Knowledge base sharing is becoming more and more important for CVRS. VRS can be
supported by the knowledge base, which stores previously asked questions and
answers and can benefit the next patron who asks the same or similar question because
s/he can be provided with a quick, relevant, and high-quality answer. If knowledge
bases can be shared efficiently, patrons may get answers from knowledge bases before
their questions are actually submitted (Jin et al., 2005). Knowledge base sharing can be
achieved in two ways:
(1) The first possible solution is real-time access to distributed knowledge bases of
member libraries. This is simple, direct and can retrieve the latest records. But it
is difficult for patrons to decide which knowledge base they should access to get
the information they need. And every knowledge base may be primarily a key
reference source customized to one library and its patrons’ needs because of
specialized local information and questions unique to local patrons. Thus there
is no doubt that it would be quite time-consuming to access distributed
knowledge bases and much unwanted local information would be returned
among the search results.
(2) Another solution to the knowledge base sharing is to build a central knowledge
base. Records in the central knowledge base are from the knowledge bases of all
member libraries and can be shared by them. Since records should be checked
for duplication and quality before they are added to the central knowledge base,
to access this single knowledge base can obviously improve the search
performance. Accessing distributed knowledge bases should happen only when
local information is needed. The key issue of this solution is how to harvest
records from distributed knowledge bases into the central knowledge base.
Recently most similar applications would adopt OAI-PMH to achieve metadata
harvesting from distributed databases. The OAI-PMH, commonly referred to as
the OAI protocol, is a client-server protocol layered over HTTP that is used to
transfer metadata records with mechanisms for periodic and incremental
updating (Suleman and Fox, 2002).