Page 221 - Building Digital Libraries

CHAPTER 9

local discovery system that works hand in hand with the traditional search
engine systems. By producing content that will work in both, organizations
can improve their chances of being indexed and optimize their content for
ease of discovery.

Current Research

While the concept of federated searching and discovery systems is not new,
active participation by libraries in the creation of new ways of displaying
and conceptualizing the user’s experience with a discovery system is a rela-
tively recent phenomenon . . . and we’d argue a welcome one. Prior to the
entry of library-vended discovery systems, cultural heritage institutions led
the way in developing new ways of visualizing the search and retrieval of
heterogeneous content. But around 2006, this changed; content providers
like EBSCOhost and ProQuest developed large managed discovery tools
that worked directly with academic content providers to enable libraries
to do full-text searching of nearly all of their vended content. These tools
represented a leap forward, and essentially stopped much of the homegrown
development and research that had been occurring around discovery system
development.
Interestingly, this started to change around 2010 when North Carolina
State University introduced users to its bento search experience. Leveraging
both vended and local data stores, NC State challenged the current paradigm
by providing results in contextual buckets rather than in a long list . . . and
the approach worked. Since that point, cultural heritage institutions have
reentered the fray, doing new and exciting research around results manage-
ment, linked data usage, language agents, and much more.

Recommender/collaborative filtering
Current federated search services query preselected or user-
selected groups of information. This assumes that the user
community making use of a particular tool consists of expert
users or others who are familiar with the resources being que-
ried. One of the strengths of federated search is the ability to
serendipitously discover information within larger datasets—
something that the current crop of federated search tools does
poorly. Within the federated search community, there is a
current drive to understand how the collaborative filtering of
results and databases can lead to a better understanding of how
target databases can be transparently selected or recommended
to the user based on query terms.

Deduplication of results
Most of the current generation of federated search tools cur-
rently provide some form of deduplication of items from a

206