Archives tests e-records

The National Archives and Records Administration is preparing to test the

intelligence of an electronic records management system. It hopes the system

will be smart enough to read documents as they are created and decide which

are important enough to be kept as official records and where and for how

long they should be stored.

How well the test system works will tell archivists a lot about how

soon their plan for a fully electronic archive might be achievable.

Developing an automated archiving system is important to NARA. As the

nation's recordkeeper, the Archives is annually inundated by literally billions

of records created by federal agencies on paper, film, magnetic tape and

in the form of computer files, e-mail and other formats.

The system the Archives is planning to test was created by Provenance

Systems Inc., a company based in Ottawa, Ontario, and Arlington, Va., that

builds automated records management systems.

The system's most important — and most remarkable — feature is artificial

intelligence, which enables it to read and recognize when a document should

be filed as a record. To do so, the system must be smart enough to be "trainable"

by archivists. " "Trainable' in this context means that the system learns

from input from records management experts," said Sam Watkins, director

of the Presidential and Administrative Projects Division of NARA.

Initially, archivists will use the system to examine documents and determine

which ones qualify as records and where they should be stored. Eventually,

through repetition, the system is expected to recognize the elements of

a record, decide which documents fit into that category and determine where

best to store them.

"The system should file automatically once the basic system configuration

is completed," Watkins said. How well it does that "should help us understand

to what extent the system is truly "trainable,' " he said.

The automated system will be tested against NARA employees. Archivists

will compare the results of employee filing with automatic filing to see

whether the two produce different results and, if so, how significant the

differences are.

The Archives "wants to learn about how these applications work so we

can better understand the issues, processes, capabilities and costs of implementing

and using these systems," Watkins said.

"This will be a very small test system," involving fewer than 30 users,

he said. Testing is to begin in October and continue for 10 months.

Eventually, archivists hope to have automated records systems that operate

"in background mode," invisible to government workers who create records.

If such systems can be created, record filing can become automatic.

U.S. Archivist John Carlin told Congress in March that an electronic

archive would take about five years and $130 million to create. The Archives

has already developed a method of "encasing" electronic records in a "digital

wrapper" so that they can be opened and read even if the software used to

create them has since been abandoned. The Archives has also demonstrated

that electronic systems can be created that are able to handle the enormous

volume of records that the federal government creates.