SYSTEM REINVENTION INITIATIVE
Introduction
The National Library of Medicine (NLM) is the world's largest
research library in a single scientific and professional field.
The Library's holdings today number more than 5 million books, journals,
technical reports, manuscripts, microfilms and pictorial materials.
Housed in the History of Medicine Division is one of the world's
finest collections of old and rare medical texts, manuscripts,
incunabula and images. NLM serves as a national resource for all
U.S. health science libraries, with lending and other services
provided through a National Network of Libraries of Medicine®. More
than 2.5 million interlibrary loan requests are filled through this
Network each year. NLM's Medical Literature Analysis and Retrieval
System (MEDLARS®) was created to provide rapid online access to the
vast store of biomedical information in the collections. The most
widely used of the MEDLARS databases, MEDLINE®, contains 7 million
references dating back to 1966. There are more than 40 files and 16
million records in the full MEDLARS family of databases.
NLM has been providing effective online access to the information in
these files for many years. New directions in software, hardware and
communications now make it possible to provide new functions which
will help users in ways not available under earlier circumstances.
With the System Reinvention Initiative, NLM is seizing the
opportunity to create scalable new systems offering users a range of
enhanced capabilities.
The System Reinvention Initiative has four primary components: the
Access Model, Retrieval Systems, File Generation and Maintenance, and
Integrated Library System. The Access Model consists of all software
systems between a user at a workstation or terminal and an NLM
database system. The Retrieval Systems are the back end retrieval
engines -- multiple systems which operate against different types of
databases. Retrieval Systems receive queries in language they can
understand, and return results. The File Generation and Maintenance
project will create a flexible, powerful environment for building,
editing and maintaining the files the Retrieval Systems run against.
And the Integrated Library System project will use a combination of
off-the-shelf software and custom modules specific to NLM's needs to
encompass the myriad tracking, processing and circulation-related
functions of a great library with collections in the millions of
items.
Central to the System Reinvention planning process is the concept of
increasing availability of Internet access both for information
delivery and to facilitate user surveys and feedback which will help
NLM refine its offerings and its services.
Access Model
The Access Model project is designed to help users find what they
need from NLM's database services. The Access Model includes all the
software tools and systems between an NLM user (which may be a person
at a machine, or a program or intelligent agent that person has
launched) and an NLM database service. An applications gateway at
NLM called the Request Manager accepts incoming data streams, scans
them, translates them into database-specific command sets when
necessary and routes them to appropriate servers running back-end
retrieval systems with the information the user wanted. The Access
Model group is building client software which interacts with the
Request Manager and an Information Sources Map to help direct users
to the right database or set of databases. The Request Manager has
available a Library of Intelligent Search Aids and a series of
Knowledge Sources (chief among them, the Unified Medical Language
System Metathesaurus®) to help users create good searches. A
proof-of-concept Access Model prototype is now running.
Retrieval Systems
ELHILL®, NLM's primary retrieval system for the MEDLARS files, is
exceptionally fast in handling multi-million record bibliographic
citation files and hundreds of simultaneous users. It provides
nearly 600,000 searches to more than 20,000 users each month. ELHILL
and its companion TOXNET® system offer online more than 40 databases
totaling 16 million records. But the nature of the NLM databases is
changing as more full-text systems (Health Services/Technology
Assessment Text, HSTAT), multimedia hypertext systems (HyperDOC), and
image libraries (Online Images from the History of Medicine Division;
the Visible Human project) come on line. New retrieval systems
running on different servers appropriate to their functions --
whether for bibliographic citations, full text, hypertext with images
and video, or 50 billion bytes of images, will be required. The
Retrieval Systems project will identify and buy or build systems
(almost certainly more than one) which can handle the growing variety
of different functions and can be scaled up incrementally as system
File Generation and Maintenance
Before a retrieval system can respond to queries from users, it must
have files to run against. The process of creating those files can
be complex; some may optimized for maintainability, some for speed of
access across large numbers of records, some for in-depth analysis or
editing among small numbers of records. Different file types and
database designs may be appropriate for differing data content such
as full text, large image files, or bibliographic citations. Some
current systems maintain copies of records in multiple files,
requiring their editing in several places when a change or update
must be made. More efficient use of valuable staff time would
suggest a system in which editing a record once is sufficient to
change it in all the places which point to it. The File Generation
and Maintenance project is responsible for creating a flexible,
distributed software environment through which NLM's major files can
be built and maintained.
Integrated Library System
The Integrated Library System project is exploring in detail a
buy-and-build approach in which a commercial off-the-shelf integrated
library software package is augmented with NLM-specific capabilities
and functions. The basic package will have public access catalog,
circulation, serials control, cataloging, acquisitions and authority
control functions which may be augmented as required. Such
additional functions as indexing, DOCLINE®, MeSH® and holdings will be
added by NLM and contractor or vendor staff. This approach is
intended as a fast-track means of bringing up the diverse set of
functions required to replace a group of software systems developed
in-house over more than two decades.
The Challenge
It will be a substantial challenge to maintain transparency to tens
of thousands of users while replacing systems central to the core
functions of the National Library of Medicine. The complex
transition from existing systems and services must occur in parallel
with the development of new ones. A System Reinvention Coordinating
Committee with representatives from all divisions of NLM will focus
on making this transition as smooth as possible while overseeing the
four projects of the System Reinvention Initiative.
NLM HyperDOC / System Reinvention Initiative FactSheet / November 1994