Netlib currently provides a search interface
that allows the user to do field-specific keyword searching.
For example, the user may search by author, filename, abstract keywords,
or GAMS classification. Search results are returned as a hypertext
list of catalog records from which the user may select files to view
or download.
Searching by keywords or classification codes often returns a large number
of search hits, leaving the user to sort through them.
Further discrimination often cannot be provided by an overall
classification scheme, but requires use of a domain specific
knowledge base. Such knowledge bases have been constructed
for specialized domains, including
differential equations [4][27][29]
and approximation [23]. We are experimenting
with providing on-line hypertext interfaces to such knowledge
bases. For example, we have provided a hypertext version of
a decision tree for approximation algorithms
.
We have also developed a prototype expert help system to assist users in selecting software within specific domains [8][7]. An advisory system for a given problem class helps the user discriminate between problem-solving software modules for that class. For a given problem class, a set of features are partitioned into a small set of feature classes, and information is encoded about how each feature applies to each software module. Prototype user interfaces have been developed that allow the user to interact with choice widgets for the various features. The system provides more specific and effective help in selecting software than a domain-independent search interface.