Page 218 - Building Digital Libraries

Thinking about Discovery

is a discovery system? Well, it depends on the type of system being evalu-
ated. Discovery systems tend to fall into one of three categories of systems:
federated systems, hybrid systems, and managed systems.
Federated search systems provide a normalized method for search-
ing multiple databases through a single query. It’s important to note that
conceptually, federated search systems have been available for a very long
time. Since the late 1990s, a number of federated search systems have been
used within the library community and outside the library community. 4
3
Federated search tools like OCLC’s SiteSearch and search engine metasearch
tools like Metacrawler are good examples of these early federated search
tools. These tools served as a searching portal, allowing users to query a
large number of resources through a single interface.
In figure 9.1, we see a diagram of a traditional federated search system.
These systems utilize a single query form that sits on top of the actual fed-
erated search engine. This engine handles the actual communication with
the various databases to be queried. Moreover, this engine will tradition-
ally handle tasks related to normalizing the resulting information from the
various databases. This would include handling tasks like sorting, merging,
and deduping the results from various databases. Today, this diagram has
changed slightly. With the advent of OAI (Open Archive Initiative) and FIGURE 9.1
other metadata-harvesting protocols, many federated search systems have Federated Search Diagram
become hybrid search systems—harvesting, normalizing, and locally index-
ing metadata for some systems—while maintaining the broadcasting search
components for resources that cannot be harvested.
Hybrid federated search systems, as diagrammed in figure 9.2,
utilize a local data store to improve indexing and response time.
These systems use a just-in-case philosophy and harvest, index,
and normalize metadata from a set of diverse databases prior to
the user’s query, in much the same way that a web search engine
crawls and indexes the Web. Today, many libraries use the hybrid
search method when creating “bento”-style discovery systems.
These tools mix locally indexed content with content retrieved via
APIs to generate a contextual-based discovery experience. This
approach is somewhat unique to the library community, in that it
seeks to replicate the different silos of library content, recognizing
that library content is very difficult to surface without some context
placed around the results. This approach attempts to provide that
context by shifting the results paradigm from a long list of results
to content buckets that are separated by the type of material being
queried.
Managed systems work by indexing all content into a central
index, as diagrammed in figure 9.3. Unlike federated or hybrid
systems, managed systems harvest and index all content locally,

FIGURE 9.2
Hybrid Federated Search Diagram

203