1. INTRODUCTION
In recent years, as more library collections are available
electronically, libraries began to adopt discovery tools that
are designed to be “one-stop” search platform for a wide
range of library collections and resources. Discovery tools
are web-based applications that search in a unified index of
metadata from article databases, library catalogs, digital repositories,
digital collections, and other scholarly information
resources (Fagan, Mandernach, Nelson, Paulo, & Saunders,
2012; Williams & Foster, 2011). The unified metadata consist
of facets, which Wynar and Taylor (1992) defined as mutually
exclusive and collectively exhaustive properties of information
items (e.g., books, journals, articles, etc.). With the utilization
of faceted browsing and searching, users can achieve higher
task accuracy and satisfaction than traditional direct search (Yeh
& Liu, 2011). Because of these potential advantages of discovery
tools over traditional library catalogs, the number of
libraries in a sample of 260 academic libraries in the United
States and Canada employing discovery tools has doubled from
2010 to 2012, increasing from 16% to 29% (Hofmann & Yang,
2012). Following the implementation of discovery tools, there
has been a growing interest of research among libraries on user
search activities and task performance changes for assessing the
effectiveness of discovery tools.
Purdue University Libraries (the Libraries hereafter) implemented
VuFind (Figure 1) to replace the traditional OPAC
(online public access catalog) in 2009. VuFind allows users
to search and browse records in the library catalog and institutional
repositories. In October 2012, the Libraries made
a move to replace VuFind with a new discovery tool, Ex
Libris PrimoTM (Figure 1; Primo hereafter), aiming at providing
additional coverage of libraries’ subscribed databases
and online journals. The decision to implement Primo was
largely due to its seamless integration with the existing library
management system. During the testing period from October
to December 2012, both VuFind and Primo interfaces were
presented on the library’s home page in order to make the
transition easier for users. Note that the “Articles” search tab
shown in Figure 1 was based on a legacy search tool that
is not part of VuFind or Primo. The testing period provided
a unique opportunity to contrast user activities of both systems.
Primo provides a single entry point for the majority of
library resources, including catalog records, institutional repositories,
databases, and online subscribed journals/magazines
(some electronic resources are not available in Primo search),
whereas VuFind is mainly for the library catalog and institutional
repositories (Figure 2). There are also differences of
user interface between these two systems due to the coverage
difference.
The goal of this study was to understand user search activities
with both VuFind and Primo in a unique time window
where both tools were available for the library searchers.
We combined transaction log analysis and user testing, which
has been rarely used in combination in studies on evaluating
search systems, especially for discovery tools. Basic search
activities such as search field selections, facet usage were
examined. We also conducted a search query analysis and
investigated the query formulation and reformulation strategies
to further understand searchers’ behavior and to provide
guidance for future discovery tool design and implementation
efforts.
422