “Common wisdom” is sometime wrong
– Many usability studies have dismissed FAQ
pages as wastes of time
• But this observation depends on context
• For a site that is well-known, it is true that FAQs truly
are rare
• However, our testing showed that this new service truly
did have a set of FAQs!
3. The last point of the story is more complex
© 2006 Frank Cervone
A simple and streamlined interface?
• What’s the goal of a federated search
engine?
– Make it simple!
– Make it like Google?
• The federated search model is different
– Actually quite different from Google
• Database? Subject Area?
• Google does not provide the same mentalmodel
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
Are strict subject categories a good
thing?
• They do help filter out some of the “noise”
from a less-than-specific search
• However, what if the searcher does not know
what the subject area is?
• What if we have not defined a subject area
that is logical from the perspective of the
patron?
• What about interdisciplinary studies?
• Strict subject categories could be an
impediment
© 2006 Frank Cervone
The answers are sometimes surprising…
• Defy the “conventional” wisdom we
may have about the best way to
approach an information retrieval task
• Usability testing of federated search
products at Northwestern and other
large research universities has
demonstrated some of the interesting
conundrums that federated search
engines raises
© 2006 Frank Cervone
For example…
• Design point - needed to be a very
simplified interface designed to appeal
to undergraduates
• Core group of databases
– Would cover the broad range of programs
at the university
– No need to know about a specific subject
area or even about specific databases
© 2006 Frank Cervone
Conflict with other goals…
• This conflicts with desire of librarians to
teach about using databases
– Which is most appropriate for a given topic
• Several factors in the decision
– Faculty and other researchers do not care about
“databases”
• Faculty are concerned with the source of the material,
that is the journal itself
• Databases change configurations regularly anyway
– FirstSearch to ProQuest
– Web of Science to SCOPUS
– Instruction is not something that can be forced
on a web visitor
© 2006 Frank Cervone
What does the research say?
• Several usability studies indicate a
clear preference for “simple search”
– Does not require the user to select a
database
– Does not provide any sort of “complex”
query statements
– User defaults to a group of databases if no
group is specifically selected
• Default typically provides broad
interdisciplinary coverage
• Works for specific areas as well
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
The advantages of quick search
• Users feel they are very successful finding resources
• Good place to start when they are beginning to
research their topic
• Unexpected advantage: includes areas that may not
be self-evident
– Graduate students and faculty often find many significant
citations they were not previously aware of
– Most participants in these tests were only familiar with a
very limited number of databases in their area
– They rarely searched in other databases outside their realm
of expertise
– This is similar to earlier studies on searching behavior
© 2006 Frank Cervone
Why?
• People become accustomed to searching in
specific native interfaces
– Are not especially interested in learning others
– Particularly when the benefits are not clear
• Researchers claim their behaviors are
different
– This is not true
– There is a surprising amount of similarity among
all types of users
© 2006 Frank Cervone
Conflict!
• Students (graduate and undergraduate)
have clear expectations about how results
should be displayed
– In relevance order
• Even when it’s not clear how relevance is determined
• Direct conflict with traditional approach
– Alphabetical within reverse chronological order
• But by forcing the traditional ordering, we
create a disconnect
© 2006 Frank Cervone
You will do as you are told…
• As a result, at NU we have decided to
display results in relevance order
• Other institutions have taken other
approaches
– Patrons are able to figure out the display order
fairly quickly
• If it is alphabetical within reverse chronological
– But they do not like it
– When they have the opportunity, they change the
order to relevance
© 2006 Frank Cervone
However, this is not without problems
• How is relevance determined?
– Federated search engine
– Individual data bases
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
© 2006 Frank Cervone
Looking for articles
• Novice researchers “fish” for articles
• Advanced researchers tend to have specific articles
in mind
– Advanced researchers say they tend to focus on databases
for these types of searches
– But when presented with a federated search, this behavior
cannot be reliably observed
• The vast majority of library patron communities,
when given the choice, prefer starting at a “Quick
Search”
• In fact, when presented with a list that allows them
to “find a database,” test participants generally find
it confusing
© 2006 Frank Cervone
© 2006 Frank Cervone
“I’d just quit and go to ProQuest”
• Because it was one of a few databases the
student knew the name of
• Seems to “provide good articles most of the
time”
• This type of response, although perhaps
articulated differently, was common
© 2006 Frank Cervone
Broad vs. focused coverage
• Delicate balancing act we must do if our web
sites are to be successful
• Finding the right grouping of databases for
“subject areas” or “interests” can be a major
problem in setting up the federated search
engine
• The natural desire to set up long lists of
items to provide comprehensive coverage
may be working against the library’s best
interest
© 2006 Frank Cervone
Too much leads to too little
• Often long lists of resources are developed
in the hope of promoting awareness of the
various resources a library purchases
• Hope that people will click through to the
native interface when it is appropriate
– In usability tests, there is no evidence that
creating these long lists does, in fact, enhance
awareness
• Students often found these long lists of
resources confusing
– “Those long lists really make me feel stupid.”
© 2006 Frank Cervone
If it’s not federatable, it’s ignored
• In general, people don’t use non-federatable
databases from within a federated search
interface
– They simply ignore these databases
• One exception is for a well-known resource
people expect to see, even if it’s not
federatable
– Hein Online or Lexis Nexis
• Alleviate the problem by having a “best
choices” group of 3 databases in each area
– Provide the rest as a secondary grouping
afterwards
© 2006 Frank Cervone
Results list should be informative
• Streamlined result lists can be
confusing
• Result lists work better when a more
“traditional” citation format is used
© 2006 Frank Cervone
© 2006 Frank Cervone
Custom combinations of databases
• Functionality for combining databases in
unique ways geared toward the specific
interests of the researcher
• A clear overriding theme: patrons find this
confusing and avoid using it for the most
part
• Default view for combining databases is
based on a subject classification model
• For people working in multidisciplinary
areas, this is not usable
© 2006 Frank Cervone
© 2006 Frank Cervone
We can’t fix that …
• Federated search software cannot be easily
customized to address this
• An important issue in usability testing sometimes, we cannot immediately “fix the
problem.”
• Solving usability issues often can only be
accomplished through a partnership of
working with our vendors to design solutions
that will work better for our patrons
© 2006 Frank Cervone
Some emerging best practices
• Predefine collections of databases
– Broad subject areas
• At least at first
• Put must frequently used databases at
the top of lists
• If possible, expose descriptive
information about databases
© 2006 Frank Cervone
Resource definition considerations
• Have correct technical information
– Is the Z39.50 configuration “up to spec”?
• Categorize resources early in the process
– Otherwise you’ll have to retouch every database record
– Kick start the process
• Appoint a small group to make the first pass at each area
– Clearly define extent of coverage
• What databases will be included in each collection?
• Will there be a maximum number?
– Think about performance characteristics
 Differences in results among databases
 Do these databases “play well” together?
 If not, what do you do?
© 2006 Frank Cervone
In summary
• OpenURL and federated searching are important
new services that we are bring to our patrons
• It is critical for us to make sure that these services
work from the perspective of the patron
• They represent extensions of the library that further
enable self-service information provision, which will
be critical to the library of the future
• Although usability testing in this area may be in its
infancy, it is critical that we vigorously explore this
area. For while we know some things about how
federated search works, there is still a lot more for
us to learn and our future depends on it
© 2006 Frank Cervone
Thanks!
Frank Cervone
Assistant University Librarian for
Information Technology
Northwestern University
Evanston, IL 60208
[email protected]
© 2006 Frank Cervone