| WilsonWeb —
Technical Overview |
|
WilsonWeb
is a sophisticated Internet-enabled, multi-tiered application
for the search, retrieval, and management of records from a
vast, and expanding, rich data repository covering numerous
disciplines. The bibliographic and biographic material includes
citations, abstracts, full-text journal articles, illustrations,
photographs, and multimedia files. This state-of-the-art
subscription service, custom-built for HW Wilson, provides information to users worldwide:
casual library visitors, librarians, students, researchers,
government workers, and corporate professionals.
The
WilsonWeb search engine was built using a combination of ATG
Dynamo (Java) and JavaScript (for the front end). Reference material
(XML) is loaded and
updated into approximately 20 - 25 Verity collections. Complex
queries (often rules based) are generated based on user input
to forms. Results are displayed in various formats, modified
to user specification using XSLT (eXtensible Style Language
Transforms) against XML (eXtensible Markup Language) data.
All subscription
and authentication data is stored in an Oracle repository
(seeded and updated from mainframe feeds). The underlying
data is stored in Verity collections which are spread across
multiple servers. These collections are optimized so that each
record is stored once, but can have different looks for each
of the products that it belongs to. A load process(Perl, XSL)
was developed in which the data loaded is transformed following
complex user rules. The load process is spread across 9 Solaris
machines (will run on NT and will scale to the size of a complex).
This process splits, re-formats, and loads approximately 12
million records and documents into Verity collections. The same
input files are used to build Browse lists (coded in C with
AVL algorithms for speed and efficiency) where users can look
through word and phrase lists. Thesaurus
and Browse functions have been provided to assist the user with
drilling down to the desired results. There are also update processes
for these collections and Browse lists. An administrative
interface exists for authorized users to customize the system's
look and feel (text, titles, toolbars, graphics), and where
features can be added or removed for certain sets of users at subscribing
institutions (universities, corporations). The Help system is also
configurable to reflect user customization.
An eCommerce feature will be available in the Fall (2002), where users can purchase
articles that are not in their subscription.
|